Student Projects

DISAL offers a range of student projects its three main areas of expertise: Distributed Robotic Systems, Sensor and Actuator Networks, and Intelligent Vehicles. For more information about supervision guidelines and different types of student projects please refer to this page.

Towards Automatic Design of Controllers: Implementation of Machine-Learnable, Cooperative Behaviors for Khepera IV Robots

With the increased deployment of mobile robotics in a multitude of real life scenarios, robots must face more and more complex situations. However, every robotic platform needs to be carefully programmed to fulfill its task, which, today, is often still done manually. Such programming operation represents often an intense time investment and requests high expertise to eventually achieve the targeted performances in the real deployment, even more even more so if the system consists of multiple robots.

There are multiple approaches for the automatic generation of controllers, a common one being to combine basic behavioral blocks, such as “go forward”, “turn left”, “follow the light”, etc., into a more sophisticated controller, either manually or through powerful meta-heuristic optimization techniques [1] [2].

This project aims to enlarge the existing library of basic behavioral blocks with a cooperative behavior introducing communication between the robots, similar to [3]. The implemented behavior shall then be verified in both Webots, an open-source high-fidelity simulator, and real robot experiments. Leveraging an existing arbitrator generator, combinations of basic behaviors can then be used to solve a more challenging task in simulation or both in simulation and reality. The resulting controller shall then be compared to a controller not using the cooperative behavior. The expected effort and outcome of this work will be adjusted based on the type of the project.

Recommended type of project: Semester project / Master project

Work breakdown: 20% theory, 50% software, 30% experimentation

Prerequisites: Broad interest in robotics, control algorithms and machine learning, good command of C/C++

Keywords: Robot control, automatic control design, simulation and real robots experiments

Contact: Cyrill Baumann

References:
[1] Francesca G., Brambilla M., Brutschy A., Trianni V., Birattari M., “AutoMoDe: A novel approach to the automatic design of control software for robot swarms.” Swarm Intelligence 8 (2014), 89-112.
[2] Duarte M., Sancho Moura O., Anders Lyhne C., “Evolution of Hybrid Robotic Controllers for Complex Tasks”, Journal of Intelligent & Robotic Systems (2015) 78: 463.
[3] Hasselmann K.,  Robert F., Birattari, Mauro. (2018), “Automatic Design of Communication-Based Behaviors for Robot Swarms.” 11th International Conference, ANTS 2018, 16-29.

Automatic Design of Behavioral Arbitrators for Khepera IV Robots: A Comparison between Finite State Machines and Artificial Neural Networks

With the increased deployment of mobile robotics in a multitude of real life scenarios, robots must face more and more complex situations. However, every robotic platform needs to be carefully programmed to fulfill its task, which, today, is often still done manually. Such programming operation represents often an intense time investment and requests high expertise to eventually achieve the targeted performances in the real deployment, even more even more so if the system consists of multiple robots.

There are multiple approaches for the automatic generation of controllers, a common one being to combine basic behavioral blocks, such as “go forward”, “turn left”, “follow the light”, etc., into a more sophisticated controller, either manually or through powerful meta-heuristic optimization techniques [1] [2].

Starting from the existing Finite State Machine (FSM) arbitrator for basic behaviors of the Khepera IV robot, this project aims to implement an Artificial Neural Network (ANN) arbitrator for the same task leveraging the same (existing) basic behaviors. The implemented arbitrator shall then be tested against the existing one in both Webots, an open-source high-fidelity simulator, and real robot experiments. Furthermore, a controller entirely based on an ANN (i.e., possibly multi-layer, and not using basic behaviors) shall be implemented and compared to the two architectures using the arbitrators. The expected effort and outcome of this work will be adjusted based on the type of the project.

