Semester projects

Total: 263

A detailed evaluation of the Via Sicura program
This project aims at updating and improving the official evaluation of the Via Sicura road safety program.
Student: Mya Jamal Lahjouji (SGC), June 07, 2023
Supervision: Nicola Ortelli, Selin Atac, Michel Bierlaire
Targeting population to evaluate COVID-19 strategies based on activity-travel behavior.
Countries everywhere in the world have been adopting different policies to lower the toll of the SARS-CoV-2 pandemic. Epidemiological and behavioral models have been widely used to guide decision-making. In this project, we aim at designing a methodology to optimize targeted COVID-19 policies. We will use a simulator provided by the TRANSP-OR lab. This simulator draws the probability of an agent getting infected given its socio-economic characteristics. We hope that this approach will provide insight on transmission and intervention that will complete what can be obtained with usual compartmental models (SIR). We will use health data (Sentinella or SARS-CoV-2) from FOPH. We will review different strategies to determine the trade-off between pharmaceutical interventions (vaccination) and non-pharmaceutical interventions. We will design these policies by accounting for socioeconomic variables of the population, such as age, gender, home address, and general health evaluation, and determine the optimal strategies according to different criteria.
Student: Antoine Goupil De Bouill, June 03, 2023
Supervision: Cloe Cortes Balcells, Tom Haering, Michel Bierlaire
Predicting activity schedules of the population considering non-pharmaceutical interventions
COVID-19 creates an undeniable bridge between the epidemiological and the transportation community. Due to the lack of a vaccine in the first stage of the pandemic, controlling mobility becomes the key to containing the spreading of the virus. Understanding how and why people perform their daily activities provides crucial information to define more targeted and less disruptive interventions. However, the link between mobility and the spreading of an infectious disease is double-edged: how people move drives the spreading of the virus, and the presence of the virus has clear repercussions on people scheduling their activities. These reschedulings can be due to a self-choice (I no longer go to the gym because I am scared), or because an external imposes it (the government closes the gyms). In this project, we will focus on the latter since we do not own data to calibrate the parameters of self-choice reschedulings. The aim is to predict the activity schedules of the population considering restrictions derived from COVID-19 presence. We will use an agent-based model provided by the TRANSP-OR lab (OASIS) which integrates the different daily scheduling choice dimensions (activity participation, location, schedule, duration, and transportation mode) into a single optimization problem. We hope this approach will provide insight into how the presence of infectious diseases impacts activity-travel behavior and reflect on potential policy implications/interventions.
Student: Arthur Nussbaumer, June 03, 2023
Supervision: Cloe Cortes Balcells, Evangelos Paschalidis, Michel Bierlaire
Optimisation et Dploiement dune mthode de calculation au sein dun groupe actif dans le second uvre
Student: Thomas Poulain, June 02, 2023
Supervision: Michel Bierlaire, Bastien Sauve, Phida Etanchit SA
Scenario generation for the multi-agent transport simulation toolkit, MATSim
CSSs are becoming increasingly popular, primarily due to their financial and environmental advantages. However, they face many operational challenges, including inventory management of vehicles and parking spots, vehicle load balancing and redistribution, pricing strategies, and demand forecasting. If these challenges are not addressed properly, the CSS risks experiencing a significant loss of customers and therefore revenue. The literature consists of variety of works on CSSs especially focused on the rebalancing operations. On the other hand, it lacks the detailed comparison of different strategies of rebalancing operations at a disaggregate level. One approach to fill this gap is using an agent-based transport simulation. Such simulation toolkits allow analysis at the individual level and testing different configurations such as city and network structure. Therefore, in this study the student will review the literature to determine the main characteristics of the CSSs. In the first part of the project, she will work on creating representative scenarios for the multi-agent transport simulation toolkit, MATSim, based on the findings in the literature. These scenarios will be used to compare different rebalancing operations strategies where an optimization module is used. The second part of the project will be implementing analysis scripts that will provide useful information on the simulation results.
Student: Anne-Valerie Preto (SGC), January 27, 2023
Supervision: Selin Atac, Nicola Ortelli, Michel Bierlaire
Comparison of machine learning techniques for synthetic activity generation
This project aims to investigate the good and the bad sides of state-of-the-art Machine Learning techniques applied in the context of synthetic population generation, focusing on the generation of attributes that de- scribe synthetic activities. Recently, there has been an increase in research focus on the application of deep generative learning algorithms for generating synthetic populations and travel behaviour modelling, including Variational Auto-Encoders (VAEs) and Generative Adversarial Networks (GANs) (Borysov et al., 2019, Wong and Farooq, 2020). Whilst initial experiments show promising results, there are still significant drawbacks with using generative modelling on population synthesis, which includes vanishing gradients, training instability or mode collapse, nor it allows for exact evaluation of the probability density of new points. Based on the literature review, we try to propose possible improvements for existing methodologies by identifying if there is a way to capture better the correlations between variables. One of the techniques that could be improved is Generative Adversarial Networks (GANs). For instance, DATGAN (Lederrey et al., 2022) is a GAN architecture that tries to give more control to the modellers when generating synthetic tabular data. It uses a Directed Acyclic Graph (DAG) to represent the correlation between variables in the dataset. This DAG is then used to define the structure of the Generator. Each variable is represented by an LSTM cell. However, due to the linear nature of LSTM cells, it is required to have a DAG instead of an acyclic graph. This directionality can be an issue when designing the DAG since it represents causality between variables. Thus, we could improve the DATGAN by removing the directionality and only specifying correlations between variables. Graph Neural Networks (GNNs) are a specific type of NN designed to mimic any type of graph, and they can be used to design the architecture for the Generator. This change would give more freedom to the user and make the method more flexible since DATGAN would not require any specific type of graph
Student: Xinling Li (SGC), January 27, 2023
Supervision: Marija Kukic, Nicola Ortelli, Michel Bierlaire
Synthetic generation of activity motifs
This project aims to work on the extension of the existing simulation framework (Kukic et al., 2021). In the existing methodology, we developed a new approach so-called Divide and Conquer Gibbs Sampler, for the generation of synthetic households. The current version of the framework can generate socio-demographic attributes of individuals (such as gender and age) grouped into households (described by their own set of features) in a realistic and representative way. However, in ABMs, there is a need to generate activities, not only socio-demographic characteristics. Some of the features that describe activities are type, location, start time, etc. However, as shown in the past, the simulation methods struggle to deliver accurate results in a reasonable time while dealing with high-dimensional datasets. According to that, instead of gener- ating disaggregated data on individuals activities, it would be better to generate aggregated data while preserving the same amount of information. To do so, we will generate synthetic activity motifs. The daily activity motifs are representative graphs where the nodes describe the activity type and location, and the links indicate the sequential order between activities. The advantage of generating activity motifs instead of disaggregated activities is that we can reveal daily mobility patterns while saving computational time and preserving privacy. The first step is to retrieve and replicate the real activity motifs at the level of individu- als. The additional step would be to investigate how to generate the activity motifs of households depending on activities that individuals performed, subject to household constraints (e.g. sharing of resources). Note that the work of pre-study should be focused on the generation of activities that describe the movements of individuals. The investigation of activity motifs at the level of households can be part of the master thesis.
Student: Bochud Quentin Philippe (SGC), January 27, 2023
Supervision: Marija Kukic, Janody Pougala, Michel Bierlaire
Modelling migration intentions worldwide
Migration across different countries is a major topic of policy debates worldwide that receives relatively little attention in the travel behaviour literature. In industrialized countries, the proportion of immigrants in the total population has increased in recent years resulting in concerns about integration, loss of national identity and economic expenses. In low-opportunity countries, international migration raises fears of loss of highly skilled workers, who are the group most likely to emigrate. Therefore, understanding the main factors that influence the location where migrants choose to settle is of primary interest for both researchers and policy makers. Location choices might be influenced by changes in the attractiveness of the country of origin, of the country of desstination, and of the alternative destinations. A recent study developed a cross-nested logit model to predict migration intentions in India based on characteristics of the individual, characteristics of the destination country and unobserved factors shared between similar destinations. However, there is still a dearth in understanding the main factors influencing migration intentions at a global scale, in countries that differ in terms of size, level of development and willingness to emigrate. The objective of this study is to assess the transferability of the previous model predicting migration intentions in India to other countries in the world. The study will identify the main characteristics of the countries of origin that have an impact on the migration intentions. The data were extracted from the Gallup World Poll (GWP) Survey. The GWP measures a series of key indicators including migration intentions and is one of the most comprehensive datasets available worldwide. Other key indicators measured in the survey are national security, nutrition, housing, job opportunities, financial situation, personal health, civil engagement and overall well-being. The GWP surveys cover more than 150 countries in the world, are usually repeated every year, and include 1000 individuals in each country. In this study, the GWP data will be analysed to select relevant countries based on the coverage, the number of waves available, the respondents per wave, and the variables available. The analysis will identify patterns per country and evolution of migration intentions over time. Mathematical methods will be implemented to understand which variables should be included to improve the model specification and to capture the perceptions of respondents about the living conditions in their country. These methods will be also used to explore the complex correlation patterns between the attributes and the destination choices.
Student: Hilda Abigl Horvth (SMA), January 27, 2023
Supervision: Silvia Varotto, Nicola Ortelli, Michel Bierlaire
A model to capture ride-sharing behavior within the agents in a household
Individuals do not schedule their day in isolation from other members of the household. There is interpersonal dependencies and intra-household interactions within household members which affect their daily schedules and travel patterns. Household car ownership, car availability limitation, joint activity participation, joint trips to joint activities, sharing of common household vehicles, and coordination of household daily rhythms are examples of these interpersonal dependencies. These interdependencies might have multiple reasons and can take multiple forms. Therefore, activity-based and travel demand models should incorporate and represent heterogeneous intra-household interactions in order to reflect reality. Focusing on joint rides, we can classify joint household tours into 3 categories: fully joint tours for shared activities, escorting tours, and synchronised activity tours. In this project, we focus on the third category; synchronised activity tours such as ride-sharing within household members. In this semester project, we will study the ride-sharing behavior within the members of a household. To this aim, we first do a literature review to see how the current studies tackle this matter and to identify the influential variables on the choice of ride-sharing or solo ride. We then develop a model to capture the choice of solo travel or shared ride with other agent(s) of the household considering various variables such as socio-economic characteristics of individuals and households, and activities attributes. We aim to study the ride-sharing behavior within the members of a household and capture the influential variables on the joint tours within a household. Finally, the student is expected to analyze and discuss the model results.
Student: Mya Jamal Lahjouji (SGC), January 07, 2023
Supervision: Negar Rezvany, Janody Pougala, Michel Bierlaire
Optimisation et Dploiement dune mthode de calculation au sein dun groupe actif dans le second uvre
Student: Bastien Sauve (SGC), December 23, 2022
Supervision: Michel Bierlaire, Bastien Sauve, Phida Etanchit SA.
Heuristic applications for rebalancing operations in one-way car sharing systems
CSSs are becoming increasingly popular, primarily due to their financial and environmental advantages. However, they face many operational challenges, including inventory management of vehicles and parking spots, vehicle load balancing and redistribution, pricing strategies, and demand forecasting. If these challenges are not addressed properly, the CSS risks experiencing a significant loss of customers and therefore revenue. The literature consists of variety of works on CSSs especially focused on the rebalancing operations. On the other hand, it lacks the detailed comparison of different strategies of rebalancing operations at a disaggregate level. One approach to fill this gap is using an agent-based transport simulation. Such simulation toolkits allow analysis at the individual level and testing different configurations such as city and network structure. Therefore, in this study the student will review the literature to determine two rebalancing strategies existing in CSSs. The student will then focus on developing heuristic algorithms to solve these problems as they are computationally expensive to solve with exact methods.
Student: Miguel Chastre Rodrigues, September 30, 2022
Supervision: Selin Atac, Michel Bierlaire
Divide and conquer one-step simulator for synthetic household generation
Transportation science today is tasked with predicting the complex mobility needs of individuals, which necessitates the use of advanced mobility and travel demand models. However, the quality of the model outputs depends on the data quality. In transport, data privacy and data availability are two limitations. Therefore, transportation scientists rely increasingly on the usage of synthetic populations. Typically, a synthetic population is generated either on the level of individuals or on the level of households using simulation or machine learning approaches. This project aims to solve the problems of existing simulation techniques for the generation of synthetic households, addressing several literature gaps. In existing methodologies, the generation of individuals and their matching into households is done separately, through two sequential processes. Although the marginal distributions of key generated attributes might show a perfect fit, the two-step household generator produces unrealistic households. The generation of illogical observations is caused by neglecting the dependencies between individuals while grouping them into households. In order to create realistic households, this paper suggests a single-step household simulator where relationships between individuals are considered simultaneously within the generation process by imposing various statistical constraints. However, as shown in the past, the simulation methods struggle to deliver accurate results in a reasonable time while dealing with high-dimensional datasets. This project changes the traditional Gibbs Sampler into so-called Divide and Conquer Gibbs Sampler that solves this problem by decomposing and parallelizing the generation process based on the level of correlation. The expectation of this approach is that it will increase accuracy and efficiency, as highly correlated areas are isolated, enabling a better representation of less probable values. This project aims to work on the extension of the aforementioned existing framework. The extension will contain several steps where we will have to add more attributes that will be generated, which includes modeling of all necessary relationships. The addition of more variables will cause the curse of dimensionality phenomena. To overcome this issue, we will work on improving the divisive Gibbs Sampler in the context of synthetic household generation. We will compare it with the traditional approach by investigating the cost of adding more attributes regarding computational efficiency. Moreover, we will focus on proving the advantages of a one-step simulator compared to a two-step simulator in terms of the realism of generated observations. The methodology will be tested using 2015 Swiss census data.
Student: Xinling Li (SGC), September 15, 2022
Supervision: Marija Kukic, Michel Bierlaire
Modeling congestion in a competitive Facility Location Problem
Assigning the right Emergency Room (ER) to persons in medical need is a pivotal mission of the Emergency Medical Dispatcher (EMD). In order to minimize the delay between the call and the medical care, EMD need to take into account the travel time and the waiting time at the an ER. The goal of opening an ER is to maximize the demand capture, either to make profit or to relieve other ERs. To achieve this, it is possible to locate the facilities and assign them capacity of service: number of beds or staff members. In this report, a non-linear integer programming formulation is established to model this type of allocation. The formulation is agent-based to encompass individual disparities through discrete choice modelling (DCM).
Student: Louise Lallemand, August 26, 2022
Supervision: Tom Haering, Michel Bierlaire
Comparing rebalancing operations in car sharing systems
A car sharing system (CSS) offers users to rent cars for a short period of time. The users are identified by an RFID card or through a mobile application. The price of the trip is generally determined according to trip duration and length. One of the challenges faced in such systems is the imbalance of the vehicles throughout the city. This problem is solved by forecasting the demand and rebalancing the vehicles to meet user demand. First, the student is going to review the literature related to rebalancing operations in CSSs. Then, she will design and/or implement different rebalancing strategies in order to test the effects of different characteristics of the city/region. In other words, different strategies (static, reactive, and proactive) for rebalancing operations will be tested for each scenario and the results will be used to drive to conclusions which explain the relation between the city structure, demand structure and the different strategies of rebalancing operations.
Student: Xinling Li (SGC), July 08, 2022
Supervision: Selin Atac, Nicola Ortelli, Michel Bierlaire
A comparitive analysis of optimization algorithms for activity-based applications
The starting point of this semester project is a Mixed Integer Linear Programming (MILP) optimization framework (Pougala et al, 2021), which simulates the scheduling of activities by considering that individuals try to optimize the utility (or satisfaction) they can gain from a given activity plan. Under a time budget constraint (e.g., 24h), the framework allows to integrate every spatio-temporal dimension of the activity-based paradigm (choice of activity participation, timing decisions, destination, mode, route...) simultaneously. The trade-offs occurring in the decision process are therefore captured explicitly. While the framework is functional, the MILP formulation proves to be limited when it comes to make it operational. Specifically, the dimension of the constraint space and the combinatorial nature of the problem significantly increases the computational time, often too high for practical use.The focus of the project is therefore on investigating a constraint programming approach, and its strengths and limitations compared to the MILP formulation. Constraint programming is a widely used optimization methodology, especially for scheduling applications, as the combinatorial nature of such problems suits the general framework of global constraints, domain filtering and constraint-propagation well. The main objective is to perform a comparative analysis of both optimization approaches on the given activity-based problems on selected case studies.
Student: Luca Bataillard (SIN), June 10, 2022
Supervision: Tom Haering, Janody Pougala, Michel Bierlaire
A discrete choice model to capture the relationship between activity type-location-modality choice of individuals
With the COVID-19 pandemic, the lifestyle and behavior of people have changed dramatically. Activities which were traditionally done out-of-home (such as work and education), are now more likely to take place in-home and remote working and studying has become an integral part of our lives. Capturing the relationship between activity type (work, study, sleep, shopping, personal care, household care, Leisure), activity location (in-home, out-of-home), and activity modality (virtual, physical) can be used to assess the tradeoff between in-home and out-of-home activities. In this semester project, we will study how the pandemic has impacted the location and activity mode choice (virtual, physical) of individuals. To this aim, we first do an analysis on the 2016-2020 UK Time Use Survey (TUS) data which encompass both pre- and during the Covid-19 pandemic daily diary data. The data analysis aims to determine the average activity duration and frequency of activity episodes per day by activity type, location, and device use. We will then develop discrete choice models to simulate the activity location (in-home, out-of-home) and modality (virtual, physical) for different activity categories (work, study, personal care, household care, shopping, leisure, sleep, organizational work). We aim to capture the preferences for virtual/physical activity participation of different activity types at specific locations. Finally, the student is expected to analyze and discuss the model results.
Student: Jingran Su (SGC), June 10, 2022
Supervision: Negar Rezvany, Janody Pougala, Michel Bierlaire
Road sections clustering
Pavement maintenance is one of the major issues of public agencies. Inefficient maintenance strategies lead to high economic expenses in the long term. Public agencies rely on pavement management system (PMS) to recommend pavement rehabilitation and maintenance. The first step of a PMS is to create homogeneous worksites of reasonable sizes (considered practical and economically relevant). This is performed by grouping road sections based on conditions data and degradation laws over time for the network. In practice, the grouping or clustering phase is done by a basic algorithm whose results then need to be manually checked and repaired. This makes the task tedious and time consuming. The goal of this project is in a first instance to implement a deterministic clustering algorithm that takes as input road sections information (position, type of surface, type of traffic, age...) together with a deterministic degradation law and regroups the sections in clusters using Machine Learning algorithms such as k-means. In a second instance, a probabilistic degradation law will be considered and a probabilistic clustering algorithm such as ExpectationMaximization will be implemented. Both algorithms will be tested on real world case study and the result will be compared to the results of an existing PMS clustering methodology.
Student: Salim Benchelabi (SMA), June 07, 2022
Supervision: Nour Dougui, Marija Kukic, Michel Bierlaire
Studying socio-economic variables through the different waves of SARS-CoV-2
The lockdown of March 2020 of several countries has changed the modern world as we know it. Back then, non-pharmaceutical interventions were put into place to mitigate the spreading of the COVID-19. Now, we count on the most effective pharmaceutical intervention, vaccination. In Switzerland, 65.9% of the population is fully vaccinated. At the time, we are dealing with the fifth wave of new cases. However, it is unclear whether the pandemic has a significant impact on the travel behaviour and whether adopting non-pharmaceutical interventions has an impact on daily activities. Therefore, it is important to take a disaggregate approach to analyze the impact of all these variables on the travel behavior of individuals using a disaggregate approach. The long-term goal is to provide insights that transportation planners and public health authorities can exploit to deal with epidemic situations. In this project, we will study the use of different ML algorithms to include the different significant variables that influence the force of infection (probability of going from being susceptible to infected) through the different waves. The main objective is to capture which socio-economic variables of the individuals and which municipality characteristics influence the force of infection and how the impact of these variables change trough the different waves. First, the student will review the existing literature related to epidemiological and activity-based models and get familiar with the tools already developed by the TRANSP-OR lab. Then, the student is expected to test different models through the different waves and check the impact of these variables on behaviour. To account for the contact matrices, we will use FOPH data together with MATSIM output of Switzerland. Finally, the student will calibrate the parameter inside the TRANSP-OR simulator.
Student: Clment Dauvilliers, June 03, 2022
Supervision: Cloe Cortes Balcells, Silvia Varotto, Michel Bierlaire
Drawing infection probabilities based on socio-economic characteristics using ML Causal Inference
Countries everywhere in the world have been adopting different policies to lower the toll of the COVID-19 pandemic. Epidemiological and behavioural models have been widely used to guide decision-making. In this project, we will study the use of causal inference as a tool to determine which variables are significant and study their causal effect on the probability to become infected. We aim to compare these results with the more traditional approach. It will account for the uncertainty inherent to the nature of the problem to decide which variables are meaningful when modelling the infection with a virus. We hope this new model will provide insight on transmission and intervention that will complete what can be obtained with usual compartmental models (SIR). To do that, each student will implement different causal discovery methods based on graphical models. For example, a score-based method like the greedy equivalence search architecture, or a non-linear method Iike a method based on the linear, non-Gaussian model.
Student: Hannah Gelblat, February 04, 2022
Supervision: Cloe Cortes Balcells, Rico Krueger, Michel Bierlaire
Drawing infection probabilities based on socio-economic characteristics using ML Causal Inference
Countries everywhere in the world have been adopting different policies to lower the toll of the COVID-19 pandemic. Epidemiological and behavioural models have been widely used to guide decision-making. In this project, we will study the use of causal inference as a tool to determine which variables are significant and study their causal effect on the probability to become infected. We aim to compare these results with the more traditional approach. It will account for the uncertainty inherent to the nature of the problem to decide which variables are meaningful when modelling the infection with a virus. We hope this new model will provide insight on transmission and intervention that will complete what can be obtained with usual compartmental models (SIR). To do that, each student will implement different causal discovery methods based on graphical models. For example, a score-based method like the greedy equivalence search architecture, or a non-linear method Iike a method based on the linear, non-Gaussian model.
Student: Mathias Nuris, February 04, 2022
Supervision: Cloe Cortes Balcells, Rico Krueger, Michel Bierlaire
Adding an Epidemiological penalty in MATSim to study the impact of SARS-CoV-2 in mobility
Countries everywhere in the world have been adopting different policies in order to lower the toll of the COVID-19 pandemic. Epidemiological and behavioural models have been widely used to guide decision making. In this project, we will study how the disease impacts the choice of activities. For that reason, we will modify the utility scoring function from the activity-based model to introduce the epidemiological penalty. We will add in the utility function, another term called Slevel of virus load to define the impact that an epidemic has on the choice of activities, including travel mode. To estimate the parameters of the model we are planning to use already-existing methodologies like maximum likelihood or Bayesian methods. First, the student will review the existing literature related to epidemiological and activity-based models and get familiar with MATSim software. Then, the student is expected to develop an environment in MATSim using data we will provide. Specifically, we will use the Sioux Falls scenario files which are public, together with google mobility data so that we can modify the base mobility according to the different policies applied during the different waves of SARS-CoV-2. Finally, he will create different scenarios in order to test the effects of different characteristics of the city/region.
Student: Florent Zolliker (SMA), January 31, 2022
Supervision: Cloe Cortes Balcells, Negar Rezvany, Michel Bierlaire
Optimal strategy for a multi-vendor, multi-buyer reverse logistics supply chain framework