Recommended type of project: Master project / Semester project

Work breakdown: 30% theory, 40% software, 30% experimentation

Prerequisites: Broad interest in robotics, control algorithms and machine learning, knowledge of ANN as well as TensorFlow will be an asset

Keywords: Robot control, automatic control design, simulation and real robots experiments

Contact: Cyrill Baumann

References:
[1] Francesca G., Brambilla M., Brutschy A., Trianni V., and Birattari M., “AutoMoDe: A novel approach to the automatic design of control software for robot swarms.” Swarm Intelligence 8 (2014), 89-112.
[2] Duarte M., Sancho Moura O., and Anders Lyhne C., “Evolution of Hybrid Robotic Controllers for Complex Tasks” Journal of Intelligent & Robotic Systems (2015) 78: 463

Validation of a High-Fidelity Simulator for Odor Sensing with a Quadrotor

Odor distribution mapping is a useful technique for real world applications such as environmental monitoring. Implementations so far have focused on 2D mapping, but a 3D approach would be very valuable, given the tridimensional nature of the plume. Researchers at DISAL are focusing on exploiting quadrotors in odor mapping algorithms to obtain a 3D map of the environment.

One of the main challenges of this work is that the quadrotor perturbs the plume that has to be detected with its propellers [1][2]. The simulation used for this project can replicate a plume, and a Computational Fluid Dynamics simulation of the propellers is being developed. However, the effect of the propellers is simulated for a hovering quadrotor and there is no comparison with the real-world perturbation. The objective of this project is to improve the existing simulation by accounting for the effect of the propellers for a moving quadrotor. Moreover, the simulation will be validated by carrying out real world experiments and compare the simulated gas distribution with the real world one.

The student will start by understanding the impact of the quadrotor on the plume and getting familiar with the pre-existing work. Afterwards, the effect of the propellers for a moving quadrotor will be modeled and integrated in the simulation, leveraging the robotic simulator Webots. The fidelity of the simulation will be tested by carrying out real experiments with the quadrotor and comparing these data with the simulation.

Recommended type of project: semester project

Work breakdown: 30% theory, 30% experiments, 40% simulation

Prerequisites: Broad interest in robotics, good knowledge of C/C++, good knowledge of fluid dynamics will be an asset

Keywords: odor source mapping, quadrotors

Contact: Chiara Ercolani 

References:
[1] Lopez, Omar D., Jaime A. Escobar, and Andres M. Pérez. “Computational Study of the Wake of a Quadcopter Propeller in Hover.” 23rd AIAA Computational Fluid Dynamics Conference. 2017.
[2] Yoon, Steven, Henry C. Lee, and Thomas H. Pulliam. “Computational Analysis of Multi-Rotor Flows.” 54th AIAA Aerospace Sciences Meeting. 2016

Dynamic Visualization of Real Data for 3D Odor Plume Mapping

Odor distribution mapping is a useful technique for real world applications such as environmental monitoring. Implementations so far have focused on 2D mapping, but a 3D approach would be very valuable, given the tridimensional nature of the plume. Researchers at DISAL are focusing on exploiting quadrotors in odor mapping algorithms to obtain a 3D map of the environment.

The experimental setup is the current focus of the research. The position of the quadrotor can be logged and the placement of the odor sensor on the vehicle is being studied. However, no tool allows for a clear visualization of the evolution of the odor map with time. This feature would enable an efficient comparison of the effect of some environmental variables such as source features and airflow effects. In addition, we would like to employ a heterogeneous system for odor mapping, thus the data visualization strategy should be compatible with other platforms (such as a mobile robot and a static sensor).

The objective of this project is to implement an effective data visualization strategy to create a dynamic odor map. The student will start by getting familiar with the odor mapping problem and defining an appropriate strategy for data visualization. Afterwards, the student will implement a dynamic map with data sets gathered during experiments. Finally, the implementation will shift towards a real-time approach, with the objective of generating a map on-line while experiments are taking place.