With the boom of e-commerce and online shopping, retailers and logistics operators must look to sustain- able process improvements in their supply chain to manage the growing volume of inventory and returns. A sustainable logistics strategy contributes to a successful transition to a low- or zero-carbon economy and improves the companys social impact. In the last-mile segment of delivery in urban centers, efficient reverse logistics strategies are essential to reduce the environmental impact of logistics operations and to reduce greenhouse gas emissions. The goal of this project is to describe a multi-vendor, multi-buyer reverse logistics supply chain frame- work, and to implement a suitable optimization algorithm to maximize revenue streams from combining new and re-manufactured products in the reverse logistics based supply chain.
Student: Olivier Laferrre (SMT), January 31, 2022
Supervision: Melvin Wong, Claudia Bongiovanni
Classification of ordinal outcomes for the analysis of injury severity using machine learning methods (ML for Science CS-433)
Ordinal scale responses capture qualitative user feedback which can be used to model individual choice preference or are employed in traffic accident analysis to evaluate accident severity. In this project, we aim to predict the outcome of the severity of the accident using machine learning methods, specifically, using a deep neural network to train the crash model. The model accuracy will be evaluated on three metrics: discrete classification accuracy, geometric mean probability of correct assignment and quadratic weighted kappa.
Student: Hilda Abigl Horvth, Julian Paul Schnitzler, Artur Andrzej Stefaniuk (SSC), December 23, 2021
Supervision: Silvia Varotto, Melvin Wong, Michel Bierlaire, Martin Jaggi
Designing and implementing rebalancing operations in car sharing systems using MATSim
In this project, we will focus on Vehicle sharing systems (VSSs). VSSs are becoming increasingly popular, primarily due to their financial and environmental advantages. However, they face many operational challenges, including inventory management of vehicles and parking spots, vehicle load balancing and redistribution, pricing strategies, and demand forecasting. If these challenges are not addressed properly, the VSS risks experiencing a significant loss of customers and therefore revenue. The literature consists of variety of works on VSSs especially focused on the rebalancing operations. On the other hand, the literature lacks the demand forecasting dimension of the VSSs since it is exhausting to collect the data to develop a demand model. This project will aim to address this problem in a different way. Instead of developing a demand model, we aim to simulate the demand structure using an open-source software MATSim (Multi-Agent Transport Simulation Toolkit). The student is expected to develop an environment in MATSim for a car sharing system (CSS) where the census data is used. Then, he will create different scenarios in order to test the effects of different characteristics of the city/region. She will then identify the cases in which the rebalancing operations are needed. The implementation of rebalancing operations will be adapted from the free floating application of MATSim. In other words, different strategies (static, reactive, and proactive) for rebalancing operations will be tested for each scenario and the results will be used to drive to conclusions which explain the relation between the city structure, demand structure and the different strategies of rebalancing operations.
Student: Alfio Simone Mosset (SGC), July 09, 2021
Supervision: Selin Atac, Stefano Bortolomiol, Michel Bierlaire
Modeling Geneva's public transport network following disruption events
The project is conducted in collaboration with the Transports Publics Genevois (TPG). The principal objective is to simulate the impact of disruptions (e.g. accidents) on the Geneva road network, and to optimize its resilience by proposing targeted and ecient intervention strategies. A particular emphasis will be put on the public transport network.
Student: Nicolas Richter (SGC), July 09, 2021
Supervision: Marija Kukic, Janody Pougala, Michel Bierlaire
Matrix factorisation methods
The aim of the project is to explore the state-of-the art of the use of matrix factorisation methods in machine learning and econometrics through literature surveys and practical implementation exercises.
Student: Ambroise Favre, June 30, 2021
Supervision: Rico Krueger, Michel Bierlaire
Application of sustainable reverse logistics in optimization of last-mile delivery systems
This project will investigate the state-of-the-art data-driven approaches to solve these problems using process mining and optimization to model sustainable reverse logistics systems. A comprehensive literature on reverse logistics, last-mile delivery, process mining and optimization techniques will be also be produced. Finally an implementation of the identified approach will be developed as a use case example for reverse logistics
Student: Jacques Roitel, June 30, 2021
Supervision: Melvin Wong, Michel Bierlaire
How to assess synthetic data?
Student: Romain Palazzo (SMA), June 13, 2021
Supervision: Gael Lederrey, Nicola Ortelli, Michel Bierlaire
MoonVR
Student: Lucas Strauss, June 13, 2021
Supervision: Claudio Leonardi
Designing a MATSIM environment to study the impact of SARS-CoV-2 in mobility
Countries everywhere in the world have been adopting different policies in order to lower the toll of the COVID-19 pandemic. Epidemiological and behavioural models have been widely used to guide decision making. In this project, we aim at building a disaggregate model (agent-based approach) that will couple two models: one with the activities of people and another one with the transmission of the virus. We hope this new model will provide insight on transmission and intervention that will complete what can be obtained with usual compartmental models (SIR). In order to do that, we will use the Matsim API (Episim), hence the student will not code everything from scratch. First, we will review the existing literature related to epidemiological and activity-based models and get familiar with MATSim software. Then, we will develop an environment in MATSim. Specifically, we will use the Berlin scenario files which are public, together with google mobility data so that we can modify the base mobility according to the different policies applied during the different waves of SARS-CoV-2. Finally, we will create different scenarios in order to test the effects of different characteristics of the city/region.
Student: Antoine Crettenand, June 11, 2021
Supervision: Cloe Cortes Balcells, Rico Krueger, Michel Bierlaire
Road Safety: Investigating the Role of Latent Behaviors in Injury Severity
Reducing the number of fatalities caused by traffic accidents is a major concern everywhere around the world. In order to do so, it is crucial to understand how various factors influence injury severity. The existing literature generally deals with explaining the influence of directly observable variables on the occurence and severity of crashes, but fails to adequately consider driver behavior. Hence, the main goal of this project is to explore approaches that enable the incorporation of latent attitudes in injury severity models. Specifically, we are interested in investigating the "propensity to taking risks" that drivers may exhibit, as well as the influence of such behaviors on the severity of injuries suffered by other road users.
Student: Daniela Spadaro (SMA), June 07, 2021
Supervision: Nicola Ortelli, Gael Lederrey, Michel Bierlaire
Designing a MATSIM environment to study the impact of SARS-CoV-2 in mobility
Countries everywhere in the world have been adopting different policies in order to lower the toll of the COVID-19 pandemic. Epidemiological and behavioural models have been widely used to guide decision making. In this project, we aim at building a disaggregate model (agent-based approach) that will couple two models: one with the activities of people and another one with the transmission of the virus. We hope this new model will provide insight on transmission and intervention that will complete what can be obtained with usual compartmental models (SIR). In order to do that, we will use the Matsim API (Episim), hence the student will not code everything from scratch. First, we will review the existing literature related to epidemiological and activity-based models and get familiar with MATSim software. Then, we will develop an environment in MATSim. Specifically, we will use the Berlin scenario files which are public, together with google mobility data so that we can modify the base mobility according to the different policies applied during the different waves of SARS-CoV-2. Finally, we will create different scenarios in order to test the effects of different characteristics of the city/region.
Student: Chantal Gressier (SMA), June 03, 2021
Supervision: Cloe Cortes Balcells, Negar Rezvany, Michel Bierlaire
The railway timetable rescheduling problem
In railway networks, unexpected disruptions may occur for different reasons and cause delays, service denial, and, consequently, passenger inconvenience. This pre-study project will look at remediation strategies such as canceling, delaying or rerouting trains in case of unexpected disruptions. This pre-study project follows two main lines of research. First, the student will familiarize with state-of-the-art methodologies and algorithms to generate railway timetables in a disrupted network. Second, the student will learn how these methodologies can be integrated in the commercial software Viriato, produced by the company SMA und Partner AG.
Student: Benoit Pahud (SGC), February 19, 2021
Supervision: Nour Dougui, Marija Kukic, Stefano Bortolomiol, Matthias Hellwig, SMA und Partner AG
Designing a MATSim environment for a vehicle sharing system as a transport mode
A vehicle sharing system (VSS) offers users to rent vehicles for a short period of time. The price of the trip is generally determined according to trip duration and length. These systems are becoming increasingly popular, primarily due to their financial and environmental advantages. However, VSSs face many operational challenges, including inventory management of vehicles and parking spots, vehicle load balancing and redistribution, pricing strategies, and demand forecasting. If these challenges are not addressed properly, the VSS risks experiencing a significant loss of customers and revenue. One of the challenges faced in such systems is the imbalance of the vehicles throughout the city. This problem is solved by forecasting the demand and rebalancing the vehicles to meet user demand. However, since it is exhausting to collect the data to develop a demand model, this project will make use of the Multi-Agent Transport Simulation Toolkit (MATSim) to build this environment. First, the student is going to review the literature related to VSSs and get familiar with MATSim software. The student is expected to develop an environment in MATSim for a VSS where the census data is used and the VSS is the only transport mode in the city other than walking. Then, she will create different scenarios in order to test the effects of different characteristics of the city/region. She will then identify the cases in which the rebalancing operations are needed. In other words, different strategies (static, reactive, and proactive) for rebalancing operations will be tested for each scenario and the results will be used to drive to conclusions which explain the relation between the city structure, demand structure and the different strategies of rebalancing operations.
Student: Paula Vogg (SGC), January 29, 2021
Supervision: Selin Atac, Cloe Cortes Balcells, Michel Bierlaire
Schedule repair in liner shipping
Large liner shipping companies operate several hundred ships worldwide. These ships carry multi-modal containers on pre-established routes with a regular schedule (typically weekly). A schedule is thus an ordered list of port calls (i.e. stops in a harbour) with associated time windows. Since there are several thousands possible ports in the world, designing routes is a complex problem. Furthermore, since a rotation takes weeks or even months, roughly four journeys out of five end up requiring adjustments to accommodate delays, breakdowns or other unforeseeable but frequent events. Due to the extent of the operations, improving from an adequate schedule to an optimal one may yield enormous benefits, not only economically but also in terms of environmental impact and of quality of life for workers. Given the nominal schedule and an additional constraint stemming from an unforeseen event, the aim of the project is to use geographical information and demand statistics in the various ports of a liner shipping network, the student applies Operational Research methodologies to generate a new optimal schedule, completing as much of the original mission as possible while minimising the ecological and economic impacts.
Student: Benoit Pahud (SGC), January 29, 2021
Supervision: Stefano Bortolomiol, Nour Dougui, Michel Bierlaire
Generating choice sets of destinations for activity based applications
Understanding human mobility behavior is essential to estimate and forecast transport demand. Activity-based models (ABM) are an increasingly popular approach to integrate complex behavioral dimensions influencing mobility choices, such as habits or social interactions. A signicant challenge in ABM research is the choice set for each individual (consisting of the type of activities, the schedules, the locations and the modes to reach them), which is highly combinatorial and thus impossible to fully enumerate. This pre-study will specically investigate the choice set of locations for individuals' daily activities, guided by (but not limited to) the following research questions: (1) How can we generate a plausible set of locations for an individual to consider, for each of their activities? (2) Which attributes can be used to characterize a location or a space (e.g. opening hours, public transport level of service, job density...) ? (3) What is the required granularity or scale to yield consistent results in terms of activity-based estimation, while maintaining a computational tractability ? (4) What are the data requirements for these tasks? What methods or tools can be used to augment datasets which lack this essential information ? The objective is to develop a methodology to generate a choice set of locations, by means of theoretical and data-driven approaches. The choice of location itself is not included in the scope of this project.
Student: Nicolas Salvad (SGC), January 29, 2021
Supervision: Janody Pougala, Tim Hillel, Michel Bierlaire
Generating choice sets of transport modes for activity based applications
Understanding human mobility behavior is essential to estimate and forecast transport demand. Activity-based models (ABM) are an increasingly popular approach to integrate complex behav- ioral dimensions influencing mobility choices, such as habits or social interactions. A signicant challenge in ABM research is the choice set for each individual (consisting of the type of activities, the schedules, the locations and the modes to reach them), which is highlycombinatorial and thus impossible to fully enumerate. This pre-study will specically investigate the choice set of transport modes for individuals' daily activities, guided by (but not limited to) the following research questions: (1) How can we generate a plausible set of modes for an individual to consider, for each of their activities? (2) What are the main determinants for including/excluding a mode in the choice set? (3) What are the data requirements for these tasks? What methods or tools can be used to augment datasets which lack this essential information? The objective is to develop a methodology to generate a choice set of modes, by means of theoretical and data-driven approaches. The mode choice itself is not included in the scope of this project.
Student: Benoit Pahud (SGC), January 29, 2021
Supervision: Janody Pougala, Tim Hillel, Michel Bierlaire
Road maintenance management optimisation
Pavement maintenance is one of the major issues of public agencies. Insufficient investment or inefficient maintenance strategies lead to high economic expenses in the long term. Under budgetary restrictions, the optimal allocation of resources becomes a crucial aspect. Therefore, public agencies rely on pavement management system that prioritize and recommend pavement rehabilitation and maintenance to maximize results within a given budget amount. The goal of this project is to produce a literature review on the existing pavement management systems and methodologies that optimize road maintenance on both network and project level. Besides, different road conditions degradation model should be investigated. Finally, a heuristic method for road maintenance should be implemented and tested on a case study while testing different scenarios of road condition degradation.
Student: Carole Basl (SGC), January 29, 2021
Supervision: Melvin Wong, Nour Dougui, Michel Bierlaire
Analyzing Machine learning regularization methods in travel mode choice prediction (ML for Science CS-433)
A problem in machine learning is to make an algorithm perform well not just on the training data, but also on new inputs. There are a number of regularization methods such as L1/L2, Dropout, max-norm that help improve generalization by reducing overfitting/overtraining. A Deep learning model is applied to a travel mode choice dataset. The objective of the project is to investigate regularization methods for improving learning accuracy.
Student: Simon Dayer, Arnaud Guibbert, Khalil Merzouk (SSC), December 17, 2020
Supervision: Melvin Wong, Michel Bierlaire, Martin Jaggi
Optimizing Organizational Chart using local search method
The aim of this project is to produce a near optimal Organisational Chart for a real case study. This Organisational Chart should optimize the span of control of Management in the company by maximizing a coverage metric using clearly-de ned manipulations of the organizational design. To achieve this goal a local search method will be used and the results should be compared to those given by an exact method (Simplex). We will in a rst phase apply the optimization method to a small sample of Organisational Chart without considering the role of the employees. In a second phase, the role of the employees should be considered and a larger Organisational Chart should be considered.
Student: Hugo Bocquet, June 30, 2020
Supervision: Nour Dougui, Selin Atac, Michel Bierlaire
Enhancing the Serial Estimation of Discrete Choice Models Sequences
In the past years, an increasing number of studies has investigated the usefulness of data-driven methods for the task of developing discrete choice models so as to alleviate modelers from the burden of manual specification. One of these methods makes use of a variable neighborhood search algorithm to mimic the way an experienced modeler would build the utility functions of a discrete choice model. More precisely, small modifications are sequentially brought to an initial specification, while the induced improvement is assessed by means of a measure of performance. Although the quality of the generated models is high, this method faces a severe limitation: the computational time for a single run counts in hours with relatively small datasets and grows quickly with the amount of considered data. The main goal of this project is to speed up the estimation of DCMs in this precise context.
Student: Youssef Kitane (SGC), June 22, 2020
Supervision: Nicola Ortelli, Gael Lederrey, Michel Bierlaire
Optimal regulation of oligopolistic markets with discrete choice models of demand
This project is inspired by the ongoing research project titled 'New generation of demand-supply interaction models', funded by the Swiss National Science Foundation. In particular, the project aims at studying models to find optimal policies to regulate markets characterized by oligopolistic competition and in which consumers make a discrete choice among a finite set of alternatives. In this framework, consumers are modelled as utility maximizers, according to random utility theory. Suppliers are modelled as profit maximizers, according to the traditional microeconomic treatment. Market competition is modelled as a non-cooperative game, for which an epsilon-equilibrium solution is sought. Finally, the regulator can affect the behavior of all other agents, for instance by giving subsidies or imposing taxes. In transport markets, these interventions might target specific alternatives, to reduce externalities such as congestion or emissions, or specific segments of the population, to achieve social welfare objectives. The objective of the project is to analyze, implement and evaluate algorithms to solve realistic applications of this problem, with particular focus on the transportation sector.
Student: Elodie Duliscouet, Paulin Raison, Yahya Basiouny (SGM), June 12, 2020
Supervision: Stefano Bortolomiol, Selin Atac, Michel Bierlaire
Robust routing and scheduling in liner shipping
Large liner shipping companies operate several hundred ships worldwide. These ships carry multi-modal containers on pre-established routes with a regular schedule (typically weekly). A schedule is thus an ordered list of port calls (i.e. stops in a harbour) with associated time windows. Since there are several thousands possible ports in the world, designing routes is a complex problem. Furthermore, since a rotation takes weeks or even months, roughly four journeys out of five end up requiring adjustments to accommodate delays, breakdowns or other unforeseeable but frequent events. Due to the extent of the operations, improving from an adequate schedule to an optimal one may yield enormous benefits, not only economically but also in terms of environmental impact and of quality of life for workers. Given geographical information and demand statistics in the various ports of a network, the project aims at applying Operational Research methodologies to automatically generating shipping schedules that make sense technically, economically and environmentally, as well as resisting perturbations. This project is proposed by Mediterranean Shipping Company (MSC).
Student: Yannis Voet (SGC), June 12, 2020
Supervision: Stefano Bortolomiol, Nour Dougui, Michel Bierlaire
Bayesian analysis of multinomial discrete choice model with t-distributed kernel errors
This project develops Bayes estimators for two robust alternatives to the multinomial probit models. Both alternative models belong to the family of robit models, whose kernel error distributions are heavy-tailed t-distributions. The first model is the multinomial robit (MNR) model in which a generic degrees of freedom parameter controls the heavy-tailedness of the kernel error distribution. The second alternative, the generalised multinomial robit (Gen-MNR) model, is more flexible than MNR, as it allows for alternative-specific marginal heavy-tailedness of the kernel error distribution. The performance of the proposed estimators is evaluated on simulated and real data.
Student: Thomas Gasos, June 09, 2020
Supervision: Rico Krueger, Michel Bierlaire
Welfare-maximizing design of a transportation system
The simulation-based linear representation of a discrete choice introduced allows to linearly approximate the expected maximum utility associated with each individual and simulation draw, which provides a measure of consumer surplus. This is of special interest in the case of advanced choice models (e.g., mixture of logit models), as the expected maximum utility presents a highly nonlinear expression. Hence, a linear formulation for the consumer surplus facilitates the derivation of a generalized objective function of some measure of social welfare as the policy objective, which can be employed to assess the performance of different urban transportation policies. The objective of the project is to provide linear (or piecewise linear) formulations to the different aspects that are included in a published non-linear optimization problem, and compare their performance with respect to the original approach, both in terms of computational time and quality of the obtained solution. Additionally, the impact of the assumptions that lead to nonlinear expressions will be evaluated.
Student: Benjamin Nicolas--Noir, May 29, 2020
Supervision: Meritxell Pacheco, Michel Bierlaire
Analysis of the value of demand forecasting within vehicle sharing systems
Vehicle sharing systems (VSSs) are becoming increasingly popular, primarily due to their financial and environmental advantages. However, VSSs face many operational challenges, including inventory management of vehicles and parking spots, vehicle load balancing and redistribution, pricing strategies, and demand forecasting. If these challenges are not addressed properly, the VSS risks experiencing a significant loss of customers and revenue. Recently, new VSSs have been introduced which use light electric vehicles (LEVs). These systems face a number of unique challenges. For instance, demand forecasting for LEV sharing systems is more complex, as locations are not fixed and journeys can start and end at any allowable location. LEV sharing systems also serve a higher portion of the population since these type of vehicles do not require a driving license. As such, the existing techniques for analysing VSSs are not sufficient for these new systems. To address this need, one needs to forecast the future demand of this novel transportation mode. Since it is exhausting to collect the data to develop a demand model, this project will aim to identify the value of a demand model by using mathematical models tailored for rebalancing operations from the literature and simulations. As LEV systems are still in their infancy, data describing them is not yet available. Therefore, the student will use alternative data, including PubliBike bike-sharing system data, accounting for any potential differences in the analysis. Based on the findings from the literature, the student will analyse different mathematical models.
Student: Jasso Espadaler (SGC), January 31, 2020
Supervision: Selin Atac, Stefano Bortolomiol, Michel Bierlaire
Tour-based mode choice modelling
This project will involve working with real world trip-diary data for London and/or Switzerland to develop sequential tour-based mode-choice models (i.e. where the choice of mode for a trip-leg is dependent on the full tour). The project will involve: Extracting tour itineraries from the trip-diary data, Investigating different tour patterns, e.g. primary return-tours, subtours, complex multilegged tours, Defining different tour-based mode-choice situations which are able to fully represent the tour patterns existing in the data, Developing appropriate sequential choice models for each mode-choice situation using Discrete Choice Models and/or machine learning.This project will involve working with real world trip-diary data for London and/or Switzerland to develop sequential tour-based mode-choice models (i.e. where the choice of mode for a trip-leg is dependent on the full tour). The project will involve: -Extracting tour itineraries from the trip-diary data, -Investigating different tour patterns, e.g. primary return-tours, subtours, complex multilegged tours, -Defining different tour-based mode-choice situations which are able to fully represent the tour patterns existing in the data, -Developing appropriate sequential choice models for each mode-choice situation using Discrete Choice Models and/or machine learning.
Student: Adrien Nicolet (SGC), January 31, 2020
Supervision: Tim Hillel, Janody Pougala, Michel Bierlaire
Risk assessment of pedestrian crossings
Student: Adrien Nicolet (SIN), January 31, 2020
Supervision: Nicholas Molyneaux, Gael Lederrey, Michel Bierlaire
Investigating daily activity patterns
This project will involve working with real world trip-diary data for Switzerland (2015 Mobility and Transport microcensus) to investigate how people schedule activities (e.g. work, lunch, sport, shopping etc) in their day. The project will involve: -Extracting and standardising daily activity schedules from the data, including representing the activity types, locations, start times, and durations with appropriate precision and/or scale, -Establishing a metric to define the similarity (or distance) between different activity schedules, -Using unsupervised machine learning (clustering) algorithms to identify prototypical activity schedules.
Student: Sergey Gasparovich (SGC), January 31, 2020
Supervision: Janody Pougala, Tim Hillel, Michel Bierlaire
Estimation of Discrete Choice Models using HAMABS 2.0
Improvement of HAMABS for choice modeling estimation using automatic switch.
Student: Linah Charif (SGC), December 20, 2019
Supervision: Gael Lederrey, Nicola Ortelli, Michel Bierlaire
Demand forecasting for a novel transportation mode
The vehicle sharing systems (VSSs) are becoming more and more popular due to both economic and environmental effects. However, these systems mainly focus on two specific types of vehicles: bicycles and cars. Recently, some other types of vehicles (e.g. light electric vehicles) with some specific properties that distinguish them from cars and bicycles are introduced for use in such kind of systems. Therefore, they lack research in each component of the framework. The aim of this project is to forecast the demand for a novel transportation mode. The student should survey different types of approaches, such as ARIMA, to forecast demand in similar systems and compare them by identifying performance measures. Since the data for this specific transportation mode does not exist yet the student will be directed to use the taxi data, which shares many properties with the system of interest.
Student: Denis Steffen (SMA), June 30, 2019
Supervision: Selin Atac, Tim Hillel, Michel Bierlaire
Calibrating pedestrian control strategies.
The objective of this semester project is to calibrate two pedestrian control strategies using heuristic methods and/or optimization methods like simulated annealing or genetic algorithms. Two strategies are under investigation: gating to prevent a gridlock situation and flow separators which prevent counter flow. The student will have to use the pedestrian simulator I have implemented and integrate it with an optimization approach like simulated annealing.
Student: Lopold Bouraux (SIN), June 30, 2019
Supervision: Nicholas Molyneaux, Gael Lederrey, Michel Bierlaire
Disruption-caused railway timetable rescheduling problem and its solution
In railway networks, unexpected disruptions may occur for different reasons and cause delays, service denial, and, consequently, passenger inconvenience. This planned master project will tackle the railway timetable rescheduling problem from a macroscopic and mesoscopic point of view in case of unexpected disruptions. The master project will consider several remediation strategies such as canceling, delaying or rerouting the trains, or introduction of other transportation modes. The other important aspect of the planned master project is its applicability in practice. Hence, the outcome of the project will be the algorithm for timetable rescheduling which will be implemented in the commercial software suite Viriato. The main functionality of Viriato is the train timetabling and it is produced by the company SMA und Partner AG. In order to prepare the candidate for the master project, this pre-study project will include several introductory tasks related to the problem comprehension, solutions approach and algorithm implementation. Therefore, the tasks of this project will be: 1) review of the relevant literature, 2) demand data preprocessing with the purpose of matching infrastructure and demand data, and 3) obtaining the knowledge of the Viriato software and its Algorithm Platform which enables extension of Viriato with third party algorithms.
Student: Oliver Mathias Buschor, June 21, 2019
Supervision: Nikola Obrenovic, Meritxell Pacheco, Michel Bierlaire
In collaboration with SMA und Partner
Location choice equilibrium - pedestrian demand analysis at EPFL campus
Modeling location choice is fundamental to understand travel behavior and to predict travel demand in urban spaces. Location choice models are often developed based on the discrete choice framework and allow the analyst to predict to which place (and when) an individual travels. While the location choice behavior has been studied a lot in the city scale, the literature in pedestrian facilities is relatively limited. In this project, we develop a pedestrian location choice model in the EPFL campus, using the data collected from Wi-Fi traces. We start with developing a simple location choice model and test its specications using the sample. Then the student collects additional data of population for aggregate forecasting, i.e. actual demand prediction at each location. Using this technique, we then extend the model to consider the congestion effect. This involves in a fixed point problem to achieve a Nash equilibrium, solved by the nested fixed point (NFXP) algorithm or the nested pseudo likelihood (NPL) algorithm.
Student: Tianyang Dong (SMA), June 03, 2019
Supervision: Yuki Oyama, Zhengchao Wang, Michel Bierlaire
Passenger satisfaction maximization under budget constraints
This project aims at characterizing a passenger satisfaction application relying on a demand-based optimization framework, which integrates discrete choice models (state-of-the-art for the mathematical modeling of the demand) n Mixed Integer Linear Programming (MILP) models, which are usually considered to address supply decisions such as the price of a service or the number of units to produce of a certain item. This formulation allows us to express passenger satisfaction directly in terms of the expected maximum utility of the future scenario, which simplifi es notably the common representation relying on the consumer surplus. The objective is the maximization of the passenger satisfaction in a short-distance commuting context while accounting for different settings with respect to road tolling and investment in public transportation.
Student: Tatiana Moavensadeh-Ghasnavi, June 03, 2019
Supervision: Meritxell Pacheco, Yuki Oyama, Michel Bierlaire
Dynamic optimization of self-service vehicles
This project is a pr-tude project that will prepare the student to working on the Master thesis. The proposed topic is related to the dynamic optimization of self-service vehicle fleets, with a possible focus on dynamic pricing.
Student: Mohamed Detsouli (SGC), January 31, 2019
Supervision: Stefano Bortolomiol, Nikola Obrenovic, Michel Bierlaire
Railway infrastructure maintenance
Following this summer tragedy in Genoa, Italy, European states and governments became more aware of the importance of keeping their countries infrastructure modern and maintained. This interdisciplinary project will particularly focus on the topic of railway infrastructure maintenance and will cover some legal, economic, social, political aspects related to this socio-economic problem. The goal of this project is to study the topic of railway infrastructure maintenance from two perspectives. Initially, a qualitative analysis will be performed to precisely define the scope of the research and the main factors that affect the planning of investments to maintain railways infrastructures. Based on the findings of the qualitative analysis, the students will then critically analyze existing mathematical models that generate maintenance schedules and optimal railway investments, and will propose a framework which will be applied on a real-life case study.
Student: Ludovica Sessa and Robert Abboud (SMGT), January 31, 2019
Supervision: Stefano Bortolomiol, Selin Atac, Michel Bierlaire
Passage au 1/4h de l'exploitation du FMA
Les Transports Public Fribourgois (TPF) exploite la ligne voie normale Fribourg-Morat-Anet (FMA). Les trains assurant ce service utilise les voies des CFF Fribourg, entre Fribourg et Givisiez, Morat, entre Morat et Muntelier et les voies du Bern-Neuchtel (BN) expoit par les BLS Anet. L'horaire de cette ligne est donc difficile mettre en place. Le projet consiste proposer une offre au 1/4h sur toute la ligne et/ou une offre combine en pseudo-1/4h en alternant des trains REG et RE. Les mesures d'infrastructures seront mises en vidence.
Student: Flix Boesch (SGC), January 31, 2019
Supervision: Daniel Emery
Variantes d'horaires sur la tangentielle Nantes-Lyon
Environ 69% de la longueur de la tangentielle ferroviaire Nantes-Lyon sont aujourd'hui lectrifis en 25kV/50Hz et 9% en 1,5kV=. Toutefois, les 143 km sparant St.Germain-des-Fosss St.Germain-au-Mt-d'Or restent non lectrifis. Le projet consiste dvelopper des variantes d'horaires entre Nantes et Lyon en tenant compte notamment des correspondances dans les noeuds de Nantes, Angers, Tours et Bourges. La problmatique des correspondances avec Clermont-Ferrand sera prise en compte et l'insertion des circulations dans le noeud lyonnais sera sommairement analyse. Les ventuelles difficults pour cette tangentielle fournir la fois une succession de services relativement locaux et une desserte de bout en bout performante seront mises en vidence.
Student: Axel Valentin Gabriel Curis (SGC), January 31, 2019
Supervision: Daniel Emery
Exploitation optimale de la ligne voie mtrique des TPF Palzieux-Bulle- Montbovon
La ligne des TPF voie mtrique Palzieux-Bulle-Montbovon doit offrir de bonnes correpondances dans ces trois noeuds ferroviaire. Le projet consiste proposer des horaires optimaux en fonction des variantes d'horaire CFF entre Palzieux et Fribourg d'une part et de l'exploitation MOB (Montreux-Oberland Bernois) d'autre part. Une desserte plus dense au Sud de Bulle et certains prolongements jusqu' Chateau d'Oex peuvent tre envisags, en fonction des besoins et des temps de rebroussement disponibles.
Student: Fabien Jacot-Descombes Jonas Gschwend (SGC), January 31, 2019
Supervision: Daniel Emery
Formulating and solving a dial-a-ride problem
The Dial-a-Ride Problem (DARP) is a problem to design a vehicular route and schedule, given the passenger requests that are characterized by origins (pickup points) and destinations (delivery points) often with the time windows. In recent years, solving the DARP is increasingly demanded, re ecting new technologies for mobility. On-demand transportation for elderly or disabled people is a typical example of application. The project is mainly dedicated to the following specific tasks. First, the literature review will be done so that the student gets familiar with the model concepts and formulations of the DARP. The student is expected to understand the difference among several types of models, such as between static and dynamic models, or between single-vehicle and multi-vehicle models. At the same time, the solution methodologies that are relevant to each type of DARP will be investigated. Second, the student should acquire the basic skills for optimization problem. Though several exercises, she will get used to coding and using appropriate softwares for solving the problem. Given the knowledge and skills, finally, she should define a DARP for a specific example. The problem will be solved by at least two methodologies and be analyzed.
Student: Rym Karime (SGC), January 31, 2019
Supervision: Yuki Oyama, Nour Dougui, Michel Bierlaire
Optimal taxi charging decision given the real-time charging station and taxi states and future uncertainties
This project aims at optimizing the charging decision of taxi drivers. Different from conventional taxi drivers, electric taxi drivers are more influenced by their refuelling strategy. This is because the refuelling process of electric taxis is longer than conventional taxis, especially in the case that electric taxis need to wait before charging. This longer charging duration will decrease the serving time of electric taxis which will, in return, influence the profits of electric taxis. Therefore, an intelligent charging decision strategy which can help electric taxis to charge optimally (maximizing the profit of taxi) is highly desirable. To build the optimal charging decision program, we assume that a taxi will make a decision at each time step. When the taxi drivers make a decision, we assume that the taxi driver knows that, in the future hours, what are the expectation and variance of revenue during each time period (let's say 30 minutes). In addition to that, the taxi driver also knows the exact position of charging station and the chargers that are installed in the station. Besides, we also assume that the taxi driver also receives information about the availability of chargers from all stations by the time he makes a decision. Moreover, the taxi driver also has a prediction of the congestion condition of the charging station in the following hours. Based on the above assumption, the taxi drivers make his charging decision at each time step, i.e. charge or not, if charge, where, and when to charge for how much. The students who work on this project should build an optimization program which can make the charging decision for the taxi driver. This project is composed of two stages. First, the students will look into the taxi tracking data to understand the taxi request characteristics. This includes the mean and variance of a taxi running distances and revenue from served taxi demands across the taxis. After, the students will learn to come up with an optimization program which can help the taxi drivers to make the charging decision given the current states of taxis and its knowledge about the future world.
Student: Julien Johan Haan (Section of GC), Loic Senser (Section of MTE) (SGC), January 25, 2019
Supervision: Zhengchao Wang, Yuki Oyama, Michel Bierlaire
Towards a techno-economic evaluation framework for regional train propulsion architectures
Railway can be the most environmental mode for land transport. However the sector faces cost challenges. Apart from main routes, tracks are often not electrified, requiring pollutive and operational expensive diesel propulsion systems. In contrary to road vehicles, there are no commercial large-scale applications of hybrid drivetrains. Furthermore, the dependencies of hybrid drivetrains on energy supplying infrastructure, like overhead wires or recharge points, has not been researched yet. In order to assess the potential of hybrid drivetrains holistically, the project Toolbox for Optimal Railway Propulsion Architectures (TORPA) has set up a framework to define drivetrain solutions, optimize them, and compare them among each other. Prior to this semester project, 58 drivetrain architectures have been defined. One of them was optimized toward the objectives of driving CO2 emissions and investment costs, on a specific use case. However, these bare results were inconclusive and expensive to compute for more than one use case. In this semester project, we downselect the number of possible architectures, outlining the currently most relevant ones. Furthermore, we include infrastructure requirements in the definition of architectures. The metric of investment cost is extended to display the full life cycle costs of architectures. The model of driving CO2 emissions is attempted to display all vehicle life cycle emissions, but left at the current state after qualitatively stating significant CO2 contributors not included in the calculation models. In order to assess the potential of the extended framework, it was required to state test cases, define tangible experiments, and prioritize them. Previously, it is researched which parameters should be chosen to be relevant for such experiments. Thereby, representative generic test tracks are conceived. We choose the question of viability of track electrification as first test case for the software framework. Thus, the currently operating vehicle architectures for diesel or full electric operation are applied. Especially the impacts of currently existing electrification infrastructure and track parameters, like distance and stop frequency, are investigated. Firstly, we imply a track with average constitution and operation. It is found, that the break even of track electrification is reached when a third of it is already electrified due to e. g. intersection with other tracks. Furthermore, we find that stop frequency is decisive for electrification break even: Tracks with shortened stop distance may be cheaper to be operated electrically, even if there is no catenary infrastructure built up yet.
Student: Florian Mueller, January 22, 2019
Supervision: Nour Dougui, Nikola Obrenovic, Michel Bierlaire
Goodness of fit in DCMs
In these last years, with the arrival and wide-spreading of Big data, the discrete choice modeling community has gained access to larger datasets, more computing power and the possibility to drastically increase model complexity. It has therefore become crucial to establish precise measures of the goodness of fit as well as techniques to detect and reduce overfitting. We propose to apply known techniques from machine learning such as cross-validation to discrete choice models. We investigate the existence of overfitting in discrete choice models using a linear and a polynomial multinomial logit model on the Optima dataset. We adopted a train-test approach to evaluate model performance. Model estimates for both models were computed on N = 100, 200 and 500 runs of random train-test splits, using an 80-20 ratio. The results show that the polynomial model has better fit on the training samples, but performs worse than the linear model on the testing set, indicating the presence of overfitting. The use of K-fold cross-validation for simple multinomial logit models has also been explored. However, model estimates obtained using K-fold cross-validation did not differ from the model estimates estimates obtained from a single model fitting, due to underfitting. The distributions of the results obtained on the Optima dataset motivated the exploration of an empirical hypothesis test to determine the presence of overfitting. The test is based on the assumption that training and testing samples of a model which is not overfitting will have the same mean. This test was implemented using the t-test for two samples with unequal variances, also known as Welch test. This empirical test presented in this report did not differentiate the models in terms of overfitting. However, it could be a starting point of our future research, which will be aimed at establishing a statistical definition of overfitting.
Student: Jessica Hopkins (SMA), December 21, 2018
Supervision: Gael Lederrey, Michel Bierlaire, Nicholas Molyneaux
Optimization of school accessibility in developing countries
The Global Program for Safer Schools (GPSS) focuses on risks linked to education infrastructure. The program aims to save lives and reduce the physical impact of disasters on school infrastructure. We are working with the Kyrgyz Republic Government to improve the capacity to respond to disasters, providing safer and quality learning environment for children, and managing the cost of disasters and climate shocks. This semester project/master thesis proposes a methodology to optimize educational infrastructure networks based on accessibility; and proposes solutions to improve the performance of educational infrastructure networks. In other words, it optimizes the location and the dimension of school buildings. Several factors are included such as transportation network, home location of the students and mobility patterns. The results is a methodology to suggest investments to reduce the risks in case of disaster. The cities of Bishkek in the Kyrgyz Republic is used as a case study. Methodologies from optimization, programming (Java) and geographical information system (GIS) will be used for completing this project.
Student: Yassine El Ouazzani, Oliver Mathias Buschor (SGC), December 21, 2018
Supervision: Riccardo Scarinci, Meritxell Pacheco, Michel Bierlaire
In collaboration with World Bank
Locating charging station for electric taxis
This project aims at locating the position of charging station for electric taxis. Currently, transportation is transforming to be more sustainable. The electric vehicle is a promising way to achieve such a goal. However, charger unavailability is among one of the main issues that hindered the adoption of the electric vehicle. To contribute to mitigating such issue, this project is de ned accordingly to optimize the charging station location. The location problem is a classic problem in the operational research domain. It is usually solved using a p-median, p-mean, set covering, or flow covering method. The goal of this research is to apply these methods to decide the location of charging stations in San Francisco using taxi tracking data and compare the di erence in the end. This project is composed of three stages. First, the student should work on mapping the taxi tracking data on the map using map-matching techniques. Then, the student should apply point-based location methods (p-median, p- mean, set-covering) to find the optimal location. The student is not required to apply all the relevant methods but chooses one or two in which he is interested. After, the students will apply and characterize a flow-based method. Finally, an analysis of the results is performed to evaluate the results obtained.
Student: Oliver Mathias Buschor, Younes Bensaid, December 21, 2018
Supervision: Zhengchao Wang, Meritxell Pacheco, Michel Bierlaire, not applicable
Sustainable & intelligent transportation evaluation and plan
This project is composed of two stages. First, the student will search for and make a summary of emerging transportation technologies that are proposed in both literature and industries. After, the student will learn to come up with feasible and promising transportation systems composed of the emerging technologies, and then quantitatively evaluate the proposed systems based on travel demand simulations.
Student: David Gunter, July 20, 2018
Supervision: Zhengchao Wang, Yuki Oyama
Cost Reduction Uisng Passenger Centric Timetabling
"To design their timetables, train-operating companies mostly focus on operational aspects and cost. In Switzerland, a paradigm of transportation planning is to create regular-interval timetables (a.k.a. cyclic timetables) which aim for maximal transfer connections, simplicity and hence user friendlyness. SBB uses travel simulation models to predict the impact of timetable changes on travel demand and revenue. Mathematical timetable optimization methods are not yet used by SBB. But a recent EPFL thesis (2016) shows that the timetable itself has a significant impact on the performance of the operator in terms of the number of transported passengers: a timetable design that considers the behavior of passengers leads to higher revenue(s), market share(s), higher value of passenger-km etc. In this project, the aim is to use the optimization methods developed by the TRANSP-OR Lab consisting in combining cyclic and non-cyclic timetables and apply it to the Swiss Federal Railways timetable design. The goal is to evaluate the performance of the current Swiss interval timetable and to compare it to the optimal one. Suitability of hybrid timetables for the Swiss railway network will be investigated".
Student: Robert Abboud (SMGT), June 29, 2018
Supervision: Virginie Lurkin, Michel Bierlaire
Design of a stated-preferences survey for a high-speed vacuum transportation mode
Since the high-speed vacuum transportation technology represents an innovative transportation mode, it is necessary to obtain data from surveys of hypothetical market/situations, the so-called stated-preferences (SP) surveys. The goal of this project is to design an SP questionnaire to evaluate the impact of this innovative transportation mode and to measure some indicators (such as the willingness-to-pay). A pilot test of the preliminary survey in a small sample will be carried out in order to evaluate the quality of the questionnaire. The main findings will be used to define the final survey that will be distributed to a representative sample of the population by a specialized company.
Student: Thibaut Richard et Mart Montesinos Ferrer, June 20, 2018
Supervision: Meritxell Pacheco, Yuki Oyama, Michel Bierlaire
Modelling competition in demand-based optimization models
The project aims at studying and understanding the interactions between supply and demand in an oligopolistic market, in which multiple operators compete for the same pool of customers. This is a common situation in the transportation sector as well as in other markets. Operators take the supply-side decisions that optimize their own performance function (e.g. maximization of revenues or profits). Such decisions are influenced both by the decisions of their competitors and by the preferences of the customers who consider purchasing one of the services offered on the market. The latter ones are modelled at a disaggregate level according to the random utility theory. The starting point of the project is a recent modelling framework that allows to include any random utility model in a mixed integer optimization formulation. In this framework, a single operator exploits its knowledge of the demand to maximize its objective function, while assuming that the decisions of its competitors are held fixed. The goal of this project is to extend the existing framework by including the case in which two operators simultaneously optimize their decisions. Such problem falls into the category of competitive games called two-player non-cooperative games. The student will refer to the concept of Nash equilibrium to answer questions such as "how does competition affect prices?", "what products should each competitor offer and in which quantity?", "should two competitors fight to attract the same group of customers or should they each target a different market segment?", among others.
Student: Charlotte Darn (SMA), June 19, 2018
Supervision: Stefano Bortolomiol, Virginie Lurkin, Michel Bierlaire
A solution approach for the Multicommodity Flow Problem within rail freight transportation
This project aims at studying the Multicommodity Flow Problem (MFP), which has several real life applications, particularly within the transportation sector. The project has been inspired by an ongoing research project conducted by TRANSP-OR in collaboration with SBB Cargo. In the context of railway freight operations, companies have limited availability of both railway lines and trains. Therefore, they have the goal to optimize as much as possible the usage of their resources in order to minimize the cost of transporting the goods. Such optimization problem can be modelled as a MFP. This project is composed of two stages. First, the student will learn the mathematical model of the MFP, in its two variants: node-arc formulation and path-based formulation. The two variants will be coded and tested on small instances by using a Mixed-Integer Linear Programming (MILP) solver such as CPLEX. Further, the student will implement a solution method to solve larger instances of the problem, by using either an exact method (such as column generation) or a heuristic method.
Student: Nicolas Pradignac (SIN), June 19, 2018
Supervision: Stefano Bortolomiol, Nikola Obrenovic, Michel Bierlaire
In collaboration with SBB Cargo
Development of a heuristic algorithm for a hub location problem
In railway networks, the location of marshalling and shunting yards, i.e. facilities used for sorting and consolidation of transported goods, determines the costs of cargo transport to a great extent. Hence, each railway company is highly motivated to determine the optimal location of these yards. Such task can be represented as a multi-level facility location problem or a hub location problem. The aim of this project is the development of an heuristic algorithm for solving a facility location problem with facilities split in multiple hierarchical layers. The project will be a part of a larger research project done in the cooperation with SBB. The student will start from understanding the exact mathematical model of the mentioned problem and encode it into a heuristic algorithm, e.g. a local search algorithm or VNS. Also, the student will have to solve the problem on a smaller data set using an exact method and compare the solution with the results of the heuristic algorithm. The student needs to have good programming skills (Java or some other OO language) and knowledge of mixed integer linear programming and heuristics.
Student: Thibaut Guillaume Marie Richard, June 19, 2018
Supervision: Nikola Obrenovic, Nicholas Molyneaux, Michel Bierlaire
In collaboration with SBB Cargo
Safer schools thanks to an improved transport access in developing countries
The Global Program for Safer Schools (GPSS) focuses on risks linked to education infrastructure. The program aims to save lives and reduce the physical impact of disasters on school infrastructure. We are working with the Kyrgyz Republic Government to improve the capacity to respond to disasters, providing safer and quality learning environment for children, and managing the cost of disasters and climate shocks. This semester project/master thesis proposes a methodology to assess and optimize educational infrastructure networks based on accessibility; and proposes solutions to improve the performance of educational infrastructure networks. In other words, it evaluates the risks based on the school network, including parameters like transportation network, home location of the students and mobility patterns. The results is a methodology to suggest investments to reduce the risks in case of disaster. Two case studies will be carry on in the cities of Bishkek and Osh in the Kyrgyz Republic. Methodologies from transportation studies, geographical information system (GIS) and programming will be used for completing this project.
Student: Zora Oswald, June 16, 2018
Supervision: Riccardo Scarinci, Meritxell Pacheco, Michel Bierlaire
In collaboration with The World Bank
Conduite Automatique des Trains entre Neuchtel et La Chaux-de-Fonds
L'European Train Control System Level 2 (ETCS_L2) autorise la conduite automatique des trains via le "paquet 44". Le projet consiste introduire dans ce "paquet" l'ensemble des instructions ncessaires pour aller, en conduite automatique, de A B tout en conservant un conducteur bord (GoA2). Le parcours entre Neuchtel et La Chaux-de-Fonds via un nouveau trac servira de "dmonstrateur".
Student: Guillaume Sauvin (SGC), June 08, 2018
Supervision: Daniel Emery
Usage d'un sous-cantonnement pour densifier la circulation aux abords d'un noeud ferroviaire important
Le projet consiste, sur un dcoupage des cantons en sous-cantons virtuels, faire circuler plusieurs combinaisons de trains signalisation latrale et de trains signalisation en cabine (ETCS_L2/3) pour dterminer les gains en dbit pouvant tre attendus.
Student: Clo Lafaye, June 08, 2018
Supervision: Daniel Emery
Offre nationale ferroviaire 2035 avec un tronon SwissMetro
Le projet consiste, notamment, dvelopper des scnarios de rseau pour le mode de transport SwissMetro, en dfinir les caractristiques principales ainsi qu'un ordre de priorit de ralisation des tronons. Dans une seconde tape, l'horaire GK12 sera adapt pour tenir compte de la prsence du tronon-pilote du(des) scnario(s) le(s) plus prometteur(s). Une comparaison exhaustive des avantages/inconvnients des horaires GK12 original et GK12 modifi(s) clturera ce travail.
Student: Lauriane Masson (SGC), June 08, 2018
Supervision: Daniel Emery
Augmentation de la rserve de capacit m1 des TL par actions limites au matriel roulant
Le mtro m1 des Transports Publics Lausannois (TL) relie le Centre-Ville de Lausanne la Gare de Renens situe dans l'Ouest Lausannois en passant notamment par l'Universit (UNIL) et l'cole Polytechnique Fdrale de Lausanne (EPFL). En heure de pointe en priode estudiantine il circule avec des intervalles de 5 minutes, soit les intervalles les plus faibles possibles compte tenu de l'infrastructure actuelle simple voie. Depuis de nombreuses anne, de nombreuses mesures touchant la demande ont t prises en vue de rduire les pointes afin de d-saturer cette ligne (dcalage des heures de dbut de cours entre UNIL et EPFL, nouvelle ligne de bus passant par Renens-Gare et les Hautes coles, ). Toutefois, avec la croissance continue de la population se rendant sur les sites universitaires, la saturation se profile moyen terme. Cette pr-tude de master quantifiera l'apport des moyens d'actions pour accrotre la capacit du matriel roulant sans avoir a modifier l'infrastructure.
Student: David Moy de Vitry (SGC), June 08, 2018
Supervision: Daniel Emery
Qualit de l'offre nationale des projets d'horaire EA2030 et EA2035
Des deux projets d'horaire nationaux (EA2030 et EA2035) sont disposition des dcideurs pour choisir entre deux variantes d'amnagement du rseau ferroviaires suisse. Le projet consiste notamment valuer la qualit de l'offre nationale offerte par ces projets en terme de temps de parcours et de frquence, l'horaire 2018 servant de rfrence. Les valeurs absolues et relatives des critres retenus (temps de parcours, frquence, nombre de correspondance, ...) seront notamment disposes sous la forme de matrice OD, chaque O(resp. D) tant une gare significative sur la plan dmographique et politique (p.ex. chef-lieu de canton)
Student: Zora Oswald (SGC), June 08, 2018
Supervision: Daniel Emery
Offre nationale ferroviaire 2035 avec tronons SwissMetro
Un projet d'horaire ferroviaire national 2035 a t publi avec les amnagements des infrastructures ncessaires exploiter un tel horaire. Le projet consiste dvelopper des variantes de ralisation de tronons SwissMetro. Pour chacune de ces variantes il sera recens les amnagement d'infrastructure prvus devenant superflus d'une part, et le projet d'horaire 2035 sera alors adapt la prsence de ces tronons d'autre part. Les avantages/inconvnients de chaque variante seront compars, le projet d'amnagement 2035 et l'horaire 2035 sans SwissMetro servant de rfrence.
Student: Lauriane Masson (SGC), June 08, 2018
Supervision: Daniel Emery
Automatic utility specification using machine learning techniques
The objective of this project is to help building a utility function using different Data Analysis functions. It is well known that modeler spend a lot of time creating their utility function for DCMs. The goal of this project is to build an automated procedure to generate Utility functions based on the data. The tools used are Data Analysis and Machine Learning (Decision Tree, Random Forest, Linear Regression, Clustering, etc.). The utility functions found using these tools will then be compared to standard and more advanced utility functions. Another direction that we can take is using Machine Learning to extract the possible nests for a Nested Logit Model. This direction, however, requires a deeper knowledge in Discrete Choice Modelling. We are open to the discussion concerning the available directions.