Recommended type of project: semester project

Work breakdown: 10% theory, 20% experiments, 70% software implementation

Prerequisites: Broad interest in robotics, very good programming skills (Python/JavaScript/HTML), knowledge of ROS would be an asset

Keywords: odor source mapping, quadrotors, data visualization

Contact: Chiara Ercolani 

References:
[1] Ishida, Hiroshi. “Blimp robot for three‐dimensional gas distribution mapping in indoor environment.” AIP Conference Proceedings. Vol. 1137. No. 1. AIP, 2009.
[2] P. P. Neumann, S. Asadi, A. J. Lilienthal, M. Bartholmai and J. H. Schiller, “Autonomous Gas-Sensitive Microdrone: Wind Vector Estimation and Gas Distribution Mapping,” in IEEE Robotics & Automation Magazine, vol. 19, no. 1, pp. 50-61, March 2012.

Source Term Estimation Algorithms for Odor Source Localization in Environments with Obstacles

       

Odor source localization is currently a developing research domain which has a large number of applications in real environments. There are currently three main classes of algorithms designed for this purpose: bio-inspired algorithms, probabilistic algorithms, and formation-based algorithms.
A promising type of probabilistic algorithms in this domain is called Source Term Estimation (STE) [1]. These algorithms use the measured concentrations as well as their locations, to calculate the probability distribution of the source position using Bayesian estimation. This algorithm has already been implemented and validated at DISAL using a single mobile robot in laminar air flow conditions [2].
The goal of this project is to replace the model that is defined based on the laminar air flow assumption with a more complex model or an online Computational Fluid Dynamics (CFD) simulation (e.g., [3]) to carry out the estimation in a more accurate way. Therefore, after reviewing the literature, the selected solution should be implemented in the open-source, high-fidelity simulator robotic simulator Webots. This solution should be integrated with the existing STE-based probabilistic algorithm. At this stage, it might be also necessary to work with a CFD simulator (e.g., OpenFOAM). Once the algorithm is validated in simulation, it can be implemented on a Khepera IV robot and evaluated in a wind tunnel.
The effort and expected outcomes of this work will be adjusted based on the type of project.

Recommended type of project: semester/master project

Work breakdown: 20% theory, 50% software, 30% experimentation

Prerequisites: Broad interest in robotics, solid base on probability theory, familiarity with probabilistic algorithms, good command of C/C++.

Keywords: odor source localization, olfactory robotics, probabilistic algorithms.

Contact:  Faezeh Rahbar

References:
[1] M. Hutchinson, H. Oh, and W. H. Chen, “A review of source term estimation methods for atmospheric dispersion events using static or mobile sensors,” Inf. Fusion, vol. 36, pp. 130–148, 2017.
[2] F. Rahbar, A. Marjovi, A. Martinoli, “An algorithm for 3D Odor Source Localization through Source Term Estimation,” IEEE Int. Conf. on Robotics and Automation, 2019. To appear.
[3] Asenov, Martin, et al. “Active localization of gas leaks using fluid simulation.” IEEE Robotics and Automation Letters 4.2 (2019): 1776-1783.

Design and evaluation of a Source Term Estimation Algorithm for Odor Source Localization using a Heterogeneous System

Odor source localization is currently a developing research domain which has a large number of applications in real environments. There are currently three main classes of algorithms designed for this purpose: bio-inspired algorithms, probabilistic algorithms, and formation-based algorithms.
A promising type of probabilistic algorithms in this domain is called Source Term Estimation (STE) [1]. These algorithms use the measured concentrations and their locations, to calculate the probability distribution of the source position using Bayesian estimation. This algorithm has already been implemented and validated for odor source localization in 2D and 3D using a single node [2] and we are currently developing it for a homogeneous distributed system consisting of mobile robots. The goal of this project is to extend the setup and the algorithm for a heterogeneous distributed system where multiple mobile robots and sensor nodes need to share their observations with each other and coordinate the motions of the self-locomoted nodes.
This project will start with simulations carried out in the open-source, high-fidelity robotic simulator Webots, in which the distributed system needs to be designed and programmed with the existing algorithm. Then it can be implemented for Khepera IV robots and a set of sensor nodes developed at DISAL and evaluated in a dedicated wind channel. The effort and expected outcomes of this work will be adjusted based on the type of project.