Student: Nicola Ortelli, June 08, 2018
Supervision: Gael Lederrey, Tim Hillel, Virginie Lurkin
Building offline and online optimization algorithms for dispatchment of teams at Nez Rouge.
Nez Rouge (http://nezrouge.ch) is a swiss charitable association with the purpose of driving people back home safely during the cold nights of December. They have to find many volunteers across the whole country and the demand never stops growing. Instead of helping them find new volunteers, I propose to solve one difficult task: dispatching efficiently the volunteers, the idea being to reduce the waiting-time of customers. Using data provided by Nez Rouge, we will first build an offline optimization problem to find a lower bound on the estimated waiting-time and compare it to the actual waiting-time. The second step will be to build an online optimization problem using the API of OpenStreetMap to solve this problem in real-time. The third and final step will be to build a webpage for the volunteers at Nez Rouge. Note: This project can be done as a master project or split between multiple semester projects, depending on the number of credits you have to do.
Student: Colin Ducommun, June 08, 2018
Supervision: Gael Lederrey, Nicholas Molyneaux, Michel Bierlaire
On the optimization of CAPEX and OPEX for the design of a full electric large capacity urban bus system
During the last few decades, environmental impact of the fossil fuel-based transportation infrastructure has led to renewed interest in electric transportation infrastructure, especially in urban public mass-transportation sector. The deployment of battery-powered electric bus systems within the public transportation sector plays an important role to increase energy efficiency and to abate emissions. RAn efficient feeding stations installation and an appropriate dimensioning of battery capacity are crucial to minimize the total cost of ownership for the citywide bus transportation net- work. The objective of this project is to extend an existing optimization model to the multiple lines case and to come up with an objective function that better reflect the real costs incurred by the operator. The objective function should include Capital Expenditures (CAPEX) and Operational Expenditures (OPEX). For the multiple line case, the central issue is to deal with the feeding stations that are shared among different lines.
Student: Guillaume Mollard (CSE), January 31, 2018
Supervision: Virginie Lurkin, Stefan Binder, Michel Bierlaire
Integrating demand and supply in the context of airlines
Student: Thibaut Richard et Gabriel Curis (SGC), January 24, 2018
Supervision: Meritxell Pacheco, Anna Fernandez Antolin, Michel Bierlaire
Infrastructure ferroviaire entre Lausanne et Genve l'horizon 2030
Le projet consiste dterminer les infrastructures ncessaires et les horaires possibles sur la ligne Lausanne-Genve l'horizon 2030. Un projet d'offre 2030 prvoit pas moins de dix paires de trains voyageurs (4 IC, 2 IR, 4 RE). L'OFT pour sa part souhaite qu'une paire de trains marchandises au minimum puisse circuler chaque heure.
Student: Clo Lafaye et Julien Thiriot (SGC), January 12, 2018
Supervision: Daniel Emery
Offre 2030 entre Berne et Lausanne rpondant aux attentes des cantons (BE, FR et VD)
Student: Benot Corday (SGC), January 12, 2018
Supervision: Daniel Emery, Jean-Daniel BURI
Amnagements futurs du complexe ferroviaire de Clermont-Ferrand
Le travail consiste premirement tablir un tat des lieux (plan et caractristiques des voies, postes daiguillages, IFTE, GOV, types de mouvements et volumtrie, roulements de matriel, quipements de maintenance et de remisage, etc.) Une analyse de ladquation des quipements aux besoins engendrs par le remplacement des rames tractes par des rames automotrices sera alors mene; et dventuelles mesures seront proposes. Si le temps le permet, les consquences sur les GOV de Clermont seront estimes si un scnario ambitieux dlectrification devait voir le jour autour de Clermont (Volvic/Le Cendre/Aulnat). de rfrence.
Student: Axel Valentin (SGC), January 12, 2018
Supervision: Daniel Emery
Mobilit lors des JO dhiver et perspectives pour Sion 2026
La pr-tude se propose premirement de recenser les dmarches dorganisation de la mobilit lors de JO prsentes par les villes candidates rcentes, notamment Sion 2002 et Torino 2006, et de dterminer et de rcolter les valeurs dimensionnantes de mobilit en lien avec les diffrentes disciplines sportives. Dans une seconde partie, la pr-tude se focalisera sur le cas de la ville candidate Sion 2026. Elle prsentera en particulier un panorama de loffre TP actuelle pour atteindre actuellement les sites retenus, ainsi que loffre envisage par le programme PRODES 2025. Le Projet De Master (PDM) proprement dit cherchera premirement atteindre les objectifs de la pr-tude non totalement atteints. Son objectif principal sera de gnrer des concepts doffre TC, voire mme des variantes dhoraires. Les lignes tudier plus particulirement seront choisies en fvrier 2018. La liste de ces lignes, non exhaustive, comporte notamment les lignes extrieures (Olympic Ring, raccords aux aroports internationaux GVA et ZRH) et les lignes intrieures au Valais (Brig-Ulrichen, Martigny-Le Chble,).
Student: Clment Sintes (SGC), January 12, 2018
Supervision: Daniel Emery, Stefano MANELLI Pierre FAVRE
From public transport vehicles to pedestrian flows.
The objective of this semester project is to extend the notion of "train induced flows" to other modes of public transport. When public transport vehicles arrive in the station, many passengers disembark. From the infrastructure's point-of-view, there is a flow of pedestrians arriving. The characteristics of such flows depend on the vehicle's characteristics (door width for example) and the characteristics of the infrastructure (width of the corridors for example). The student will explore the different interactions and characterize the different flows by using models similar to the "train induced flows" model, already published. The specifications will be different based on the various modes and platform setups. The second objective is to explore the feasibility of using a Poisson distribution to model the pedestrians arriving onto platforms. Similarly to alighting flows, different configurations can lead to different model specifications. The student will start by exploring the literature then, based on the pre-existing models, some new formulations can be proposed. Data is available for calibration of pedestrian flows boarding trains. Synthetic data can be used for the calibration of other models. This project can be done during the spring 2018 semester.
Student: Rodolphe Farrando, January 12, 2018
Supervision: Nicholas Molyneaux, Gael Lederrey, Michel Bierlaire
Alternative activity pattern generation for stated preference surveys
Student: Nicola Marco Ortelli (SGC), December 22, 2017
Supervision: Anna Fernandez Antolin, Gael Lederrey, Michel Bierlaire
Identifying the objectives of car drivers route choice behavior
The goal of this project is to shed light on the underlying objectives of car drivers' behavior --- pertaining to their route choices --- when traveling from one location of a transportation network to another. The student will use clustering methods, defined on the basis of different measures of similarity, that are associated with potential objectives such as the minimization of the travel time, the length or the complexity of a route. The results of the clustering will be analyzed to draw insights into the use of specific objectives depending on the features of the trip, such as the origin and destination points, and the departure time. The project involves an initial stage of data processing. The student needs to have good programming skills (Matlab or equivalent) and good knowledge of statistics. Familiarity with SQL is a plus.
Student: Nicola Marco Ortelli, December 22, 2017
Supervision: Evanthia Kazagli, Marija Nikolic, Michel Bierlaire
Analysis of pedestrian group behavior based on tracking data and pattern recognition methods
The objective of this project is to analyze the group behavior among pedestrians based on individual trajectory data. The data is collected in Lausanne train station, where a large-scale network of smart sensors has been used to track pedestrians. The project aims to improve the understanding of group behavior among pedestrians and its impact on pedestrian dynamics. It involves the following steps: (i) Development/selection of suitable methods (e.g. data mining/pattern recognition) for identification of individuals walking together, based on pedestrian trajectory data, and their implementation. (ii) Analysis focused on the behavior of identified groups; (iii) Comparison of the findings with the existing empirical basis, as well as with the proposed theories and models that take group dynamics into account. The student needs to have good programming skills (Scala/Matlab), and knowledge of statistical analysis.
Student: Montesinos Ferrer Mart, December 22, 2017
Supervision: Marija Nikolic, Evanthia Kazagli, Zhengchao Wang
Modeling purchases of new cars for 2015: a comparison between countries
We are interested in analyzing the behavior of people when they face the decision to buy a new car. We have a large dataset containing information of new car purchases, where respondents answered to several questions related to their socioeconomic characteristics and to the attributes of their recently purchased car. We have developed a framework that has been applied to the data corresponding to France in 2014. The objective of this semester project is to apply it to 2015 for several countries (France, Germany, Italy and Spain) and to obtain and predict market shares in each of the countries as well as to compute willingness to pay towards different attributes. Moreover, the student will analyze the difference between the different countries and between the different years.
Student: Mart Montesinos (SGC), June 19, 2017
Supervision: Anna Fernandez Antolin, Meritxell Pacheco, Michel Bierlaire
Planning of feeding station installment for a full electric large capacity urban bus system
The integration of customer behavioural models in optimization provides a better understanding of the preferences of clients (the demand) to policy makers while planning for their systems (the supply). These preferences are formalized with discrete choice models, and the corresponding optimization models where supply and demand closely interact are associated with (mixed) integer optimization problems. One concrete application of this integration consists of an operator selling services to a market, each service at a given price to a finite number of customers, called the capacity of the service. We are interested in finding the best strategy in terms of pricing and capacity allocation in order to maximize the revenues of the operator. The project objectives can be tailored for both Master or Semester project students.
Student: Adrien Ruault (SIN), June 16, 2017
Supervision: Virginie Lurkin, Marija Nikolic, Michel Bierlaire
Speed profile of an innovative catenary-free electric bus
The TOSA bus system is a revolutionary catenary-free electric bus concept that includes small short-range on-board batteries and a series of fully automated fast charging stations installed at some bus stops. The automatic fast-charging stations partially replenish the bus batteries in a few seconds whenever a bus arrives at the bus stops, while avoiding any interference with the bus schedules and operations. This system has been defined during the project myTOSA 1.0 and implemented in a pilot test in the city of Geneva in 2013. The project myTOSA 1.0 is currently extended by the project myTOSA 2.0. myTOSA 2.0 is composed of several modules, the most relevant for this semester project is the traffic simulation. The traffic simulation represents the movement of the buses in a network, and requires a representation of the infrastructure as an input. This project aims to define the typical speed profile of an electric bus moving in an urban network. The speed profile includes acceleration and deceleration patters for different driver behaviour, network configurations and traffic levels. The influence of traffic lights, bus priority, roundabouts on the speed profile should be investigated. The first step to achieve this aim is to define the probability distribution of the number of intersections crossed by the bus. This step is necessary to complete this project successfully. Further steps in the directions of the definition of the speed profile are considerate optional. Methodologies from transportation modelling, geographical information system (GIS) and programming (MATLAB) will be used for completing this project.
Student: Valentin Axel Olivier & Nicolet Adrien (SGC), June 16, 2017
Supervision: Riccardo Scarinci, Virginie Lurkin, Michel Bierlaire
Organisation de lamlioration de la performance de loffre ferroviaire dans les mtropoles dAuvergne Rhne Alpes: Retour dexprience du nud ferroviaire lyonnais et enseignements tirer pour les autres mtropoles
Ltat franais et SNCF Rseau ont sign un contrat de performance permettant de donner une trajectoire financire pour la dcennie venir et qui demande dorganiser lamlioration de la performance, notamment de la rgularit. Dautre part, dans le cadre de Schma Rgionaux dAmnagement, de Dveloppement Durable et dEgalit des Territoires (SRADDET), SNCF Rseau entend positionner linfrastructure ferroviaire comme vritable pivot de la mobilit au niveau rgional. Ltude (pr-tude et travail pratique de Master) doit permettre de sapproprier les mthodes de travail utilises pour le NFL ( audit, diagnostic, bilan des circulations, mobilisation des acteurs internes ou externes ) et les outils correspondants permettant tous les acteurs de participer lamlioration de la performance en appliquant des plans daction court et moyen terme ( organisation du projets, identification des phases dcisives, formalisation ders plans daction, ...) Une analyse dans la cohrence des plans daction court terme mais galement long terme sera dvelopp en prenant en compte la stabilit des volutions des hypothses de loffre ferroviaire. La mthode de travail ainsi conceptualise par un travail personnel et autonome et une reformulation des facteurs cls de succs et les outils correspondants seront appliqus, titre de dmonstration pour la mtropole grenobloise ou clermontoise. Des plans dactions court, moyen et long terme seront dduits et proposs
Student: Jean-Baptiste Landes (SGC), June 09, 2017
Supervision: Daniel Emery
Emotions and risky discrete choices
This CSE project aims at understanding, analyzing and quantitative modeling of the roles of emotions in risky choice making
Student: Le Sueur Ccile Christianne Anny Dborah Suzanne (SMA), June 02, 2017
Supervision: Matthieu de Lapparent, Michel Bierlaire
Modeling public transport transfers in the new Lausanne train station
Exploration of the tracking data collected in the main station of Basel. The objective is to calculate passenger-centric variability indicators and perform some simple modelling tasks.
Student: Jos Ramn Rodriguez, June 02, 2017
Supervision: Nicholas Molyneaux, Riccardo Scarinci, Michel Bierlaire
Building an integrated model for modelling pedestrian movements inside hubs
The objective of this project is to build a simulator for measuring the impact of management strategies inside transportation hubs.
Student: Charles Jeanbart, June 02, 2017
Supervision: Nicholas Molyneaux, Riccardo Scarinci, Michel Bierlaire
Models for pedestrian movements based on integrated and sequential clustering
The focus of this study is the modeling of speed-density relationship for pedestrian movements using the potential of available data. The data set considered in this project contains pedestrian trajectories collected in the train station in Lausanne. To track pedestrians, a large-scale network of tracking sensors is installed in the station (Alahi et al., 2014). The analysis we have performed reveals a high scatter in the data. To characterize the observed scatter we have developed a multi-class model of the speed-density relationship based on the latent class modeling approach (Nikolic et al., 2017). The latent class approach allows for modeling the segmentation in the population and the movement behavior simultaneously. The aim of this project is the derivation of the model representing speed-density relationship based on a two-stage (sequential) approach. The first stage involves segmentation of pedestrian trajectory data, by using machine learning techniques (clustering). In the second stage, a separate speed-density model is to be estimated for each cluster discovered in the first stage. The performance of the approach will be tested using real data, and compared to the more integrated, latent class approach.
Student: Konde Romain Olivier Bondo, June 02, 2017
Supervision: Marija Nikolic, Iliya Markov, Michel Bierlaire
Investigating the role of attitudes in the purchase of new cars
Attitudes and perceptions play an important role in decision-making processes. We are interested in new car purchases, and how attitudes and perceptions play a role in this context. We have a large dataset containing information of new car purchases, where respondents answered to several attitudinal questions. The objective of this semester project is to analyze these attitudinal questions, starting with principal component analysis. The study will be expanded to see how these attitudes affect purchases of new cars, and of electric vehicles in particular.
Student: Nicola Ortelli, January 23, 2017
Supervision: Anna Fernandez Antolin, Meritxell Pacheco, Michel Bierlaire
Mobilit Rail+Bus 2020 dans les Alpes vaudoises
Une communaut dtudes, comprenant notamment la CITAV, a tudi de manire globale un territoire gographique appel Alpes vaudoises 2020 . Le volet mobilit na t trait que de manire globale, tel que lon peut notamment le constater dans le rapport final de juillet 2013. Aprs lecture attentive des documents Alpes vaudoises 2020 et des horaires actuels, les tudiants dvelopperont et valueront des variantes doffre Rail+Bus sur la base du projet dhoraire trains 2025 le plus actuel.
Student: David Moy de Vitry Jean-Baptiste Landes, January 13, 2017
Supervision: Daniel Emery
Rgulation de vitesse par la signalisation pour garantir un croisement actif
Dans le cadre de la densification dhoraires cadencs sur lignes simple voie, il y a souvent ncessit que les trains se croisent hors gare. Il est donc intressant de prvoir un lot de croisement actif, cest--dire sans quaucun des deux trains nait besoin de sarrter. L'tude consiste premirement tablir une dmarche pour tablir les diffrentes topologies devant tre tudies. Dans la partie principale, le projet consistera faire varier de manire systmatique et logique, sur un tronon de ligne suisse voie troite fictif, la vitesse de ligne, la longueur de l'ilot deux voies, les endroits de dtection des trains, les endroits de rduction de la vitesse pour le train en avance sur l'autre, et la valeur de cette(ces) rduction(s) de vitesse. La signalisation sera de type latral trois aspects avec, si besoin, chiffre complmentaire d'annonce ou d'excution de vitesse. L'outil de simulation OpenTrack sera utilis.
Student: Benot Corday Marc Zimmermann, January 13, 2017
Supervision: Daniel Emery
Price of anarchy in public transit networks
Student: Marc-Edouard Schultheiss (SGC), December 23, 2016
Supervision: Stefan Binder, Toms Robenek, Michel Bierlaire
Une modlisation de la grande mobilit
Student: Pauline Hosotte, December 23, 2016
Supervision: Matthieu de Lapparent, Michel Bierlaire, E. Ravalet, V. Kaufman (LASUR)
Grandes mobilits rversibles. Approche par la modlisation dynamique de choix discrets.
Student: Alexis Gumy, December 23, 2016
Supervision: Matthieu de Lapparent, Michel Bierlaire, E. Ravalet, V. Kaufman (LASUR)
tude dtaille des contours de rfrence ferroviaires et drogations
Dans le cadre du projet Clip-Air (http://clipair.epfl.ch/) il est envisag d'acheminer des capsules par le rail. La pr-tude au projet de Master consiste tudier trs prcisment les gabarits de rfrence ferroviaires pour voie cartement UIC tant au niveau suisse, qu'europen et amricain non seulement en alignement mais aussi en courbe. Les relations avec les gabarits limites des obstacles seront mises en vidence. Les procdures suisses pour les transports exceptionnels par rail sur voie cartement UIC seront dcrites.
Student: Martin Ellwanger (SGC), December 23, 2016
Supervision: Daniel Emery
4. Modeling Route Choice in Quebec City Using Mental Representations
In this project, we build upon previous work proposed by Kazagli et al., 2016 to model route choice for a big network and dataset concerning the city of Quebec. We aim first to define the mental representation items (MRIs) for the case study and then to derive operational route choice models based on them. A GPS dataset from the city of Quebec is available and will be used for this purpose. The focus of the project is methodological.
Student: Mathieu Plourde, December 23, 2016
Supervision: Evanthia Kazagli, Matthieu de Lapparent, Michel Bierlaire
Passengers' connection time preferences in airline itinerary choice
As noted by Theis et al. (2006), "Network airlines traditionally attempt to minimize passenger connecting times at hub airports, assuming that passengers prefer minimum scheduled elapsed time for their trips. However, minimizing connecting times creates schedule peaks at hub airports. These peaks are extremely cost-intensive in terms of additional personnel, resources, runway capacity, and schedule recovery. Consequently, passenger connecting times should be minimized only if the anticipated revenue gain of minimizing passenger connection times is larger than the increase." Prior work has used (small) stated preference surveys to estimate customers' connection time preferences, and explicitly whether this function is nonlinear with connection time (i.e., customers avoid very short connections and very long connections). However, little is known as to how these connection time preferences vary as a function of other characteristics, including flight frequency (or schedule delay), length of haul, whether the flight is the last flight of the day, prior on-time performance of flight legs. As part of this project, the student will estimate airline itinerary choice models using a large ticketing database provided by the Airlines Reporting Corporation (ARC). The student will build off of prior MNL models that have been estimated which have corrected for price endogeneity and focus explicitly on refining the utility function related to connection time preferences.
Student: Elisabeth Zbinden (SGC), December 23, 2016
Supervision: Virginie Lurkin, Matthieu de Lapparent, Michel Bierlaire, GARROW
Demand based rolling stock allocation
Providing a high level of service for the passengers is one of the most important requirements of passenger railway company. In practice during rush hours passengers cannot be transported according to usual service standards because of a shortage of the rolling stock capacity. The purpose of this pre-project is to determine an efficient method to allocate the rolling stock taking into account passengers demand.
Student: Salma Derouiche (SIN), December 23, 2016
Supervision: Yousef Maknoon, Toms Robenek, Michel Bierlaire, Simon Landureau
Generalization and policy analysis for a rich inventory routing problem
We solve a rich logistical problem inspired from practice, in which a set of trucks is used for collecting recyclable waste from large containers over a finite planning horizon. Each container is equipped with a sensor, which communicates its level at the start of the day. Given a history of observations, a forecasting model is used to estimate the point demand forecasts as well as a forecasting error representing the level of uncertainty. The problem falls under the framework of the stochastic inventory routing problem. We introduce dynamic probabilistic information in the solution process, which impacts the cost through the probability of container overflows on future days and the probability of route failures. To solve the problem, we implement an adaptive large neighborhood search algorithm, which integrates a specialized forecasting model, tested and validated on real data. The student will analyze the performance of the algorithm and its results for various scenarios and collection and routing policies. In particular, we would like to see the impact on the best solution of the full probabilistic model with various cost parameters against simpler policies, such as one using buffer truck and container capacities to handle stochastic demand. We would like to apply the model on a rolling horizon basis and analyze the expected value of perfect information, the properties of the stochastic solution, etc. Other practical features such as open tours, multi-product problems, etc. may also be explored. We can also work on improving the algorithmic performance per se through new operators and more intelligent search strategies in the current implementation. This project is appropriate for a student with excellent coding skills in Java and knowledge of a good software for plotting and data analysis such as Matlab or R. The focus of the project will be suited to the student's interests, coding abilities, and the number of credits he or she needs.
Student: Prisca Aeby (SIN), December 23, 2016
Supervision: Iliya Markov, Yousef Maknoon, Michel Bierlaire
Measure of user-oriented service variability in hubs
This project's goal is to explore the service variability (reliability) in public transport hubs. Three operator oriented service characteristics are common in literature (robustness, reliability and punctuality). As these measures are operator oriented, they might not reflect actual level-of-service (LOS) experienced by users (pedestrians). To account for this difference, user-centered measures will be defined then observed using an already secured data set. Some examples of variability measures are OD travel time, passenger transfer times, passenger waiting times, train punctuality, etc. The main data sets used for this project is the pedestrian tracking data from the Lausanne and Basel train stations and possibly some effective train time tables.
Student: Anouk Allenspach and Dieynaba Dia (SGC), December 23, 2016
Supervision: Nicholas Molyneaux, Riccardo Scarinci, Michel Bierlaire
Accounting for dynamics in pedestrian multi-class speed-density relationship
The relationship between speed and density plays an important role in modeling of pedestrian traffic. It is useful for planning and design of pedestrian facilities, and it is also a required input or calibration criterion for models of pedestrian dynamics. The relationship is specified under the assumption that the traffic system is at equilibrium (stationary and homogenous). The analysis we have performed, based on the data collected in the Lausanne train station, rules out the use of a unique equilibrium relationship due to a high scatter in the data. This scatter may be explained by the violation of the equilibrium assumptions, as documented in the literature. To characterize the observed scatter we have developed a multi-class model of the speed-density relationship based on the latent class modeling approach. The model is derived by relaxing the homogeneity assumption of equilibrium relationships. It is assumed that pedestrian population is heterogeneous (e.g. different trip purpose, different time to departure, etc.) and that this heterogeneity leads to the existence of multiple pedestrian classes that are characterized by different behavior. There are two specification issues related to the panel data set (data collected over multiple time periods for the same sample of individuals) that we use in our analysis. The first is serial correlation across the observations of the same individual due to unobserved individual factors that persist over time. The second is related to dynamics, meaning that the speed in one period may depend on the speed values in the past. We have addressed the first issue by introducing an agent effect in the model that captures individual related unobserved factors. We term this model the static model with agent effect. The aim of this project will be to deal with the second issue, or dynamics. We will start with the simplified assumption that the speed at time t is influenced by the speed at time t - 1 only. We term such a model a dynamic model with agent effect. The performance of the approach will be tested using real data.
Student: Marc-Edouard Schultheiss (SGC), December 23, 2016
Supervision: Marija Nikolic, Matthieu de Lapparent, Michel Bierlaire
Pricing and capacity allocation strategies for a demand-based revenues maximization problem
The integration of customer behavioural models in optimization provides a better understanding of the preferences of clients (the demand) to policy makers while planning for their systems (the supply). These preferences are formalized with discrete choice models, and the corresponding optimization models where supply and demand closely interact are associated with (mixed) integer optimization problems. One concrete application of this integration consists of an operator selling services to a market, each service at a given price to a finite number of customers, called the capacity of the service. We are interested in finding the best strategy in terms of pricing and capacity allocation in order to maximize the revenues of the operator. The project objectives can be tailored for both Master or Semester project students.
Student: Jonathan Lachkar (SGC), December 23, 2016
Supervision: Meritxell Pacheco, Virginie Lurkin, Michel Bierlaire
Energy consumption of an innovative catenary-free electric bus
Electric busses help to decrease pollution in city centers. However, they need to be constantly attached to a power source. This limits their mobility and brings visual pollution due to continuous catenaries throughout the city. Focus of this project is a revolutionary catenaryfree electrical bus that includes short-range on-board batteries and a series of fast charging stations installed at some bus stops capable to charge the bus batteries in a few seconds without interference with the bus schedules. This project aims to develop a traffic simulation of this electric bus moving on a simple road network. A discrete event simulation representing the fundamental components of the system (e.g. bus, passengers load, on-board battery, charging station, storage level) should be developed, calibrated and validated using traffic data (such as number passengers on board, traffic and congestion information). Then, key indexes representing the system performance should be defined (e.g. travel time, dwell time, battery charge level). Methodologies from transportation modelling, simulation and programming (Java) will be used for completing this project.
Student: Xiaoran Yu, December 23, 2016
Supervision: Riccardo Scarinci, Yousef Maknoon, Michel Bierlaire
Modeling evolution of sales of alternative fuel vehicles in Europe
Student: Christophe Paillard (SMA), June 17, 2016
Supervision: Matthieu de Lapparent, Michel Bierlaire
Analysis of External Effects on Cyclicity in Passenger Railway Service
In the passenger railway service, cyclicity is in general perceived as a beneficial attribute. The main allegation being memorability of such timetables by the passengers. However no quantitative proof has been presented to support such statement. In this project, the student will become familiar with a passenger satisfaction (that is based on utility theory) and further extend this concept by the attribute expressing the perception of the cyclicity. Student will then solve an optimization problem maximizing the passenger satisfaction and/or train operating company's profit. The benchmark will be done on a small part of Swiss network, followed by a test on a full scale network of Israeli Railways. Basic knowledge of programming is required.
Student: Lucia Montero (CSE), June 17, 2016
Supervision: Toms Robenek, Shadi Sharif Azadeh
Traffic simulation model of an innovative catenary-free electric bus
Electric busses help to decrease pollution in city centres. However, they need to be constantly attached to a power source. This limits their mobility and brings visual pollution due to continuous catenaries throughout the city. Focus of this project is a revolutionary catenary-free electrical bus that includes short-range on-board batteries and a series of fast charging stations installed at some bus stops capable to charge the bus batteries in a few seconds without interference with the bus schedules. This project aims to develop a traffic simulation of this electric bus moving on a simple road network. A discrete event simulation representing the fundamental components of the system (e.g. bus, passengers load, on-board battery, charging station, storage level) should be developed, calibrated and validated using traffic data (such as number passengers on board, traffic and congestion information). Then, key indexes representing the system performance should be defined (e.g. travel time, dwell time, battery charge level). Methodologies from transportation modelling, simulation and programming (MATLAB) will be used for completing this project.
Student: Romain Meyer (SGC), June 17, 2016
Supervision: Riccardo Scarinci, Yousef Maknoon, Michel Bierlaire
Solar Decathlon: strategies for a sustainable mobility to achieve the goal of a 2000-Watt society
Solar Decathlon is an international competition that challenges twenty university teams from around the world to build and operate solar-powered houses. This project is part of the EPFL multidisciplinary team of students participating at this competition. The global objective is to transform a regular district in Fribourg into a sustainable eco-district using the solar-powered building as an activator. This building would promote sustainable actions, reduction in energy consumption and soft mobility. This project aims to evaluate the necessary strategies to incentive sustainable mobility inside the district. The goal is to suggest concrete actions that would incentive the modal shift toward more sustainable modes of transport. The effects of these strategies should be quantified and analyzed. The total energy consumed for mobility should respect the limit suggested by the 2000-Watt society. This concept proposes a way of living where nobody consumes more than a continuous 2000 W in comparison with the 6000 W currently consumed in Europe. Knowledge of transportation systems, environmental impacts and strong quantitative analysis skills are required to complete this project. The student will also interact with the other members of the Solar Decathlon competition.
Student: Charles Albert Jeanbart (SGC), June 17, 2016
Supervision: Riccardo Scarinci, Michel Bierlaire
Modeling route choice in Qubec using mental representations
In this project, we aim at simplifying the route choice problem by modeling the strategic decisions of people --represented by the mental representations of their itineraries-- instead of the operational ones --represented by paths. We define an abstracted graph based on what we denote mental representation item (MRI) and we derive operational route choice models based on it. A GPS dataset from the city of Quebec is available and will be used for this purpose. Hence, the project is divided in two main parts; the first and technical part which involves data processing, and the second methodological part which involves the development of a model for the MRI network.
Student: Mathieu Plourde, June 03, 2016
Supervision: Evanthia Kazagli, Matthieu de Lapparent, Michel Bierlaire
Strategic Energy Planning under Uncertainty
The main goal of the project is to classify parameter uncertainty (i.e. defining ranges of variations or probability distribution functions) for the Swiss energy system in the year 2035 and evaluate its impact on energy planning decisions.
Student: Cyprien Say (SGC), June 01, 2016
Supervision: Stefano Moret, Shadi Sharif Azadeh, Michel Bierlaire
Hedonic pricing of car attributes: a comparison across European countries
Hedonic pricing is a revealed preference approach that is used for valuation of constituent characteristics of a good or a service. This project aims at pricing car attributes using disaggregate data on car purchases in 5 European countries from 2010 to 2014. Attention will be paid on estimation of market price gradients and willingness-to-pay for engine types, fuel consumption, weight to power ratio, etc., while accounting for heterogenous preferences of consumers and controlling for market segments, changes in quality, and brand effects. The candidate will take a strong interest in nonlinear regression methods, issues in econometric modeling, analysis and preparation of large datasets, and programming.
Student: Anna-Katharina Clodong (SGC), January 30, 2016
Supervision: Matthieu de Lapparent, Anna Fernandez Antolin, Michel Bierlaire
Specification testing of fundamental diagrams for an anisotropic pedestrian network loading model
Student: Joel Mateus Fonseca (SSC), January 30, 2016
Supervision: Flurin Hnseler, Marija Nikolic, Michel Bierlaire
Online estimation of pedestrian origin-destination demand in train stations using Kalman Filtering
Student: Marc Solsona Bernet (SGC), January 30, 2016
Supervision: Flurin Hnseler, Stefan Binder, Michel Bierlaire
Accelerating moving walkways as a transport mode of the future: system optimization and management
In a hypothetical future where the use of private cars will be limited in cities, the need for movement will be satisfied by a mix of transport modes such as public transport, cycling, walking and other innovative systems. One of these possible futuristic systems, focus of this project, is an urban network of Accelerating Moving Walkways (AMW), i.e. a moving conveyor system for pedestrian similar to the one used in airports capable to reach speeds up to 15km/h. This project aims to optimize the network design of this innovative transport system of AMWs on the city of Geneva. Given the origin-destination demand, pedestrians are assigned to the city road network in order to obtain route choice and trip distributions. This information is used to identify the optimal configuration of link equipped with AMWs and their capacity. For this, a specifically developed optimization framework is used. Empirical mobility data and the road network (both already available) should be used as input of the optimization framework. Methodologies from mathematical optimisation, traffic assignment, programming (MATLAB) and Geographic Information Systems (GIS) will be used for completing this project.
Student: Alexandre Petit, January 15, 2016
Supervision: Riccardo Scarinci, Michel Bierlaire
Development of a novel pedestrian walking model applicable to congested flows
We are in the process of developing a novel pedestrian walking model that can describe multi-directional and congested pedestrian flow. Given a certain demand, the model is supposed to predict travel times and density levels as accurately as possible. The basic idea thereby consists in discretizing walkable space into cells and links, and to compute for each link a speed based on an empirical density-speed relation. The goal of this project, which can be carried out as a Bachelor's, semester, or Master's thesis, is (i) to understand and improve the existing mathematical model, (ii) to accordingly update the computational model (written in Java), and (iii) to consider a real-world case study involving either a Swiss railway station, a Dutch bottleneck experiment or a pedestrian crossing in Hong Kong, China. In the long term, this model will be useful for real-time crowd monitoring and control, as well as for infrastructure dimensioning of e.g. a train station. Skills in object-oriented programming and a basic knowledge of statistical mathematics are required; knowledge of parallel computing is a plus. Semester: Spring 2015. If interested, please contact us at {flurin.haenseler,marija.nikolic}@epfl.ch. We're looking forward to hearing from you.
Student: Gael LEDERREY, June 19, 2015
Supervision: Flurin Hnseler, Marija Nikolic, Michel Bierlaire
In collaboration with SBB-CFF-FFS
Development of an aggregate route choice model for a big network
The use of random utility models for route choice analysis involves challenges stemming from the high requirements in data and data processing, the physical overlap of paths, and the large size of the choice set. These factors increase the complexity of the models significantly. In order to simplify the problem, a novel approach based on an aggregate representation of route choices has been proposed. The conventional representation and modeling approach is based on path alternatives constructed as link-by-link sequences on the network. This approach entails a very large number of possible paths connecting a given origin and destination (OD), and high correlation among the alternative paths. In this work, we claim that a path is solely the manifestation of the route choice, i.e. the way the traveler implements her decision to take a specific route, and we replace the paths with aggregate elements that we denote as Mental Representation Items (MRIs). This key feature allows us to reduce the complexity of the model and at the same time is more behaviorally realistic. The aim of this project is to extend this approach and apply it in a big network for which information about congestion is available. A GPS dataset from the city of Quebec is available and will be used for this purpose. Knowledge of statistical mathematics and familiarity with MATLAB and SQL are required. Geographic Information Systems (GIS) software will be used for completing this project.
Student: Arriagada Diego Alexandre, June 19, 2015
Supervision: Evanthia Kazagli, Matthieu de Lapparent, Michel Bierlaire
In collaboration with McGill University
Pedestrian movement in train stations: modeling speed-density relationship for different classes of passengers
The increased number of passengers in train stations is causing congestion not only on trains, but also on platforms and underpasses. To better design and manage these infrastructures, pedestrian movement should be understood and modeled accurately. Fundamental Diagram (FD) plays an important role in the representation of pedestrian dynamics. FD models the relationship between density of pedestrians and the speed at which they are able to move. This macroscopic model represents the average behavior and does not take into account differences among pedestrians. They can differ in terms of their purpose of the trip (e.g. business/leisure), personal characteristics (e.g. age, gender), presence of luggage, etc. This heterogeneity can lead to different walking behavior of pedestrians and should be adequately modeled by the FD. This project aims to develop a Multi-Class Fundamental Diagram (MC-FD) assuming the existence of different classes of passengers in a train station. As a case study we will use the Lausanne train station where detailed pedestrian trajectories have been collected. The key elements of the project are: (i) literature review of MC-FD for both pedestrians and vehicles; (ii) development of the modeling assumptions employing empirical, qualitative and sociological considerations; (iii) model specification; (iv) sensitivity analysis, calibration and validation of the proposed model based on empirical data. Knowledge of statistical mathematics and familiarity with programming (MATLAB) are required.
Student: Laure Emma Rosine, June 19, 2015
Supervision: Marija Nikolic, Riccardo Scarinci, Michel Bierlaire
Routing of a mixed fleet of electric and diesel trucks: Analysis of solution approaches
A transportation company wants to study the potential benefit of replacing some of the diesel trucks in its fleet with electric ones. Electric and diesel trucks have comparable characteristics in terms of capacity and electric trucks are capable of performing an average daily itinerary on a single charge. The price of an electric truck is higher, while its operating cost is much lower compared to a diesel truck. Therefore, we want to find an intelligent deployment of the mixed fleet so as to minimize cost. The goal is the development of a vehicle routing model and/or heuristic, taking into account the characteristics of the various trucks, including fuel/energy consumption, speed, capacity, compatibilities, working time, service times, etc. The integration of the truck load's impact on fuel/energy consumption and the analysis of battery replacement/recharging strategies could also be considered if time permits. Solution quality can be compared against benchmarks results and data from the state of practice. As a previous semester project, a student has already worked on defining the problem, developing a heuristic algorithm and obtaining some results. The current project is intended to do a deeper analysis and implementation of solution methodologies. This project is appropriate for a student with an interest in operations research and experience in programming (Java preferred). The workload can be adjusted to the number of credits.
Student: Noortje Verstegen (SIN), May 29, 2015
Supervision: Iliya Markov, Stefan Binder, Michel Bierlaire
In collaboration with HES-Fribourg, CREM
Implementation of a futuristic transport system based on accelerated moving walkways: optimization on a real case study
In a hypothetical future where the use of private cars will be limited in cities, the need for movement will be satisfied by a mix of transport modes such as public transport, cycling, walking and other innovative systems. One of these possible futuristic systems, focus of this project, is an urban network of Accelerated Moving Walkways (AMWs), i.e. a moving conveyor system for pedestrian similar to the one used in airports. This project aims to apply an optimization framework to identify the optimal network design of this innovative transport system of AMWs on a real case study. Given the origin-destination demand, pedestrians are assigned to a road network in order to have route choice and trip distributions. On this scenario, the optimal configuration of link equipped with AMWs and their capacity is define. For this purpose, a specifically developed optimization framework is used, and it should be applied to a real city. Lausanne presents appropriate transportation and geographical characteristic to be used as a case study. Empirical mobility data and the city network should be used as input of the optimization framework. Methodologies from mathematical optimization, traffic assignment, programming (MATLAB) and Geographic Information Systems (GIS) will be used for completing this project.
Student: Raphal Luthi (SSIE), May 29, 2015
Supervision: Riccardo Scarinci, Iliya Markov, Michel Bierlaire
Planning tool for the admission of medical graduates into the Otolaryngology training program in the CHUV
Student: Guillaume Lopez, January 15, 2015
Supervision: Yousef Maknoon, Shadi Sharif Azadeh, Michel Bierlaire
A destination choice model for EPFL Campus
Based on WiFi traces from access points, the goal of this semester project consists in developing binary discrete choice models for the choice of attending classes.
Student: Loc Tinguely (SGC), December 19, 2014
Supervision: Antonin Danalet, Matthieu de Lapparent
Accounting for attitudes in modeling demand for electric vehicles
The purpose is to characterize and to model determinants of electric vehicle acceptance and adoption by individuals. Special attention will be paid to additional integration of attitudinal drivers and barriers to adherence to such technology. It aims at understanding how consumers accept the financial and lifestyle investments associated with the leap from traditional to electric powertrains, focusing on battery electric (BEV) and plug-in hybrid electric (PHEV) vehicles. To this extent, the 2012 Renault-Nissan Alliance survey on electric vehicles will be used. Data are collected over 5 European countries: France, Italy, Germany, Spain, UK. It focuses on current owners of a car bought new between 6 months and 5 years ago and who intend to buy a new car in the next 5 years. State-of-the-art and extended discrete choice models will be developed to account for formation of latent processes and existence of endogeneity issues.
Student: Maurin Baillif (SGC), December 19, 2014
Supervision: Matthieu de Lapparent, Anna Fernandez Antolin, Michel Bierlaire
In collaboration with Nissan
Routing of a mixed fleet of electric and internal combustion trucks
A transportation company has introduced a number of electric trucks into their current fleet of internal combustion trucks. Electric and internal combustion trucks have similar characteristics in terms of capacity and electric trucks are capable of performing an average daily itinerary on a single charge. The goal is the development of a vehicle routing model and heuristic, taking into account the characteristics of the various trucks, including fuel/energy consumption, speed, capacity, compatibilities, working time, service times, etc. Various objectives could be considered and analyzed, such as the minimization of time, energy consumption and cost. The integration of the truck load's impact on fuel/energy consumption and the analysis of battery replacement/recharging strategies could also be considered if time permits. Data is available for benchmarking with the state of practice and assessing the heuristic's performance against the mathematical model. This project is appropriate for a student with an interest in operations research (mixed integer linear programming), some experience with mathematical programming languages (e.g. AMPL or CPLEX OPL) and good programming skills (e.g. C++, Java or Matlab).
Student: Thomas Cibils (SMGT), December 19, 2014
Supervision: Iliya Markov, Stefan Binder, Michel Bierlaire
In collaboration with HES-Fribourg, CREM
3D routing approach for air navigation planning for small sized planes
This projects aims to study and visualize the aviation data using factors such as, geographical information of the plane as well as weather condition for a flight most specifically during cruise phase. MATLAB is a powerful software with high quality of visualization of data, however, when the size of data gets bigger, the performance of MATLAB degrades and it becomes slower. Recently, some new software such as, Tableau has been introduced that provide powerful tools to visualize the big data.
Student: Jamal El Rhazi (SGC), December 19, 2014
Supervision: Shadi Sharif Azadeh, Michel Bierlaire, Stefan Binder
Development and Implementation of a Decision Making Tool in Quality Control Networks using Quantitative System Modeling Techniques
In an increasingly complex business environment, managers have to grapple with problems and issues, which range from relatively trivial to the strategic. This project proposes to develop and implement a decision making tool to optimize quality control networks. To do so, several metaheuristics will be used as an efficient approach to improve global value chains towards lean manufacturing applied to a production process of Nestle. The control mechanism of each control point determines the conformity (or non-conformity) of the product in the supply chain defining a sample randomly chosen units and indicating the maximum number of bad units for each batch to pass control points. As control points become decision variables, the detailed examination of control mechanisms at each control point requires the use of nonlinear algorithms. The following decision variables have to be optimized in the considered networks: (1) determine the set of control points; (2) determine the control plan parameters of each control point. That point requires understanding basics of statistical quality control. The objective function contains a performance and a risk component, and a budget constraint has to be satisfied. The following metaheuristics will at least be investigated: tabu search, variable neighborhood search and adaptive memory algorithms. In addition, an exact method relying on CPLEX should be compared to the proposed metaheuristics.
Student: Pierre Jullien de Pommerol (SGC), July 18, 2014
Supervision: Michel Bierlaire, Prof. Nicolas Zufferey, EPFL STI LGPP
A two-step approach for estimating pedestrian demand in a congested network
Student: Eduard Rojas (SIN), May 30, 2014
Supervision: Flurin Hnseler, Antonin Danalet, Michel Bierlaire
In collaboration with SBB-CFF-FFS
Train Management in SNCF Application
Student: Cao Huu-n (CSE), May 30, 2014
Supervision: Toms Robenek, Stefan Binder, Michel Bierlaire
Optimisation of the network design of a futuristic transport system based on moving walkways
In a hypothetical future where the use of private cars will be limited in cities, the need of movement will be satisfied by a mix of transport modes like public transport, cycling, walking and other innovative systems. One of these possible futuristic systems, focus of this project, is an urban network of moving walkways, i.e. a moving conveyor system for pedestrian like the one used in airports. This project aims to study and optimize the network design of this innovative transport system of moving walkways, exploring various network configuration and system characteristics to understand which specification could satisfy at best the increased demand. Given the origin destination demand and trip distribution, the variables of the optimization problem should be defined such that indexes of performance, e.g. overall traveling time and cost, are minimized. Example of design variables for this innovative transport system are: number of links to be equipped with moving walkways, moving pathway speed and system capacity that can be related to energy consumption and construction cost. Methodologies from mathematical optimization, transport network design, simulation and scripting will be used for complete this project.
Student: Guillaume Lopez, May 30, 2014
Supervision: Riccardo Scarinci, Jianghang Chen, Michel Bierlaire
Demand/supply coupling in pedestrian traffic estimation
The main purpose of this paper is to show a mathematical framework for solving the fixed-point arising in joint demand estimation / traffic assignment problems. As a case study. a congested corridor with known outflow is considered, for which the inflows is predicted using the developed framework. The model is based on a coupling between the travel demand and the network supply (the infrastructure available) in an analogous way as what is done in economics.
Student: Jrmy Rabasco (CSE), January 31, 2014
Supervision: Flurin Hnseler, Michel Bierlaire
Schedule-based estimation of pedestrian travel demand within a quasi-uncongested railway station
In many railway stations, capacity limits are reached at peak hours and congestion in pedestrian facilities occurs. There is thus a general need to analyze and model pedestrian flows in train stations. In this process, estimating pedestrian origin-destination demand and, more precisely, route flows is a major challenge. The ultimate goal of this research project is to dynamically predict pedestrian travel demand within a train station based on train time table and train track assignment. To this end, a preliminary methodology has been developed. This framework needs to be refined and extended in many ways, which can be an interesting and suitable task for a semester or Master's thesis. Your thesis would be part of a comprehensive project employing pedestrian tracking, flow modeling and infrastructure optimization. All ideas and model concepts will be tested, parametrized and validated on a real case study.
Student: Quentin Mazars-Simon (SSC), January 31, 2014
Supervision: Flurin Hnseler, Amanda Stathopoulos
In collaboration with SBB-CFF-FFS
Implementation of Tabu Search in Quality Control Networks
This project proposes to optimize quality control networks using the tabu search algorithm as a novel solution to improve global supply chains towards lean manufacturing applied to the Nestle chocolate supply chain. The control mechanism of each control point determines the conformity (or non-conformity)) of the product in the supply chain defining a sample of randomly chosen units and indicating the maximum number of bad units for each batch to pass a control point. As control points become decision variables, the detailed examination of control mechanisms at each control point require the use of non-linear algorithms. In a first phase, control points as decision variables necessitate the examination and identification of (sub)optimal control mechanisms. This is followed by the proposition of a first model using the tabu search algorithm. This model will demonstrate links among suboptimal solutions, but also synergies among risk, performance and budget distribution in quality control networks. In order to efficiently allocate quality control network resources, this project proposes to globally optimize the control mechanism for equi-distribution of risk, performance and budget using the tabu search algorithm.
Student: Pierre Jullien de Pomerol (SGC), January 20, 2014
Supervision: Michel Bierlaire, Prof. Nicolas Zufferey, EPFL STI LGPP
Optimisation des tournes de ramassage des employs de l'aroport de Genve (Pre-Project)
Student: Isabel Tovar (SGC), January 20, 2014
Supervision: Toms Robenek, Stefan Binder, Michel Bierlaire, Philippe Quaglia
In collaboration with Geneva Airport
Vehicle dispatching problem in physical internet hub and spoke model
The call for sustainability in logistics section promotes the idea of Physical Internet (PI). The approach of PI aims to universally interconnect logistics networks as the digital internet did with computer networks. In the world of physical internet, consumer goods are encapsulated in smart and secured containers (i.e., PI containers) and are routed in logistics networks with the similar pattern that digital packets are handled in the internet.
Student: Alexis Dubil (SGC), January 17, 2014
Supervision: Jianghang Chen
Vehicle routing problem coupling with bin packing
The call for sustainability in logistics section promotes the idea of Physical Internet (PI). The approach of PI aims to universally interconnect logistics networks as the digital internet did with computer networks. In the world of physical internet, consumer goods are encapsulated in smart and secured containers (i.e., PI boxes) and are routed in logistics networks with the similar pattern that digital packets are handled in the internet. In this student project, we focus on the last mile problem in the PI. After PI boxes arrive their final designated PI hub, given clients addresses and the information of the trucks (e.g., truck weight and volume capacities) available at this hub, the decision makers aim to assign the minimal number of trucks to deliver all the arrived PI containers to their destinations in a cost-effective way. This is a typical Vehicle Routing Problem (VRP) but with the side constraint that we need to take the dimensions of the PI boxes (in 3D) into account when loading and/or unloading occur. This problem is termed vehicle routing problem coupling with bin packing. Its essential constraints that the student need to consider are: 1. The available number of vehicles is bounded; 2. Vehicle weight and volume capacities cannot be exceeded; 3. The loading and unloading need to follow the policy of Last-in-first-out.
Student: Buytaert Gabrielle (SSC), January 17, 2014
Supervision: Jianghang Chen
Identify Users Locations of Interest from Smartphone WiFi Data
Smartphone is a powerful and convenient tool for collecting a variety of data (location, social interaction, and more) that can be useful for individual mobility analysis. Most of the ongoing research relies on GPS to acquire accurate location. GPS though is expensive in terms of battery consumption. In addition, the GPS sensor embedded in mobile phones fail in practice more often than dedicated GPS devices. Yet, smartphones are endowed with multiple sensors from which location can be inferred. Wifi sensors provide stable indoor location data and have been identified as a potential alter- native to indicate location that can well serve in addition and complementary to GPS data to advance mobility learning. The location extraction from Wi-fi records permits the detection of users regular places of visit, the frequency and intensity (number of times) of visiting these places, as well as patterns such as the time of day and duration of visit. This kind of information is relevant for the identification of activity locations making use of clustering techniques, and after further analysis they can support measuring the activity space and intensity of activity participation of the user. By fusing the information about activity locations with land use and points of interest (POI) data, as well as with more data acquired from the smartphone sensors (e.g. phone status, Bluetooth, charging, phone interaction etc.), and taking into account temporal dimensions, it is then possible to infer the type of activity at these location (including home and work locations) and subsequently trip/ activity purposes can be disclosed. In addition to the above, and building on the identification of meaningful clusters, WiFi data have the potential to assist the identification of origins and destinations of trips (often missing in case of GPS). This information can then be used for improving and validating trip detection algorithms, which currently appear to be problematic and hence introduce biases in the subsequent stage of map-matching (e.g. one a known trip is segmented in several due to the limitations of the algorithm and GPS data). This project aims at exploiting the advantages of the WiFi data discussed above for individual mobility analysis.
Student: Amlie Buisson (GC) (SGC), January 17, 2014
Supervision: Evanthia Kazagli, Antonin Danalet, Michel Bierlaire, Francisco Pereira
Mode choice analysis from a large smartphone dataset
Smartphone is a powerful and convenient tool for collecting a variety of data (location, social interaction, and more) that can be useful for individual mobility analysis. One of the advantages of smartphone data is the ability to collect data over longer periods of time (panel data) without burdening the respondent. Having such resolution of information enables us to gain better insight into the general mobility style of people than it would be possible by means of traditional one- or two-day travel surveys. It becomes clear that smartphone data have great potential with respect to analysing travel- ers profiles and disclose systematic mobility (e.g habit, morning or afternoon routines etc.) and switching behavior patterns (departure time, chosen route, mode). This project focuses on such higher-order mobility styles in order to shed light on travel behavior.