Recommended type of project: semester/master project

Work breakdown: 40% theory, 40% software, 20% experimentation

Prerequisites: Broad interest in robotics, solid base on probability theory, familiarity with probabilistic algorithms, good command of C/C++.

Keywords: odor source localization, olfactory robotics, probabilistic algorithms.

Contact:  Faezeh Rahbar

References:

[1] M. Hutchinson, H. Oh, and W. H. Chen, “A review of source term estimation methods for atmospheric dispersion events using static or mobile sensors,” Inf. Fusion, vol. 36, pp. 130–148, 2017.

[2] F. Rahbar, A. Marjovi, A. Martinoli, “An algorithm for 3D Odor Source Localization through Source Term Estimation,” submitted for publication in IEEE Int. Conf. on Robotics and Automation, 2019.

Automatic Pre-Launch Self-Test Framework for AUVs

Deploying multiple robots in harsh outdoor environments is a technical as well as a logistical challenge. Therefore, we intend to automate various aspects of the process of launching AUVs. The most critical of them is initial compass calibration. The other aspects include the so-called “power on self-test”, where the AUVs should self-analyze operation of various subsystems, attempt to reboot individual subsystems if needed and report errors if necessary.

The project will begin with implementation of initial compass calibration. Calibration parameters should be automatically computed after a series of “figure of 8” movements and applied to the attitude estimation module.

Subsequent work will involve implementing automatic testing and diagnosis of errors in the AUV subsystems. The AUV comprises of a main unit (called “power board”) which is responsible for managing other subsystems such as navigation, sensing payload, acoustic modem etc. Before the start of the mission, the power board should ensure that each of these subsystems are functioning as expected, and take corrective measures (usually reboot) if necessary. Access to information about the functioning of these devices is available already.

A good command of C/C++ and prior experience in programming is required. Experience with hardware/embedded systems is a bonus.

Recommended type of project: semester project

Work breakdown: 20% theory, 70% software, 10% experimentation

Prerequisites: Good command of C/C++, experience with embedded systems.

Keywords: distributed underwater robots, mobile robots, field robotics

Contact: Anwar Quraishi

Cooperative Localization for a Group of AUVs 

We have developed Autonomous Underwater Vehicles (AUVs) for gathering environmental data in water bodies. The AUVs are equipped with a suite of sensors that measure various quantities (temperature, concentration of various substances, turbidity, etc.). The robots are intended to be used in a coordinated formation, gathering measurements in 3D at a high spatial resolution.

AUV localization is difficult because GNSS does not work underwater. As such, the AUVs have to periodically surface to receive a GNSS update, and perform dead reckoning between such updates. However, since we intend to operate them as a group, an AUV can exploit GNSS updates received by other AUVs. Hence, multiple AUVs could cooperatively navigate, without frequently surfacing for GNSS reception.

We are also developing an acoustic communication system, which can be used to exchange data as well as estimate range and bearing to the transmitting AUV based on time-of-flight. Thus, new information (such as GNSS update) received by an AUV can be propagated to other vehicles in the group, which can then refine their own position estimates [1]. The challenge then lies in correctly incorporating this information into a vehicles position estimate, and correctly computing the new uncertainty in its position.

The goal of this project will be to address one or more of the challenges in cooperative navigation. The methods developed will be tested in simulation as well as in the lake with real robots.

Recommended type of project: master project

Work breakdown: 40% theory, 40% software, 20% experimentation

Prerequisites: Background in mobile robotics, prior experience with (or knowledge of) Kalman and particle filters, good command of C/C++, experience with embedded systems.

Keywords: distributed underwater robots, state estimation, sensor fusion, cooperative localization

Contact: Anwar Quraishi