Student: Mikael Nicolas Xavier Friederich (SGC), January 17, 2014
Supervision: Evanthia Kazagli, Marija Nikolic
Exploration of smartphone users trip data to investigate travel behavior
Smartphone is a powerful and convenient tool for collecting a variety of data (location, social interaction, and more) that can be useful for individual mobility analysis. One of the advantages of smartphone data is the ability to collect data over longer periods of time (panel data) without burdening the respondent. Having such resolution of information enables us to gain better insight into the general mobility style of people than it would be possible by means of traditional one- or two-day travel surveys. This project aims at unveiling frequent behaviours in terms of space, i.e., the regions of space traversed during movements and understand the processes generating them. The work elaborated in the project is expected to support our effort in incorporating travelers mental/ spatial represen- tation items (MRI) in route choice modeling. More specifically, the goal of this bigger project is to develop a modeling framework where the route choice decisions will take place in a higher/ concep- tual level. Path alternatives will be constructed as replaced by sequences of mental geo-marked items akin to the concept of anchor points as elements of travelers mental maps in cognitive science. We are interested in car route choice, hence the scope of this semester project will also focus on car trips. In this context, processing and analysing the trips of the smartphone users can reveal the most frequently traversed points or segments of the network (both in an aggregate but also in an individual level) that can signify such items in the network.
Student: Mikael Nicolas Xavier Friederich (SGC), January 17, 2014
Supervision: Evanthia Kazagli, Marija Nikolic
Exploring pedestrian mobility using video tracking data in Lausanne train station
Pedestrians, contrary to the other modes, do not have defined network and they do not follow strict constraints. A characteristic feature of pedestrian route choice is that routes are continuous trajectories in time and space - pedestrians choose a route from an infinite set of alternatives. As a consequence of these facts we have observed the distribution of walked distances and travel times for specific origin-destination pairs at the train station in Laussane, during the morning rush hour. The goal of this project is to analyse the distribution of walked distances and corresponding travel times and speed values based on pedestrian trajectories collected at the train station in Lausanne. Here we would like to answer the questions such as: 1) Is the motion of pedestrians random or does it follow a specific pattern? How can we characterize it? 2) What is the impact of congestion on the mentioned observables? 3) Do familiarity with the place, the attractiveness of shops and signs found along the corridors affect pedestrian motion behaviour and how? What are the other factors that cause deviations from a straight line motion? Programming skills and knowledge of statistical mathematics are required.
Student: Babel Hugo Louis, January 17, 2014
Supervision: Marija Nikolic, Evanthia Kazagli
Modeling the demand for financial products
This MA pre-project focuses on developing a demand model for financial products. The work is carried out in collaboration with a trading service partner an based on real data on commodity portfolio investments. The project will cover the following: i) a literature review on the demand for financial products, behavioural finance on trading choices, ii) development of a demand model on the client data that is able to qualify the impact of different determinant (price, spread, client features, external events) on decisions to buy/sell the commodity, iii) testing the model on different scenarios (sensitivity to different changes in conditions) and validation. The work comprises: conceptual modeling, data elaboration/exploration and empirical model development, hence this pre-project and the related MA project requires skills in mathematics, management of large data, statistics/econometrics related to discrete choice models and some knowledge of behavioural theories in finance.
Student: Billal Mahoubi (SGC), January 10, 2014
Supervision: Amanda Stathopoulos
Optimisation des transports scolaires Grandson II
Analyser les problmatiques, actuelles et futures, et proposer des solutions d'amlioration dcoulant de l'organisation et de la gestion de l'ensemble des transports ncessaires la bonne marche du groupement et de l'arrondissement scolaire de Grandson (18 communes, 1'350 lves, 70 classes, 12 sites scolaires). Ces problmatiques rsultent principalement d'une augmentation significative des cots lis l'introduction de la communaut tarifaire vaudoise (Mobilis), de contraintes lies l'entre en vigueur d'HarmoS, des projets de constructions de l'association intercommunal, de l'volution du rseau rgional des transports voire galement des synergies dvelopper avec le milieu parascolaire.
Student: Franka Tholen (SMA), June 07, 2013
Supervision: Michel Bierlaire, Toms Robenek, Bilge Atasoy, Yves Guilloud, Association Intercommunale du Groupement et de l'Arrondissement Scolaires de Grandson
Mobility learning from smartphone WIFI data
Smartphone is a powerful and convenient platform for collecting location data for mobility behavior research. Most of the research relies on GPS to get accurate location, or GSM to get massively available but imprecise location. This project investigates the feasibility of using WIFI data for mobility behavior research. The dataset is collected from 200 smartphones over 2 years. The data collection application (EPFLScope) records all nearby WIFI access points (AP) every 3-5 minutes. Some of the WIFI access points are associated with location information. Therefore, it is possible to identify the location of the smartphone according to the nearby WIFI AP. With this location information, we can further understand spatial and temporal mobility patterns.
Student: Amlie Buisson (SGC), June 07, 2013
Supervision: Jingmin Chen, Evanthia Kazagli, Marija Nikolic
Activities in Palo Music Festival from Bluetooth
Can we guess the activities of spectators in a music festival from Bluetooth traces collected from 10 people with smartphones used as antennas? This project will study an existing dataset and existing map data from Palo music festival in Nyon, Switzerland. Depending on the results, the report will either draw conclusions about spectators' behavior, or recommend other data collection specifications for this particular case study. For more information, don't hesitate to contact the assistant of the project.
Student: Elisaveta Kondratieva (SSC), June 07, 2013
Supervision: Antonin Danalet
Visualization of pedestrian demand in a 3D graph
This project aims at visualizing flows in a pedestrian graph in 3D. The graph is coded as a postGIS database (postgresql). The goal consists in representing the density at each node and eventually the flows between the nodes (data coming from our own algorithm). The visualization needs to be both in 3D and in 2D for printing. An IC student would ideally fit the goal of this project. The first step would be to compare the pros and cons of Processing and Google Map/Earth. Then it would consists in visualizing the pedestrian graph in the chosen tool. As a final step, the student will include density data of signals (without treatment) and the results of our algorithm in the map (treated data) and generate final maps and animations from data. As an inspiration of how it could possibly look like this (but with pedestrian as dots instead of trains)
Student: Javier Lopez-Montenegro Ramil (SIN), June 07, 2013
Supervision: Antonin Danalet, Bilal Farooq
Associations generation in synthetic population
In recent years significant advancements have been made in the research related to agent based urban systems modelling, especially in the area of activity based large scale travel demand models. Synthesis of different types of agents (person, family, and household) is an integral part of the input preparation for such models and micosimulations. In addition to agents generation, these recent developments also require synthesizing associations between di fferent types of agents. For instance, which individual person is married to whom, who is the father of a 7 years old male going to a particular school, etc. Such information is very useful in realistic modelling and microsimulation of short term decisions like who gets to take the car for work and long term decisions for instance, what type of dwelling a particular household will buy. A Markov chain Monte Carlo simulation based agents synthesis approach has been proposed by Farooq et al. (2012). At core, the approach used Gibbs sampler to draw from the joint distribution of agent attributes using the available data. Among other advantages, they also demonstrated that the approach can reproduce the joint distribution better than the conventional fitting based approaches. Here we extend the simulation based synthesis approach so that it can also generate the association among drawn agents of di fferent types. Required skill set: Experience in data analysis and pre-processing; probability and statistic; coding in Java or C#
Student: Paul Anderson, June 07, 2013
Supervision: Bilal Farooq, Dimitrios Efthymiou
Enhancement of Naville's Press Delivery Regulation Tool: An Exploratory Analysis
The aim of this project is to explore data driven complementary strategies to the existing allocation strategy of Naville, in order to provide guidance for potential improvements of the existing allocation tool of the company. In this project the student will investigate the potential benefits of including a wider set of data in order to identify press demand patterns over time. Accounting for consumer profiles, detailed sale points data, geographical area, external events, etc., will enable a more accurate understanding and anticipation of demand fluctuations. Drawing from the results of this analysis we will provide recommendations for improvements of the existing tool used by Naville. The proposed exploratory project can form the basis for a further work on the challenging theme of modeling and predicting demand and optimizing distribution strategies in the context of declining and uncertain evolution of the press market.
Student: Lea Kissling (SMA), June 07, 2013
Supervision: Evanthia Kazagli, Amanda Stathopoulos
In collaboration with Naville
Optimization of Paris' Metro System
The Metrolab (a joint venture between RATP (Rgie Autonome des Transports Parisiens) and Alstom) would like to conduct analysis of their current metro system in terms of metro schedules and pedestrian congestion in the stations. The aim of this project is to use optimization techniques to design more flexible dynamic timetables and partially to handle the pedestrian congestion in the stations. This project will be offered only for spring semester 2013.
Student: Thomas Cibils (SMA), June 07, 2013
Supervision: Toms Robenek, Jianghang Chen, Michel Bierlaire
In collaboration with MetroLab
Modelling car ownership duration
Household vehicle ownership influences many aspects of travel demand, with pronounced impact for energy consumption, travel mode distribution, residential location patterns and city attractiveness. Despite a large number of vehicle choice models focusing on the type of car chosen, formal modelling of the timing and duration of ownership has received relatively little attention. Statistical models known as duration (hazard) models can be used to estimate the distribution of vehicle ownership lengths in a population. A central interest is to identify the determinants of the time that elapses between two automobile transactions, including characteristics of the car and household and macroeconomic/policy variables. In addition, the influence of less tangible factors such as expectations regarding the market, future vehicle launches, propensity for planning and personal aspirations may have a large impact on the timing of renewal and vehicle ownership duration. We have a large dataset available containing rich information on vehicle acquisitions and are looking for a student to collaborate on data elaboration, exploration and development of a duration model for vehicle ownership. Several assumptions for the hazard function and innovative determinants of ownership spells will be considered. The student should be familiar with management of large data, statistics/econometrics and have some notion of behavioral theories of decision-making.
Student: Natalie Sauerwald (SMA), June 07, 2013
Supervision: Amanda Stathopoulos, Aurlie Glerum
In collaboration with PSA
Movement patterns of pedestrians on platforms
To understand how people access a train station by means of trains, the arrival and departure pattern of pedestrians on platforms is of interest. In particular, the embarkation/disembarkation behavior of train passengers is an important factor for congestion in pedestrian facilities. The departure pattern of pedestrians awaiting to board a train has been shown to follow a beta distribution. In contrast to that, arriving passengers usually enter a train station as a very dense crowd, causing a theoretical inflow that exceeds the capacity of pedestrian facilities. The goal of this work is to mathematically describe the arrival and departure patterns caused by single trains as well as for platforms on which multiple trains follow each other in close succession. Very detailed pedestrian tracking data of platform 3/4 in Gare de Lausanne will form the basis of this analysis. Basic programming skills or a strong willingness to acquire such knowledge is desired.
Student: Nicholas Alan Molyneaux, May 31, 2013
Supervision: Flurin Hnseler, Amanda Stathopoulos
In collaboration with SBB-CFF-FFS
Development of a novel pedestrian flow simulator
Due to a general increase in travel demand, pedestrian flows in railway stations are gaining importance. Space is getting scarce, pedestrian density is reaching critical levels, and walking times are getting longer. To better understand these phenomena, we are currently developing a cell-based pedestrian flow simulator. Key features of this framework include its ability to i) realistically reproduce pedestrian density waves caused by arriving trains, ii) adequately describe multi-directional flows often present in public spaces, iii) consider differences in walking speeds among characteristic groups of pedestrians (such as passengers with luggage, handicapped people, travelers in a hurry, etc). To extend and implement this framework, basic programming skills or a strong willingness to acquire such knowledge is needed. Familiarity with fundamental fluid dynamics would be a plus. Extensive support and insight into ongoing research are provided.
Student: Thomas Mhlematter, May 31, 2013
Supervision: Flurin Hnseler, Bilal Farooq
Movement patterns of pedestrians on platforms prior to/after train departures/arrivals in Gare de Lausanne: Exploitation of pedestrian tracking data
In order to better understand how people access and leave a train station by means of trains, the arrival and departure pattern of pedestrians on platform is of interest. In particular, the embarkation/disembarkation behavior of train passengers is an important factor for congestion in pedestrian facilities. The departure pattern of pedestrians awaiting to board a train is expected to follow a beta distribution. In contrast to that, arriving passengers usually enter a train station as a very dense crowd, causing a theoretical inflow that exceeds the capacity of pedestrian facilities. This saturation effect needs to be taken into account when modeling arrival patterns. The goal of this work is to mathematically describe the arrival and departure patterns caused by single trains as well as for platforms on which multiple trains follow each other in close succession. Case studies for RER, RE and ICE/EC trains will be considered in order to better understand the influence of train size and type on these patterns.
Student: Isabel Tovar, January 12, 2013
Supervision: Flurin Hnseler, Marija Nikolic
Dynamic estimation of pedestrian origin-destination within train stations: Exploitation of pedestrian tracking data and comparison to travel surveys
OD demand will be estimated dynamically (i.e., as a function of time) based on observed pedestrian trajectories. Key in this process is a diligent choice of origins/destinations and their cordons, as well as a smart handling of lost pedestrians, i.e., people of which the algorithm looses track due to overcrowding or bad light conditions. Another challenge is finding a way to deal with multi-destination trips. For instance, a person entering a train station which goes to a ticket vending machine before boarding a train falls into this category. To help us tackle these issues, VisioSafe grants us access to some of their codes such that we can see how they have solved similar problems. Besides developing an OD estimation algorithm based on tracking data, visualization of OD demand is an important part of this thesis. Furthermore, it might be interesting to investigate how the estimated OD demand compares to previous estimates based on travel surveys (Anken et al., 2012) and pedestrian count data (ASE).
Student: Malle Zimmermann (SMA), January 12, 2013
Supervision: Flurin Hnseler, Antonin Danalet
Job Shop Scheduling in a Medical Parts Production Factory
Student: Nathan Scheinmann (SMA), January 11, 2013
Supervision: Michel Bierlaire, Nitish Umang
Robustness and recovery in airline scheduling
Student: Jonathan Blaiberg (SGC), January 11, 2013
Supervision: Michel Bierlaire, Bilge Atasoy
Process Optimization and School Schedules: New Legistlation, New Constraints
Student: Ailin Zhang (SMA), January 11, 2013
Supervision: Michel Bierlaire, Nitish Umang
Optimisation des transports scolaires Grandson I
Analyser les problmatiques, actuelles et futures, et proposer des solutions d'amlioration dcoulant de l'organisation et de la gestion de l'ensemble des transports ncessaires la bonne marche du groupement et de l'arrondissement scolaire de Grandson (18 communes, 1'350 lves, 70 classes, 12 sites scolaires). Ces problmatiques rsultent principalement d'une augmentation significative des cots lis l'introduction de la communaut tarifaire vaudoise (Mobilis), de contraintes lies l'entre en vigueur d'HarmoS, des projets de constructions de l'association intercommunal, de l'volution du rseau rgional des transports voire galement des synergies dvelopper avec le milieu parascolaire.
Student: Clement Massart (SMA), January 11, 2013
Supervision: Toms Robenek, Bilge Atasoy, Michel Bierlaire, Yves Guilloud, Association Intercommunale du Groupement et de l'Arrondissement Scolaires de Grandson
Pedestrian flow simulation in Lausanne train station
This project is about modelling, simulating, and analyzing the future expansion scenarios, in the context of pedestrian flows within Lausanne Gare (train station). We have already developed the modelling and simulation test-bed in a pedestrian flow simulator, called VisWalk. Here we would like to use more detailed data for existing and future demand in our models and analysis. A considerable amount of time will be spent to perform the calibration of various parameters using aggregate level speed and density related data. If time permits, we would also be interested in doing small infrastructure changes, specifically tapering of the intersections between PIs and ramps in order to evaluate their effects of the pedestrian flows. Ideal student should have some experience in using CAD software (AutoCAD, Microstation); should have some knowledge of databases; and be able to do basic programming. The student working on the project will have a great and unique opportunity to work directly with SBB CFF FFS, a major transportation software company called PTV AG, and a rapidly growing startup from EPFL called VisioSafe.
Student: Nicolas de Lamberterie, January 10, 2013
Supervision: Bilal Farooq, Marija Nikolic
In collaboration with SBB CFF FFS, PTV AG, VisoSafe
Population synthesis for large-scale agent based microsimulation
Dynamic microsimulation of cities, including transportation, land use, and energy, require an initial dissaggregate population of agents (households and persons) as a key input. Due to privacy reasons, most of the times, governments do not provide access to full information on the population in census. This requires synthesizing the population from available datasets. In principle two datasets are necessary for such synthesis: a) disaggregate sample of households/persons; b) demographic summaries for all the zones in the study area. Techniques like Iterative Proportional Fitting (IPF) method are developed to use these datasets and generate a synthetic population of households and persons. A new technique for population synthesis has been developed at Transportation and Mobility lab. The scope of this project is to a) prepare datasets from the census and various other micro data samples b) run the simulations code based of the developed technique, for various scenarios (code is written in c++ and will be provided to the student. Only a very basic programming knowledge is required from the student) c) analysis and comparison of the results with other methods. This project will require developing understanding of the general methodology of population synthesis, preprocessing the datasets, familiarity with MySQL, GIS software, and some programming experience is recommended.
Student: Lovisa Arnesson, January 10, 2013
Supervision: Bilal Farooq
In collaboration with European Union funded, SustainCity project
Modelling the choice of vehicle in an extended framework
It is becoming increasingly clear that models of vehicle choice need to find a way to measure complex consumer substitution patterns across products and over time, while accounting for a wealth of features that modulate the buying experience. This means that planning, acquisition and use of the vehicle is linked through features such as attitudes, life-event changes, status aspirations and perceptions of making a good deal. We have a large dataset available containing rich information on vehicle acquisitions and are looking for a student to collaborate on data elaboration, exploration and modelling that factors in richer behavioral variables than is common in traditional vehicle choice models. Methodologically the choice of vehicle will be modeled using random utility discrete choice demand systems. The student should be familiar with management of large data, statistics/econometrics and have some notion of behavioral theories of decision-making. The work will start with exploring the structure in the data, formulating simple models to identify some constructs and segments in the data then move towards a more formal choice model with perceptual/attitudinal components.
Student: Areg Gevorgyan (SMGT), January 07, 2013
Supervision: Amanda Stathopoulos
Generation and Simulation of MATSim plans for Brussels
The objective of this project is to generate travel plans for a synthetic population for the city of Brussels and simulate their travel behavior with the transport microsimulation software MATSim. The plans will be generated based on available census and travel survey data. The network for Brussels will be provided. The project considers the implementation of a prototype (simplified) model and a sensitivity analysis of the simulation results
Student: Sona Hunanyan (SMA), July 30, 2012
Supervision: Ricardo Hurtubia
Optimizing Security Staff Operations at Geneva Airport
Geneva airport is the second largest airport in Switzerland. Airport is capable of easily handling 80-100 landings and take-offs per hour. However one of the major bottlenecks today is the passenger handling capacity. The airport terminal area can handle up to 3000 passengers per hour. In recent times, passenger congestion is frequently building up at the airport. Because of the geographical reasons, it is not easy to expand the terminal building. As a result, it may be fruitful to attempt reducing congestion by efficient handling of the passengers and not letting the passengers wait too long to be serviced. Security screening is the most important area for departing passengers. Optimizing the operating costs for security personnel at airports is a complex problem due to a number of reasons. Each passenger and their corresponding hand baggage need to undergo security screening before entering the boarding gates. Even though the passengers and the baggage need to move together, the rate of service for the two can be vastly different. Security personnel work over specific shift times and need to be provided with suitable meal breaks. Staff members can work full-time or part-time, with a minimum of four hours per day. Passenger arrival pattern at an airport can be extremely uneven. Flight activities also tend to be non-uniform while passenger service expectations and staffing inflexibilities due to shift durations can make the problem even more complex. In this work, we would use the flight schedule and service criteria to develop a method to find optimal shift timings and the mix of full-time and part-time workers such that the overall costs are minimized and the service criteria conditions are fulfilled. Term Project (Master Project Pre-Study) Deliverables Determine global parameters for passenger arrival rates, service times and shift times at different times of the day and days of the week for both passengers and baggages Develop and implement a cost optimization MIP model with discrete time intervals Masters Project Deliverables Adapt the deterministic linear model to account for non-linear passenger arrival and service rates and implement the same Build a robust optimization model to capture vastly varying arrival and service rates. Determine the various ways to capture model stability with special emphasis on recoverability Perform sensitivity analysis and compare the results of the stochastic model with simulation Use the results to make recommendations: For example, create a separate queue or sub-queue for passengers without baggage, or passengers whose flights depart in 20 mins should be brought ahead of the queue, etc. Develop and implement the model for staff rostering at the security screening Apart from expertise in modeling, the student is expected to be able to implement his model and write his own codes in a standard programming language, such as C or C++ Desired Masters Project Deliverables Evaluate the tendency for staff to take leaves (including sick leaves) and vacation (subject to data availability) Plan for contingencies in your model Project Organization The student would be expected to update us on the progress and revert to us for queries on a pre-decided day of the week. The student would be expected to update us on the progress and revert to us for queries on a pre-decided day of the week. A midterm presentation would be scheduled on Thursday, 10.11.2011. At the completion of the term project, a report will be delivered (both in paper as well as electronic form) before the 13.01.2012. The final oral presentation for the term project will be scheduled on 11.01.2012. Similarly, the Masters project would be due before the 01.06.2012 while the oral presentation for Masters project will be scheduled between 18.06.2012 and 29.06.2012 or a mutually convenient and agreed date. All the files, programs, codes, data and report associated with the project must be delivered before these date. All reports will be submitted in two copies to TRANSP-OR Lab and GVA Airport. References Dowling, D., Krishnamoorthy, M., Mackenzie, H., and Sier, D. (1997). Staff rostering at a large international airport. Annals of Operations Research, 72:125-147. Ernst, A., Jiang, H., Krishnamoorthy, M., and Sier, D. (2004). Staff scheduling and rostering: A review of applications, methods and models. European journal of operational research, 153(1):3-27. Gilliam, R. (1979). An application of queueing theory to airport passenger security screening. Interfaces, 9(4):117-123.
Student: MAHMOUD KHAROUF, June 20, 2012
Supervision: Michel Bierlaire, Prem Kumar, Nitish Umang, VINCENT-RUBEN JIMENEZ, GVA AIRPORT
Sensitivity analysis for a new generation of aircraft: Clip-Air
Clip-Air is an innovative air transportation system based on modularity. By design, loading units (capsules) can be detached from the carrying unit (wing) which has several advantages in terms of fleeting operations. To quantify these advantages an integrated schedule design and fleet assignment model is developed for both standard planes and Clip-Air. The comparative analysis has shown that there is a potential increase in the transportation capacity thanks to modularity of Clip-Air. Since Clip-Air is only exists in a simulation environment, the parameters regarding the design of the Clip-Air are based on estimation. A sensitivity analysis should be carried out to see the impact of the estimated parameters on the results. The objective of this project is to identify the parameters to which the model is more sensitive.
Student: Joseph Abisaleh (SGC), June 08, 2012
Supervision: Bilge Atasoy
Positioning Clip-Air among other transportation systems as a multi-modal flexible aircraft
Clip-Air is an innovative air transportation system based on modularity. By design, loading units (capsules) can be detached from the carrying unit (wing) which has several advantages in terms of fleeting operations. To quantify these advantages an integrated schedule design and fleet assignment model is developed for both standard planes and Clip-Air. The comparative analysis has shown that there is a potential increase in the transportation capacity thanks to modularity of Clip-Air. To see the impacts of Clip-Air with a systematic perspective it is important to build analogies with existing transportation systems. The novel feature of Clip-Air is its flexible transportation capacity and it is similar to the case of railways where the train cars are assigned to the locomotives. Moreover, the multi-modal aspect of Clip-Air has strong similarities with maritime transportation where standard load units are transported in different transportation modes. Starting from these analogies we can analyze the similarities and differences between the existing transportation systems and new air transportation system Clip-Air. Similarities will strengthen the basis of this new idea and the differences will justify the need for new models and methodologies to analyze the added value of Clip-Air.
Student: Jonathan Blaiberg (SGC), June 08, 2012
Supervision: Bilge Atasoy
Tracking Pedestrians with WiFi Traces
Gathering data about pedestrian localization and tracking indoor is a hot topic today. WiFi traces to track pedestrian paths have been collected on EPFL Campus. The poor quality and the scarcity of WiFi localization precludes the use of traditional map matching methods. This project aims at adapting methods recently developed for Smartphone GPS data.
Student: Yusen Bian (SMA), June 08, 2012
Supervision: Antonin Danalet
Transportation mode choice models including word data
Recently, there has been an emphasis in the discrete choice literature on the introduction of attitudes and perceptions into discrete choice models. We are interested to see how these factors impact on choice.

A joint work with social scientists has led to the development of new surveys which include questions on the perception of transportation modes. In the framework of a survey conducted with CarPostal, the following question was asked:

"For each of the following transportation modes, indicate 3 adjectives that characterize them the best:

The car is: 1)________ 2)________ 3)________

The train is: 1)________ 2)________ 3)________

etc."

After this survey, a second online survey was conducted to collect information on how individuals would situate the adjectives reported in the first survey (e.g. stressful, relaxing, full, etc.) on a scale of comfort. For example, respondents were ask to rate adjective 'relaxing' on a scale from -2 to 2, with -2 indicating a total discomfort and 2 indicating a total comfort. In addition to this, they had to report some socio-economic information (languages spoken at home or work, mother tongue, level of education, etc.).

This project has two goals:

1. Model of the effect of the socio-economic information of the respondents of the second survey on their ratings of the adjectives.

2. Integrate the model developed in 1. into a discrete choice model to explain the transportation mode preferences.
Student: Peng Cui (SMA), June 08, 2012
Supervision: Aurlie Glerum
Analyse des frquentations de bus en Suisse
Ce projet de semestre porte sur une analyse des donnes de frquentation de plusieurs lignes de bus travers la Suisse. Il s'agit d'analyser le lien entre la frquentation et diffrents autres facteurs tels que la frquence de la ligne, des variables temporelles (jour de la semaine, heure de la journe), le type de ligne, le bassin (densit de population, emplois) ou encore le lien avec une gare CFF (prsence, mais aussi ventuellement frquence des trains dans cette gare, ou du moins type de gare). Ce projet permet de travailler sur des donnes d'une grande richesse, au sein d'un projet de recherche multidisciplinaire en cours en lien avec l'industrie et d'autres laboratoires de l'EPFL, en particulier la CEAT.
Student: Suzy Polka (SSC), February 17, 2012
Supervision: Antonin Danalet, Ythier Jeanne
Mode choice model for the city of Nice
The goal of this project is to implement a mode choice model for the city of Nice. This model will be used by a public transport operator in order to identify the potentials for increasing market shares for public transport in this area, based on a better understanding of the demand structure between public transport and car. The model will be estimated on data collected from a recent Household Travel Survey which sampled around 6500 households. Information on travel times and costs is available for trips performed by public transport but similar information has to be collected for private transport modes. The model will be implemented and estimated using BIOGEME, an estimation software developed by the Transport and Mobility Laboratory. The project will be conducted in collaboration with Veolia Transport.
Student: My Hang Nguyen (SGC), January 17, 2012
Supervision: Ricardo Hurtubia
In collaboration with Veolia Transport
Robustness and Recovery in Berth Allocation Problem
Student: Wei Li (SMGT), January 13, 2012
Supervision: Michel Bierlaire, Nitish Umang
Multi-modal transportation modeling for a new generation of aircraft: Clip-Air
Clip-Air is an innovative air transportation system based on modularity. By design, loading units (capsules) can be detached from the carrying unit (wing) which has several advantages in terms of fleeting operations. To quantify these advantages an integrated schedule design and fleet assignment model is developed for both standard planes and Clip-Air. The comparative analysis has shown that there is a potential increase in the transportation capacity thanks to modularity of Clip-Air. As a further investigation, the performance of Clip-Air is needed to be analyzed from a network perspective. Repositioning of Clip-Air capsules by other means of transport, specifically railways, will give this multi-modal network viewpoint. The repositioning is believed to increase the profit especially in case of unbalanced demand between airport pairs. Furthermore, since Clip-Air capsules are completely detached from the wing, the storage and transfer costs are expected to be reduced. The aim of this project is to develop an appropriate model for the repositioning of Clip-Air capsules and obtain an integrated fleeting model. The tasks of this semester project are the following: 1. Get familiar with the concept of fleet assignment and already developed models. 2. Make a literature search on empty container management models. 3. Develop an appropriate model for the repositioning of capsules to be integrated into the existing fleet assignment model. Repositioning model will determine the flow of empty capsules throughout the network considering the carrying costs of capsules by rail and storage cost of capsules at airports. 4. Obtain preliminary results for a few data instances. 5. Evaluate the effect of repositioning.
Student: Jonathan Blaiberg (SGC), January 13, 2012
Supervision: Bilge Atasoy, Matteo Salani
Optimizing Train Design in Capacitated Railroads
The problems of Train Design, Railroad Blocking and Train Assignment are fairly complex and often observed in the railroad industry. Efficient solution methods to these problems bring a huge potential to achieve enhanced operational performance and dramatically reduced costs. Blocking is defined as an activity where a set of shipments arriving at or commencing from a certain node station and departing to another particular node station, or further, are grouped together and sent across as the same train to minimize costs and exploit economies of scale. This problem has marked similarities with the airline scheduling which operates flights across a predetermined hub and spoke network. The problem considered here not only necessitates determining the right hubs and right trains to be scheduled on the network, but also scheduling the shipments on appropriate trains between the hub station yards and spoke station yards so that the overall costs are minimized. In the problem considered by us, we are given a network comprising a set of nodes and arcs. We are also given a set of shipments with their origin and destination nodes. We are given a range of costs such as the cost of car travel per mile, cost of train travel per mile, cost of starting a train, cost of grouping (also referred to as classifying or blocking) shipments at an intermediate station of a train, cost of train imbalance, cost of crew imbalance and the cost of a missed car that is not transported. It is also required that each train path overlaps one or more crew segments completely. Crew segment between two nodes will also always follow the shortest path between those two nodes. Thus a train cannot run on a section which is not on the path of a crew. Train imbalance is defined as the imbalance generated due to the fact that the number of outgoing and incoming trains at a node differ. Total train imbalance for the network is computed as the sum of imbalances at every node. Crew imbalance is generated due to the fact that a crew operates on a segment, but cannot find an operational train to return to their base. In addition to these considerations, there are specific requirements relating to the maximum number of trains that a shipment can travel on, the maximum number of blocking (or classifications or work events as referred in the problem statement) allowed for each train, the maximum number of trains allowed on any arc and the limitation on the maximum train length and tonnage. The main objective of our efforts would be to find a cost minimizing set of feasible trains that operate on one or multiple crew segments completely. We would also need to determine the least cost assignment of shipments to these trains. Term Project Deliverables Develop and implement a cost optimization single or multi-stage mixed integer program model Solve the model using a mix of relaxations and heuristics and report results on two data sets Prepare a comprehensive report on solution methods and future research tracks on this problem Apart from expertise in modeling, the student is expected to be able to implement his model and write his own codes in a standard programming language, such as Java or C# Project Organization The student would be expected to update us on the progress and revert to us for queries on a pre-decided day of the week. A midterm presentation would be scheduled on Thursday, 10.11.2011. At the completion of the term project, a report will be delivered (both in paper as well as electronic form) before the 13.01.2012. The final oral presentation for the term project will be scheduled on 11.01.2012. All the files, programs, codes, data and report associated with the project must be delivered before these date. All reports will be submitted in five copies to TRANSP-OR Lab, LUTS Lab, Dr. Prem Kumar and Mr. Burak Boyaci. One copy is for the student.
Student: Stefan Binder, January 12, 2012
Supervision: Michel Bierlaire, Prem Kumar, BURAK BOYACI, LUTS
Mobility identification from smartphone GPS data
We record data from smarpthones over 2 years from about 180 N95 smartphone user. The data includes, GPS, nearby Bluetooth devices, nearby WIFI spots, calendar, etc. This project aims at identifying mobility history (trips and destinations) from raw data. The project will be based on some existing solutions (with existing Java and Python code). And the student will improve or develop a new software to process the data, and generate and visualize the data. You can choose among Python, Java or C++ for the project. The objectives of the project are following (You can choose among them): 1. Identify trips (origin, destination and time) from the GPS data. 2. Detect transportation mode (car, or public transport, etc) from the trips.
Student: Denis Garcia (SMA), January 11, 2012
Supervision: Jingmin Chen, Ricardo Hurtubia
Measuring Passenger Satisfaction using [email protected] Survey
M1nd-set is an international market research agency specialized in air travel surveys and research. The company conducts regular passenger satisfaction surveys among the major fullservice airlines on different international market areas. M1nd-set manages reasonable passenger response rates and over a third of their survey is administered on business-class travelers who value passenger service.

At present, [email protected] survey captures the passenger satisfaction levels for different parameters, such as waiting times for check-in, boarding and in-flight services, staff friendliness, seat comfort, in-flight entertainment, meals, etc., that are perceived to drive the overall satisfaction of flying with a particular airline company. M1nd-set uses a specific method of calculating weights of these parameters, in the order of importance, and arriving at overall airline satisfaction levels. However we see some scope in applying more advanced scientific models to this data to gain a deeper perspective on the relative significance of these parameters. In addition, it is critical to understand if higher satisfaction levels with a particular carrier are actually translating into increased demands and thus increased revenues. This could also mean that we can test the sufficiency of the currently captured parameters and if there are some additional parameters that could be silently driving the demand.

M1nd-set is seeking fresh, innovative and out-of-the-box ideas to analyze the data mine at hand. They are willing to give access to the data base (under the umbrella of the enclosed NDA) to the student who brings an interesting research idea. Ideally the outcome of this research / analysis should be presented to airlines on conference(s) and/or published.

This project will be composed of the following stages: 1. Understanding basic airline industry domain, various players and [email protected] survey in that perspective 2. Literature review and discussion on new ideas 3. Preliminary data analysis 4. Application of fresh methodologies for data analysis, such as using Discrete Choice Models for analysis 5. Inclusion of additional questions on the survey that could provide new insights 6. Final presentation and report

Project will commence around mid February and would continue for about 4 1/2 months
Student: Lidija Stankovikj (SMA), June 30, 2011
Supervision: Bilge Atasoy, Aurlie Glerum, M1nd-set
Battery life modelling
In a context where electric vehicles are going to be released soon on the market, a high importance is given to the analysis of the degradation of their batteries. A key aspect of this research is to identify the factors that are responsible for a fast battery degradation, in order to give drivers advices on a battery-friendly driving.

The aim of the project is two-fold:

1) Refine an existing model of the state of charge of the battery, in order to identify which are the factors explaining the battery discharge and how they affect it.

2) Analyze the impact of different driving patterns and other factors on battery degradation. The student will have to get familiar with an existing battery life model owned by ZEM.
Student: Franois Anken (SMA), June 30, 2011
Supervision: Aurlie Glerum, Responsable externe ZEM
In collaboration with ZEM
Travel behavior models: an Abu Dhabi case study
In this project, the student aims at studying the state-of-the-art demand modeling method- ologies in the transportation field. She will learn discrete choice models, and relevant tech- niques, such as simulation, and apply them in the study of mode choice and/or driving behavior. An Abu Dhabi case study will be carried out.
Student: Laurne Aigrain (SGC), January 30, 2011
Supervision: Michel Bierlaire, Jingmin Chen
Prototype MATSim model for Brussels
Student: Sohrab Sahaleh (SGC), January 20, 2011
Supervision: Gunnar Fltterd, Ricardo Hurtubia
An analysis of a potential implementation and investigation Singapore's road pricing system within the MATSim transportation microsimulation
Student: Youssef Mezdani, January 20, 2011
Supervision: Gunnar Fltterd, Alexander Erath
Quantitative analysis of urban sustainability indicators
Student: Timothe Vincent (SGC), January 20, 2011
Supervision: Gunnar Fltterd, Ricardo Hurtubia
Estimation and simulation of bid-auction and choice location models
Student: Thibaut Dubernet (SGC), January 20, 2011
Supervision: Ricardo Hurtubia, Thomas Robin
The urban development effects of the construction of a metro line
Student: Aurlien Odobert (SGC), January 20, 2011
Supervision: Ricardo Hurtubia
Integrating demand functions inside optimization model
For the new air-transportation system CLIP-AIR we want to have a fleeting model which includes a demand model. The project requires 1. understanding of the concept, 2. implementing the model in a general purpose math language, 3. using a solver for the nonlinear mixed integer problem, 4. performing preliminary tests with given instances.
Student: Nancy Moret (SMA), January 05, 2011
Supervision: Bilge Atasoy, Matteo Salani
Integrating the latent attitudes into mode choice
In the classical mode choice modeling we have the modal attributes like time and cost, and the socio-economic information as explanatory variables. However there are unobservable variables like attitudes, perceptions, lifestyle etc. which are effective in travel behavior. In this project aim is to come up with integrated choice models including these latent variables.
Student: Lidija Stankovikj (SMA), January 05, 2011
Supervision: Bilge Atasoy, Aurlie Glerum
Analysis of Electric Vehicle Data
The project aims to apply statistical analytics and basic mathematical techniques to analyze a unique data set of electric vehicle usage, and a cutting edge battery life model. The student will analyze the impact of different driving patterns and other factors on battery degradation a key question facing the transportation industry today. The student will be exposed to the latest developments in mobility and electric vehicle development, and will have constant access to supervisors that will guide him / her to better understand the technical and industrial context of the project. The student will have great autonomy in choosing the mathematical and analytic techniques he/she is familiar with to run simulations, stress the data sets, and incorporate new analytical models as necessary.
Student: Parmeet Singh Bhatia, January 05, 2011
Supervision: Aurlie Glerum, Responsable externe ZEM
In collaboration with ZEM
Optimizing Staffing Plans at Airports
Minimizing operating costs for maintaining ground personnel at airports is a complex problem due to uneven flight activities, passenger service expectations and staffing inflexibilities due to shift durations. In this work, we would use the flight schedule and service criteria to develop a method to find optimal shift timings that considers non-productive time due to activity changeovers, the mix of full-time and part-time workers and passenger waiting time criteria.
Student: Franois Anken (SMA), January 05, 2011
Supervision: Prem Kumar
Insertion d'un nouveau moyen de transport arien sur la base d'avions de transport existants
Insertion d'un nouveau moyen de transport arien sur la base d'avions de transport existants Il s'agit de crer une base de donnes comprenant les avions de ligne les plus reprsentatifs utiliss par les diffrentes compagnies dans les transports de passagers et de fret. Il faudra remplir un tableau par types et constructeurs d'avions: en tonnages, surfaces alaires, masses max et min, charges, nombre de passagers, altitudes, rayon maximal de chacun, ainsi que les vitesses de dcollage, de croisire et d'atterrissage. La recherche, la mise en place et le classement de l'ensemble de ces donnes permettront de faire des correspondances par type de missions, comme les courts courriers, moyen-courriers et long-courriers. Ceci constitue l'tape N1 Cette classification permettra de dfinir l'tape N2, qui sera exprime par une srie de graphiques faisant correspondre les caractristiques aux missions. Par exemple rechercher des correspondances de surfaces alaires et de poids ou de surfaces et de masses en corrlation un rayon daction maximum, etc. Les tudiants chercheront et exprimeront ces concordances entre ces types d'avions par une srie de graphiques explicites sachant que les donnes collectes sont optimises aux maximums par les lois de la mcanique des fluides, vritable base commune tout constructeur d'avions. L'tape N3 consistera, avec les donnes qui seront remises aux tudiants, faire un tableau runissant un maximum dlments concernant le projet Clip-Air. Ceci au travers des deux publications existantes qui seront remises aux tudiants. Ces donnes seront compltes avec l'aide des personnes travaillant dans le projet. L'tape N4 sera d'insrer le projet Clip-Air dans les graphiques de l'tape N2. Ceci permettra aux tudiants de tirer des conclusions qui situeront le projet Clip-Air par comparaison, dans: - la classe oprationnelle la plus proche -son potentiel de transports -d'valuer "les manques" ou "les trop", qui permettrait de correspondre une classe oprationnelle donne (en tenant compte de facteurs spcifiques au projet Clip-Air qui seront discuts cette tape).
Student: Laurene Aigrain & Dethier Daphn (SGC), January 05, 2011
Supervision: Claudio Leonardi, Bilge Atasoy
Destination choice models for a free bicycle system
Several cities around the world are implementing free bicycle systems that consist of a network of stations were bicycles are available for users to pay and ride. Usually these systems allow users to return the bicycle in a different station from the one where it was taken. The project consists in estimating destination choice models that should allow to predict the station where a bicycle will be returned, given the origin, the time of the day, the length (in time) of the trip and the characteristics of the surroundings of each station. The project considers 3 main stages: Data analysis, model estimation (with the software BIOGEME) and model validation.
Student: Zehra Onen (SMA), June 30, 2010
Supervision: Ricardo Hurtubia, Thomas Robin
Simulateur de mouvements de pitons
Les modles de mouvement de pitons permettent de simuler le comportement de foules. Ils sont utiles dans le domaine de la scurit, de l'vacuation, mais galement pour la plannification urbaine. En effet de nos jours pour des raisons environementales, de nombreux centres-ville sont interdits aux voitures et transforms en zones pitonnes. Ces modles servent simuler le comportement des pitons dans des contextes urbains prdfinis afin de choisir l'amnagement qui sera le mieux adapt. Ce projet concerne l'implmentation d'un modle de mouvements de pitons, modle dvelopp l'EPFL au sein du laboratoire Transp-or
Student: Viljami Laurmaa (SMA), June 20, 2010
Supervision: Thomas Robin, Javier Cruz, Mamy Fetiarison
Optimisation d'horaires ariens
Dans le cadre du transport arien, qui est sujet de nombreux retards, il est essentiel de planifier les horaires de manire mieux contrler ces retards. Le but de ce projet est d'tudier, implmenter et tester un modle d'optimisation qui maximise le temps de connexion des passagers, afin de rduire le nombre de connections rates dans le cas o des retards sont observs.
Student: Sezin Afsar (SMA), June 04, 2010
Supervision: Niklaus Eggenberg
Algorithme de gnration de vols de repositionnement pour amliorer la rparation d'un horaire perturb
Dans le domaine du trafic arien, un problme rcurrent est de rparer un horaire perturb suite des vnements tels que de mauvaises conditions mtorologiques, une dfaillance technique d'un avion, etc. La problme de rparation d'horaire est largement tudi dans la littrature. Malheureusement, si la majorit des approches existantes considre la possibilit d'intgrer des vols dits "de repositionnement" (des vols additionnels qui ne sont pas prvus initialement), aucune d'entre elles n'tudie la gnration des vols considrer. Il s'agit donc d'laborer un algorithme permettant de gnrer de bons candidats amliorer la solution de rparation, sans pour autant considrer tous les vols de repositionnement possibles.
Student: Sabine Luisier (SMA), June 04, 2010
Supervision: Niklaus Eggenberg
The Tactical Berth Allocation Problem: hierarchical vs integrated models in the context of container terminal operations
Container terminal operations have received increasing interest in the scientific literature over the last years and operations research techniques are more and more used to improve terminal's efficiency and productivity. In particular, the simultaneous optimization of decision problems that are usually solved hierarchically by terminal's planners represents nowadays a promising research trend. The Tactical Berth Allocation Problem (TBAP) deals with the integration of the berth allocation problem (BAP) and the quay crane assignment problem (QCAP). It aims to schedule incoming ships over a time horizon, assigning them a berthing position and a certain quay crane profile (i.e. number of quay cranes per working shift). These decisions are strictly correlated, since the number of quay cranes assigned to a ship affects its expected handling time, and thus has impact on the scheduling in the berth allocation plan. The problem has been modeled as a mixed integer program and housekeeping costs generated by the berth assignment are taken into account by a quadratic term in the objective function. The aim of this project is twofold: (i) to compare the integrated approach (modeled in TBAP) to the standard hierarchical approach (first solve BAP, then QCAP); (ii) to analyze the impact of different BAP objective functions on yard operations, by taking into account housekeeping costs.
Student: Luca Furrer (SMA), June 04, 2010
Supervision: Ilaria Vacca
Testing the algorithm for generating path observation from GPS data
This project aims at using and testing an innovative probabilistic path observation generation algorithm on location data and comparing against state-of-the-art map matching algorithms.
Student: Jensen Anders Fjendbo (SGC), January 15, 2010
Supervision: Michel Bierlaire, Jingmin Chen
Visualization of Cell Phone Data on Google Earth
This project aims at providing to our data collection campaign participants a convenient and lively way of visualizing cell phone data which is collected from their cell phones and stored on web servers. Various kinds of data are display on google earth to show chains of users activities, with time-space information provided by GPS data.
Student: Raoul Neu (SIN), January 15, 2010
Supervision: Jingmin Chen
Analysis of Transport Mode Choice in Trieste
In the transport planning context, survey data have been collected between 2002 and 2003 in Trieste. Three modes of transport have been considered : car, motorcycle and bus. Transportation is a major application field of Discrete Choice Models (DCM) since they can capture lots of situations where a choice is performed. In order to study the mode choice of transport users, DCMs have been developed and calibrated with these collected data. Analysis of the results pointed out some prediction inaccuracy. This means that improvement of the model still has to be made. The aim of this project is to study the travelers choice behavior with the existing models and investigate possible improvements of the models specification.
Student: Alexandre Khelifa (SGC), January 15, 2010
Supervision: Mamy Fetiarison
Un modle prliminaire d'UrbanSim pour Lausanne
Depuis longtemps l'importance de l'interaction entre la forme urbaine, les infrastructures de transports et la demande de transport a t reconnu. Par contre ce n'est que rcemment que des mthodes analytiques tel que des systmes de modlisation de transport et d'occupation du sol ont t dveloppes pour pouvoir mieux comprendre ces interactions. UrbanSim est un systme de modlisation d'occupation du sol qui devient de plus en plus utilis. L'objectif de ce projet est de monter un modle prliminaire d'UrbanSim pour la ville de Lausanne.
Student: Sarah Droz (SGC), January 15, 2010
Supervision: Ricardo Hurtubia
Modeling the effects of spatio-temporal flexibility in activity scheduling
In travel demand modeling there is an increasing interest in understanding and modeling the planning or scheduling of activities over space and time. This is due to the fact that a better understanding of activity scheduling processes will contribute to the development of activity scheduling models and an understanding of the short (week) dynamics. In the literature, very little is known about the activity scheduling processes and the spatio-temporal flexibility of activities. For an individual, each (work, leisure, shopping) activity in a day can be classified according to its level of flexibility in routine, pre-arrange and spontaneous in the two dimensions, space and time. Based on his spatio-temporal constraints, an individual is making a decision to assign its degree of flexibility. The objective of this research is to investigate how socio-demographics of individual/households, life-stage, ICT access (cell phone/internet), accessibility to services, and activity attributes (time, duration, week day) affect the spatio-temporal flexibility of out-of home activities. The analysis will be done using data from a survey conducted in Qubec City from 2002-2006 to estimate activity-choice models.
Student: Laetitia Bettex (SGC), January 15, 2010
Supervision: Ricardo Hurtubia
Modeling the link between transport and land-use with UrbanSim
Location of activities in the city has an important effect on travel demand and on the transport systems performance. Land-use models are used to forecast location in the city, in order to help decision-making for urban and transport planning. Because of its flexibility, UrbanSim is an increasingly popular alternative for integrated land-use and transport modeling. However, UrbanSim is not exactly an integrated model but a land-use model that works together with a transportation model. This makes relevant to understand if the interaction between transport and land use is properly modeled by UrbanSim. The objective of this project is to implement the latest stable version of UrbanSim for the city of Brussels, using data that was already collected to implement an older version of the model. The project also considers a deeper analysis on how UrbanSim accounts for the relation between transport and land use.
Student: Peter Goodings (SGC), January 15, 2010
Supervision: Ricardo Hurtubia
Game theory applied to ambush avoidance
This project continues along the stream of the previous project "Minimizing risk of ambush for vehicle routes". The goal of the project is to apply game theory to determine the optimal mixed strategy to select as set of vehicle routes. Each route comes along with the expected maximal payoff when the route is implemented. A strategy to determine the game matrix is to be conceived.
Student: Peter Goodings Swartz (SGC), January 15, 2010
Supervision: Matteo Salani
Finance and Discrete Choice Models
Default risk evaluation for credit cards and mortgages, credit ratings determination, probabilities estimation of acquisitions and mergers of firms (used to guide investment), are very complex financial tasks. They require a very good expertise of the field. This is due to the high number of information that decision makers need to account for. In addition, information could be very heterogeneous in term of source and impact on the decision itself. Models can help financial decision maker in their tasks. For instance, classification methods are mainly used to model financial decisions. Discrete Choice Models (DCM) have been successively applied to many fields in the past 40 years, such as transportation or marketing. Compared to machine learning methods they have the advantages to explicitly model causal relations between dependant variables and exogenous attributes, and to avoid data over-fitting. DCMs start to be used in the financial field. The aim of this project is to investigate DCM use in a financial context, then find relevant problem to which it can be applied and finally illustrate those concepts on a real case study.
Student: Ins Azaiez (SMA), January 10, 2010
Supervision: Thomas Robin
Mode choice modelling with qualitative aspects inclusions
Discrete Choice Models (DCM) are widely used in transportation to explain mode choice. Objective attributes have been commonly used to describe the different alternatives, such as travel time or travel cost. Kaufmann et al. (2001) have conducted some transportation mode choice surveys. In addition to the classical data collection, they have included questions about experiences and perceptions of transportation modes. The obtained database is large and contains lots of respondents sociological aspects, difficult to apprehend in a classical DCM, due to their qualitative nature. In order to go through such issues, J.L.Walker (2001) has proposed a new modelling framework based on DCM and latent concepts. She used schemes to describe links between variables. The aim of the project is to identify complex causal links between sociological aspects and transportation mode choice, based on the data cited above, and then derive a new modelling framework.
Student: Zehra Onen (SMA), January 10, 2010
Supervision: Thomas Robin
Exprience des moyens de transport et choix modal
Lorsque lon choisit dutiliser un moyen de transport plutt quun autre, indpendamment des temps de dplacements et des cots compars des alternatives en prsence, lusager a des prfrences. Celles-ci sont lies son exprience des moyens de transport et la perception quil en a. Les modlisations du choix modal ne rendent quimparfaitement compte de cette dimension, intrinsquement qualitative. Pour la traiter, le LaSUR a dvelopp depuis une dizaine dannes un dispositif de citation dadjectifs, appliqu de gros chantillons de population lors denqutes de mobilit. Nous proposons comme sujet de Master de modliser sur la base des donnes disponibles, limpact de lexprience sensorielle des moyens de transports sur le choix modal.
Student: Seyed Tavakoli (SGC), June 30, 2009
Supervision: Michel Bierlaire, Thomas Robin
Minimizing risk of ambush for vehicle routes
Vehicle routing problems are known problems in which a fleet of vehicles is routed through a network to collect (or deliver) items from customers. The objective of routing problems in an hostile environment is to prevent ambushes (for valuables deliveries). The goal of routing under potential attacks is to design safe routes in order to minimize the risk of being exposed to dangers. This project is designed to investigate and implement innovative routing strategies.
Student: Gatan Duyckaerts, Peter Goodings Swartz (SGC), June 19, 2009
Supervision: Matteo Salani
Analysis of consumer behaviour in terms of product preference/choice.
This project will be conducted in collaboration with the Nestl Research Center. Quantitative studies are frequently used in the food industry in order to determine whether a product is superior to its competitor(s). Many types of studies exist, but this project focuses on a specific type of study involving only two products and in which consumers are asked to evaluate both products with regard to different criteria (such as flavour, appearance, texture, etc.) and to choose the product they prefer. Socio-demographic characteristics as well as information on consumption habits are also collected during the test. The aim of this project is to explore a database containing about 100 studies that have been conducted over the past years on a same product category in different countries. Such studies are usually analyzed by the means of descriptive statistics in order to answer the two following questions: -Which product do consumers prefer? -Why do consumers prefer this product? In this project, we propose to analyze the studies by making use of discrete choice modelling. This technique allows understanding and modelling the behaviour of consumers in a quantitative way when they are exposed to such choice situations, i.e. the choice between two products. Discrete choice models should allow identifying and quantifying the effect of both product attributes and consumer characteristics on the preference.
Student: Aurlie Glerum (SMA), June 12, 2009
Supervision: Thomas Robin, Michal Thmans
In collaboration with Nestl
Effets sur le trafic d'une nouvelle jonction autoroutire Chavannes
Actuellement, il ny a quune seule jonction autoroutire entre celles de la Blcherette et de Malley pour desservir lOuest Lausannois : la jonction de Crissier. Or deux nouvelles jonctions sont en projet Ecublens et Chavannes. Selon les choix effectus pour les voies daccs lautoroute, ces jonctions peuvent diminuer les nuisances en rduisant les kilomtres parcourus, surtout hors autoroute ou les augmenter en amenant du trafic de transit indsirable dans des zones prserver. Le projet consiste examiner des variantes damnagement du rseau dans le sud de Renens, ventuellement en suggrer de nouvelles, et valuer ces diverses variantes en saidant du logiciel Emme de simulation du trafic.
Student: Annie Faniry Andriamanorohasinjafiniarivo Ravalitera (SGC), June 05, 2009
Supervision: Jean-Pierre Leyvraz
Implementation of attribute processing strategies in advanced discrete choice models
Student: Carol Kirchhofer (SMA), February 03, 2009
Supervision: Michel Bierlaire, John Rose, ITLS, University of Sydney
Extending the framework for MEV discrete choice models
Student: Lorenza Santini (SMA), February 03, 2009
Supervision: Michel Bierlaire, Mogens Fosgerau, DTU, Denmark
Estimation of household location choice models
Location of activities in the city has an important effect on travel demand and on the transport systems performance. Land-use models are used to forecast location in the city, in order to help decision-making for urban and transport planning. To predict the location of agents its first necessary to estimate location choice models (which are usually a part of the land-use model). These models are estimated through maximum-likelihood methods from observed location data. The aim of the project is to estimate several different household location choice models for the same city, using different specifications for the households utility function and testing different methods to deal with attributes thresholds in the utility. The estimations will be made using Biogeme, databases will be provided.
Student: Antonin Danalet (SMA), January 12, 2009
Supervision: Ricardo Hurtubia
Est-il possible de rduire le nombre de panlistes dans une tude DTS (Dominance Temporelle des Sensations) sans altrer la qualit des rsultats?
La mthode de la Dominance Temporelle des Sensations (DTS) permet de suivre lvolution des perceptions sensorielles en bouche au cours de la mastication dun produit alimentaire. Elle consiste simplement demander au panliste de slectionner la caractristique dominante du produit au cours de la mastication. Par exemple, si le produit est une barre de crales, le panliste doit indiquer au cours de la mastication dune bouche si le produit est plutt croquant, friable, collant ou pteux. Cette caractristique dominante peut changer au cours du temps. Sur lensemble du panel, cette mthode permet de dessiner les courbes dvolution de chaque descripteur sensoriel au cours du temps. Cette mthode est relativement rcente et les premires tudes ont t ralises avec un nombre de panlistes trs important (environ 50 valuations de chaque produit). Lobjectif de ce projet est de tester si cette mthode donne des rsultats aussi satisfaisants avec un nombre de panlistes rduit. Le travail consistera donc : se familiariser avec la mthode DTS et la construction des courbes, dvelopper un programme permettant de construire ces courbes pour un sous-ensemble de panlistes, gnrer les rsultats pour plusieurs tailles de panel en utilisant les donnes de plusieurs tudes DTS existantes au centre de recherche Nestl, proposer une mthode pour comparer les rsultats des panels de taille rduite avec le panel complet, appliquer cette mthode et analyser les rsultats, rdiger un rapport prsentant la dmarche et mettant en avant les principales conclusions.
Student: Glerum Aurlie (SMA), January 11, 2009
Supervision: Thomas Robin, Michal Thmans, Nicolas Pineau (Nestl Research Center)
In collaboration with Nestl
Minimizing risk of vehicle routes in valuables collection from banks
Vehicle routing problems are well known combinatorial problems in which a fleet of vehicles is routed through a network to collect (or deliver) items from customers. The objective of routing problems is to minimize operational costs in terms of distance traveled such that all customers are visited and all operational constraints are respected. In the context of logistic services delivered to credit institutes, e.g. the collection of valuables or distribution of coins, the total distance traveled is no more the primary objective to optimize. For logistic companies is more convenient to design {\em safe} routes in order to minimize the risk of being robbed. Beside risk minimization, logistic companies commit to provide a certain level of service to credit institutes in terms of pick-up frequencies.
Student: Fabrice Piat (SGC), January 11, 2009
Supervision: Matteo Salani
Is it possible to reduce the number of panelists in a TDS (Temporal Dominance of Sensations) study without altering the quality of the results?
Student: Aurlie Glrum (SMA), January 06, 2009
Supervision: Thomas Robin, Michel Bierlaire, Michal Thmans
In collaboration with Nestl Research Center
La modlisation de transport de la rgion Lausannoise avec PTV Vision
La modlisation de transport est un aspect fondamental de la planification de transport. Plusieurs diffrents logiciels existent pour modliser la performance des rseaux de transports urbains. Les logiciels EMME, TransCAD et PTV Vision sont parmi les plus utiliss au monde pour modliser le trafic. Au sein du laboratoire TRANSP-OR existent dj des modles EMME et TransCAD de Lausanne. Pour faire connatre aux tudiants de l'EPFL les diffrentes options disponibles pour la modlisation de transport, le dveloppement d'un modle PTV Vision de Lausanne est prvu. Le but de ce projet est d'utiliser les donnes existantes pour monter un modle PTV Vision de la ville de Lausanne.
Student: Chen Lu (SSC), July 15, 2008
Supervision: Zachary Patterson
Dconvolution de signaux gochimiques
Le but gnral de la mthode est de connatre, partir du relev de la concentration d'un lment chimique dans l'air, la part naturelle, et la part anthropique(celle de l'homme, par exemple la pollution). Concrtement partir de donnes existantes, nous voulons implmenter et rsoudre une maximisation de vraisemblance, afin de dterminer les deux composantes du signal.
Student: Gfeller Nicolas (SMA), June 30, 2008
Supervision: Thomas Robin
Dynamic Traffic Assignment in Lausanne
Dynameq is a new breed of equilibrium dynamic traffic assignment (DTA) for use on large, congested networks, distributed by INRO (www.inro.ca). Dynameq gives planners a view into dynamic traffic conditions, and provides rational scenario comparisons that are only possible with an equilibrium-based solution. It has been successfully used in cities like Montral. The purpose of the project is to develop a Dynameq model of Lausanne. The challenge consist in gathering relevant available data, and identify additionnal data which would need to be collected. The objective is to obtain a prototype model for which the missing pieces are well identified. The project will be conducted at the Centre de Recherche sur les Transports (University of Montral), in close collaboration with the staff of INRO.
Student: Vidaud Marine (SGC), June 20, 2008
Supervision: Michel Bierlaire, Mahut Michael
Development of mode choice models in Trieste
The discrete choice models are very relevant concerning the transport mode choice. A survey on the daily trips has been conducted in Trieste, Italy, in 2000-2001, led by the municipality. The study of the resultant data base will permit to understand the Triestans habits, along with developing transport mode choice models. The aim of this project is to analyse real data (the dataset described above), perform a complete modelling process, together with statistical tests, and prepare appropriate documentation.
Student: Violin Alessia (SMA), June 17, 2008
Supervision: Michel Bierlaire
Optimization of Lausanne's traffic signal timings
The aim of this project is to optimize the traffic signal settings of the network of the city of Lausanne.
The phases of the project are as follows. The student will:
- provide an overview of the existing formulations for this optimization problem;
- propose a formulation, this formulation may be taken from exisiting studies;
- implement the problem on a subnetwork of the city of Lausanne, the student will be given access to a Matlab code which may need to be adapted according to the formulation;
- analyse and discuss the results.
Basic knowledge of Matlab is required for this project.
Student: Yanjun Zhang (SGC), June 06, 2008
Supervision: Carolina Osorio Pizano
Analyse gographique pour l'implmentation d'un prototype de modle intgr de transport et d'occupation du sol pour la rgion lausannoise
UrbanSim est un systme de modlisation transport et occupation du sol qui devient de plus en plus utilis. Se projet se fera dans le cadre d'un projet plus large pour dvelopper un prototype d'un modle intgr de transport et d'occupation du sol pour la rgion lausannoise. Ces modles demandes beaucoup de donnes socio-dmographiques ainsi que gographiques pour la rgion d'tude. L'objectif de ce projet est de contribuer au dveloppement des donnes pour le tableau `Gridcells' du modle en dveloppement. En particulier le projet aura comme but : l'identification des donnes manquantes pour ce tableau, l'identification des sources pour les donnes manquantes, le traitement de ces donnes et l'incorporation dans le tableau `Gridcells.'
Student: Bettex Laetitia (SGC), May 30, 2008
Supervision: Zachary Patterson
Algorithmes On Line pour le CVRP avec Demandes Alatoires
EN - The project copes with the capacitated vehicle routing problem (CVRP) in the case of unknown demand volumes of the customers. The aim is to abord the problem with an on line strategy. The student will first familiarize with CVRP and on line algorithms and then develop and test an on line strategy.

FR -Le projet propose d'aborder le problme de trounes de vhicules avec contraintes de capacit (CVRP) dans le cas o les volumes des demandes des clients ne sont pas connus. Il s'agit ici d'aborder le problme avec une approche on line. L'tudiant devra d'abord se familiariser avec le CVRP et les problmes on line puis dvelopper et tester une stratgie.
Student: Boukriba Sami (SSC), May 23, 2008
Supervision: Niklaus Eggenberg
Dterminer les vols de repositionnements pour un horaire arien perturb
Lors de l'excution d'un horaire, il arrive souvent que ledit horaire devienne irralisable. Dans le cas du transport arien, pour ragir un tel vnement, il s'agit de retarder ou d'annuler certains vols, voir mme faire des trajets " vide" pour repositionner les avions. Dans ce projet, il s'agit d'laborer une technique permettant d'identifier, tant donn l'horaire devenu irralisable et l'tat des avions, quel(s) vol(s) de repositionnement sont considrer pour minimiser autant les cots que les annulations de vols et les retards.
Student: Karker Amin, Tournier Sebastien (SSC), May 23, 2008
Supervision: Niklaus Eggenberg
Analyse de donnes requises pour un modle intgr de Lausanne
Depuis longtemps l'importance de l'interaction entre la forme urbaine, les infrastructures de transports et la demande de transport a t reconnu. Par contre ce n'est que rcemment que des mthodes analytiques tel que des systmes de modlisation de transport et d'occupation du sol ont t dveloppes pour pouvoir mieux comprendre ces interactions. UrbanSim est un systme de modlisation qui devient de plus en plus utilis. Ce projet reprsente une phase prparatoire pour le dveloppement d'un modle UrbanSim pour la rgion lausannoise.
Student: Maret Jonathan (SGC), January 18, 2008
Supervision: Zachary Patterson
Calibration of on Integrated Transportation Land-use Model - UrbanSim for Brussels
Transportation and infrastructure planning requires a good understanding of future transportation demand. Traditionally, transportation demand analysis has focused primarily on the transportation system independent of its relationship with land-use and urban form. Integrated Transportation and Land-use Models (ILUMs) explicitly model these interactions. The TRANSP-OR laboratory currently has an operational land-use model (UrbanSim) for the city of Brussels in Belgium. The purpose of this project is to is to fine-tune UrbanSim for Brussels by improving the location, developer and land-price models. The analysis will take advantage of GIS land-use data for Brussels. The project will include an application of the modeling system to a practical transportation and land-use planning problem. The project will be undertaken in cooperation with Stratec of Brussels.
Student: Zemzemi Fatima and Stoitzev Iordanka (SMA), January 18, 2008
Supervision: Zachary Patterson
Dynamic Traffic Assignment in Lausanne (prtude)
Le logiciel Emme permet de comparer des variantes de rseaux routiers urbains en se basant sur une affectation des dplacements des automobilistes de leur origine leur destination. Cette affectation est macroscopique (on considre des flux globaux de dplacements) et statique (les conditions dcoulement sont censes tre constantes durant toute la priode considre). Dautres modles comme Dynameq et AIMSUN se basent sur une affectation microscopique (chaque vhicule est modlis) et dynamique (les conditions peuvent changer au cours du temps). Ces modles demandent plus de donnes, mais permettent en change dvaluer des mesures comme des changements dans la gestion des carrefours et de reprsenter lvolution au cours du temps. Lagglomration lausannoise dispose dune banque de donnes Emme et, pour sa partie la plus centrale, dune banque AIMSUN. Le projet de master, qui se droulera chez INRO Montral, consiste crer partir de ces donnes et ventuellement dautres donnes disponibles, une banque de donnes Dynameq, de documenter le processus utilis et didentifier les donnes supplmentaires ncessaires pour amliorer la qualit de cette banque.
Student: Vidaud Marine (SGC), January 14, 2008
Supervision: Michel Bierlaire, Jean-Pierre Leyvraz
Modle de choix discret
Les modles de choix de discret sont trs pertinents concernant le choix de mode de transport. Une enqute sur les trajets a t mene daot dcembre 2003 par lETHZ, Frauenfeld et ses abords dans le canton de Thurgovie, en Suisse. Celle-ci concerne 230 personnes, suivies pendant 6 semaines et provenant de 99 mnages. Ltude de la base de donnes rsultante permettra de mieux connatre les habitudes des voyageurs, ainsi que de dvelopper un modle de choix de mode de transport. Le but de ce projet est de sinitier lanalyse des modles de choix discret leurs applica- tions sur des donnes relles, en loccurence les donnes dcrites ci-dessus.
Student: Weber Caroline (SMA), January 09, 2008
Supervision: Thomas Robin
Modle de classes latentes en analyse de choix discret
Le but de ce projet est de s'initier l'analyse des modles de choix de classes latentes et leurs application sur des donnes relles en vue de prparation d'un projet de master concernant le choix de billets d'avion.
Student: Kirchhofer Carol (SMA), January 07, 2008
Supervision: Michel Bierlaire
Modles GEV et MEV
Les modles de choix discrets ont t prdominants dans l'analyse de transport durant ces dernires annes. Le modle MNL(Multinomial Logit) est particulirement apprci grce ses capacits calculatoires et sa forme analytique. Il est obtenu partir de la loi de Gumbel, appele aussi loi des valeurs extrmes de type I. D'une part, une version multivarie (MEV) a t propos par McFadden (1978). D'autre part, diffrentes lois monovaries aux valeurs extrmes peuvent tre rassembles en utilisant la spcification dite gnalise (GEV). Le but de ce projet est d'une part de matriser les concepts lis aux modes MEV (Mul- tivariate Extreme Value) et GEV, afin d'investiguer la possibilit de combiner les deux. D'autre part, les modeles d'utilit multiplicatifs (Fosgerau & Bierlaire 2007) permettent de faire un lien avec les modles GEV ( Fosgerau, short note). Ce lien sera aussi investigu, en vue de la ralisation d'un projet de master li ce sujet avec Mogens Fosgerau Copenhague.
Student: Santini Lorenza (SMA), January 07, 2008
Supervision: Michel Bierlaire
Graphical interface for a JAVA simulator
Project aim: The aim of this project is to implement a graphical interface for an existing discrete event simulator.
Description:
Numerical simulators are often used to evaluate the impact of changing the characteristics of an already existing system (e.g infrastructure improvements in highways). Such computer models can also be used to evaluate modifications under hypothetical scenarios that would be difficult to observe in the real world (e.g. predict future congestion levels, based on demographic forecasts.) These numerical simulators mimic the behavior of complex systems, yielding performance measures that are rich in information. In order to take full advantage of this detailed information, and to be able to summarize it in a user friendly manner, a graphical interface is crucial.

The aim of this project is to implement a graphical interface for a numerical simulator.
The student will be given a JAVA implemented discrete event simulator that studys the flow of units (e.g. vehicles, pedestrians) through a network (e.g. highways, corridors). The aim will be to implement an attractive graphical interface that will summarize and illustrate the main performance measures that the simulator yields.
Student: Fetiarison Mamy Nirina, August 31, 2007
Supervision: Carolina Osorio Pizano
Etude du comportement d'achat des consommateurs
Le comportement des consommateurs durant la dmarche d'achat est un phnomne compliqu et propre chaque personne, qui regroupe des paramtres observables, comme le choix final, et des composants latents, comme p. ex. le raisonnement avant ce choix. La modlisation de l'impact de marketing cibl (via design des panneaux publicitaires) est du plus grand intrt pour la comprhension de ce comportement. Le but de ce projet est d'unifier les donnes venant des sources diffrentes et d'amliorer le modle de la dmarche complte d'achat bas sur les donnes d'une tude relle de traage des yeux.
Student: Alexandre Xavier (SMA), June 29, 2007
Supervision: Michel Bierlaire
Study of Optical Flow techniques for motion estimation in video sequences
In image processing, when dealing with video sequences, it is usually very useful to have an estimation of the motion in order to obtain spatio-temporal information. To achieve this, the most common approach is to compute the optical flow, which is the velocity field of the apparent motion between frames.
The aim of this project is to study the most relevant approaches to compute the optical flow and implement them using C, C++ or JAVA.
Student: Epely-Chauvin Gal (SGC), June 29, 2007
Supervision: Javier Cruz
Deadlock detection
The spillback phenomenon frequently arises in urban traffic networks under congestion. If the network contains loops then spillback is a potential source of dealocks, also known as gridlocks. In this project the student will be given acces to a network simulation tool. Based on empirical studies of simple networks under deadlock the student will:
1) identify preformance measures that indicate the presence of deadlocks
2) propose analytical performance measures that may help detect deadlocks
3) validate the proposed performance measures
Student: Anken Nicolas (SMA), June 29, 2007
Supervision: Carolina Osorio Pizano
A Survey of Active Integrated Land-use Models around the World
Traditional travel demand modeling has tended to ignore the important interactions between urban form and transportation demand and the performance of the transportation system. Integrated transportation land-use models (ILUMs) explicitly model these interactions. These models are becoming increasingly popular. Better understanding these models requires a better knowledge of how they are applied in different contexts. The purpose of this project is to undertake research to provide a summary of active ILUM applications around the world.
Student: Vignon Olayitan (SGC), June 29, 2007
Supervision: Zachary Patterson
Simulation of finite capacity queueing networks
The aim of this project is to reproduce via numerical simulation the behaviour of a network of finite capacity queues. The simulation tool is to be both implemented and validated. The tool shall be used to analyse the HUG (Geneva University Hospital) hospital room network.
Student: Meier Pirmin (SMA), February 17, 2007
Supervision: Carolina Osorio Pizano
An Integrated Land Use Model Application to Brussels using UrbanSim
The interrelationship between urban form, transportation infrastructure and transportation demand has long been recognized. More recently, analysts have been developing Integrated Transportation and Land-use Models (ILUMs) to better understand and quantify these interrelationships. As a result, several modeling options are now available to develop ILUMs. One such model has already been applied to the city of Brussels in Belgium. UrbanSim is an increasingly popular alternative for integrated land-use modeling. The current study is intended to test the feasibility of applying UrbanSim to a city for which a fair bit of transportation and land-use data already exist, within a medium-term project timeline. The project is being undertaken in collaboration with Stratec in Brussels.
Student: Samartzis Lefteris (SMA), February 17, 2007
Supervision: Zachary Patterson
Modles de choix discret pour la reconnaissance des expressions faciales statiques
Student: Danalet Antonin (SMA), February 16, 2007
Supervision: Michel Bierlaire
Analysis of a Recovery Network for an Airline Recovery Method
Airline schedules are often disrupted because of unforeseen events like lateness or bad weather. If the schedule becomes unfeasible, the scheduler has to recover from the actual state in order to get a new feasible schedule, either by canceling or postponing flights or even rerouting planes. The model often used is a time-space recovery network which encodes several possible recovery decisions for each plane in the fleet. It's main disadvantage is its size. The student will have to familiarize with airline scheduling and network modeling in order to analyze and simplify as much as possible the recovery network. The aim would be be able to end up with a reduction algorithm that throws network of smaller size by means of number of nodes and arcs amd eventually test the reduction algorithm on existing instances.
Student: Messina Daniele (SMA), February 16, 2007
Supervision: Niklaus Eggenberg
Etude des modles de mouvements de pitons
Etude des modles de mouvements de pitons, revue de la littrature, recensement des modles existants dans divers domaines tels que l'architecture, planification urbaine, gestion de grands vnements, vacuation...etc. Le but est d'identifier ces divers modles, dterminer leurs hypothses, contextes d'utilisation, et faire une tude comparative.
Student: Li Xiangchun (SMA), February 16, 2007
Supervision: Thomas Robin
Validation de modles de choix de route
Student: Anken Nicolas (SMA), February 15, 2006
Supervision: Emma Frejinger
Detection of behavioral inconsistency in Revealed Preferences surveys
Student: Laurence de Torrent and Ariane Wenger (SMA), February 01, 2006
Supervision: Michal Thmans, Michel Bierlaire
Detection of behavioral inconsistency in Revealed Preferences surveys
Student: Julie Marc (SMA), June 23, 2005
Supervision: Michal Thmans, Michel Bierlaire
Discrete choice models: development of case studies
Student: Frdric Anken (SMA), June 22, 2005
Supervision: Michal Thmans, Michel Bierlaire
Analysis of choice sets for route choice models
Student: Fortn Garcia Vernica, June 20, 2005
Supervision: Emma Frejinger
Analyse de l'approche sous-rseau pour la modlisation de choix de route
Student: Gilliron Fanny (SMA), June 20, 2005
Supervision: Emma Frejinger
Deterministic Correction of the Multinomial Logit Model for Route Choice Analysis
Student: Regis Cline (SSC), June 20, 2005
Supervision: Emma Frejinger
Different strategies for trust-region size management
Student: Emmanuel Leclercq (SMA), June 20, 2005
Supervision: Michal Thmans, Michel Bierlaire
Stated Preferences Survey for the choice of exchange university
Student: Eric Von Aarburg (SMA), June 20, 2005
Supervision: Michal Thmans, Michel Bierlaire
Utilisation d'un systme gographique pour un problme de logistique
Student: Emery Sarah and Roth Isabelle (SMA), February 15, 2005
Supervision: Emma Frejinger
Route Choice Analysis using GPS Data
Student: Lindbck Rolf (SSC), February 15, 2005
Supervision: Emma Frejinger, Michal Thmans
Optimisation of the Operating Suit
Student: Pivin Edward (SMA), February 15, 2005
Supervision: Emma Frejinger
Calibration of models for transportation demand in the context of real-time applications
Student: Mohamed Rhmari Tlemani (SSC), June 25, 2004
Supervision: Michal Thmans, Michel Bierlaire
Adaptation of GSM for unconstrained nonlinear optimization
Student: Anouck Brossard and Sarah Degallier (SMA), June 23, 2004
Supervision: Michal Thmans, Frank Crittin, Michel Bierlaire
On the convergence of multi-dimensional filter methods
Student: Lionel Dumartheray (SMA), June 23, 2004
Supervision: Michal Thmans, Michel Bierlaire
Identification of Coherent Behavior using Linear Programming
Student: Ittig Oliver (SMA), June 20, 2004
Supervision: Emma Frejinger, Michal Thmans
Dynamic O-D matrices estimation with iGSM
Student: Olivier Grandjean (SMA), February 01, 2004
Supervision: Frank Crittin, Michal Thmans, Michel Bierlaire
Testing GSM on unconstrained nonlinear optimization problems
Student: Mina Adel Latif (SSC), June 20, 2003
Supervision: Michal Thmans, Frank Crittin, Michel Bierlaire
Discrete choice models to capture drivers behavior in response to real-time traffic information
Student: Thomas Quentin Maillard (SMA), June 20, 2003
Supervision: Michal Thmans, Michel Bierlaire
Empirical analysis of the correlation in discrete choice models
Student: Steve Salom (SSC), February 01, 2003
Supervision: Michal Thmans, Michel Bierlaire
Activity organization and ICT
Selon indications ultrieures.
Student: Laetitia Bettex (SGC)
Fatal error: Uncaught TypeError: mktime(): Argument #4 ($month) must be of type ?int, string given in E:\Inetpub\transp-or.epfl.ch\web\wordpress\php\changeTimeStamp.php:49 Stack trace: #0 E:\Inetpub\transp-or.epfl.ch\web\wordpress\php\changeTimeStamp.php(49): mktime() #1 E:\Inetpub\transp-or.epfl.ch\web\wordpress\php\changeTimeStamp.php(43): getMonthName() #2 E:\Inetpub\transp-or.epfl.ch\web\wordpress\completedProjects.php(43): printDate() #3 {main} thrown in E:\Inetpub\transp-or.epfl.ch\web\wordpress\php\changeTimeStamp.php on line 49