Student Projects

Master’s Thesis/Semester Projects Proposals (Fall 2025)

We have updated the thesis/semester project offers. Please e-mail the listed contacts directly.  

Please tell us your motivation to the specific project, your skillsets and past experiences + something that shows that (CV, website, photos, videos, github, etc), and your transcript.

Students from all research backgrounds are welcome – for example, mechanical engineering, bioengineering, computer science, materials etc.

You can also find this webpage here in another format 

Updated August 2025

[closed] Self-healing material as robotic joints

This project focuses on the development of a reconfigurable, soft robotic joint using self-healing materials. The goal is to move beyond conventional applications of these materials in soft robotics (such as grippers) to create a dynamic joint that can be repeatedly “broken” and “re-formed” in new configurations. The primary challenge involves identifying or fabricating a material that possesses robust and repeatable self-healing properties. The project will involve designing a joint that can be controllably “healed” using an external stimulus, like heat, and then deploying this mechanism onto a fabricated prototype.

The project will start with material selection and small-scale testing to verify repeatable healing cycles. The final prototype will demonstrate the joint’s ability to be reconfigured multiple times, with its mechanical properties characterized after each healing event. This research aims create robots which are more versatile and resilient to the environment that they operate in.

10% material identification, 60% prototyping, 30% testing.

Requirements – strong background in CAD (fusion 360, solidworks, openSCAD), electronics, independent experience in mechanical design and fabrication, and interest in bioinspired robots. Self-driven.

Contact – Nana Obayashi, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

[closed] Bio-inspired underwater loco-manipulation

Bio-inspiration is crucial for designing underwater robots that operate in complex natural environments and interact with marine life. Traditional Autonomous Underwater Vehicles (AUVs) often create noise and wake that disturb ecosystems. This has led to the development of robots inspired by various marine creatures like fish, octopuses, and jellyfish, primarily for monitoring and locomotion. However, to significantly increase their utility, these robots need manipulation capabilities to interact with their surroundings and anchor themselves. The goal of this project is to add manipulation features to a lightweight, scalable, and soft robotic fish that uses a single motor for tail propulsion. This involves the design, prototyping, and experimental validation of both the manipulation mechanism and the entire integrated system.
70% prototyping, 30% testing.

Requirements – strong background in CAD (fusion 360, solidworks, openSCAD), electronics, independent experience in mechanical design and fabrication, and interest in bioinspired robots. Self-driven.

Contact – Nana Obayashi, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Semester Project) Enhanced Robot Navigation Through Multi-Sensor Integration

Description – Service robots operating in dynamic environments face critical challenges when encountering large obstacles, moving objects, and complex spatial configurations that standard sensors cannot adequately detect. This project focuses on augmenting a Keenon service robot with advanced sensor arrays to create a robust, adaptive navigation system that ensures safe and efficient operation in human-occupied spaces. The enhanced system will integrate multiple sensing modalities including LiDAR, depth cameras, and ultrasonic sensors to build a comprehensive environmental perception framework. Students will develop real-time obstacle detection algorithms, dynamic path planning systems, and sensor fusion techniques that allow the robot to adapt its navigation strategy based on changing environmental conditions. The goal is to significantly improve the robot’s ability to detect and respond to both static and dynamic obstacles while maintaining smooth, efficient movement patterns.This project involves literature review and system analysis (20%), sensor integration and algorithm development (40%), and experimental validation and performance testing (40%).

Requirements – The project will teach you skills in Python, Computer Vision, System Integration and APIs.
Contact – Max Polzin, [email protected]

To apply, please include your CV and academic transcripts.

(Master Thesis) Force estimation and control of a spring loaded robotic arm

Description — The ability to estimate and control the force is critical for robotic arms with applications where environmental interaction is critical. While most conventional robotic arms achieve this via torque estimation at the joint level via torque sensors directly attached to the joints, we developed a tendon driven 7-dof robotic arm where physical springs are embedded which provides an inherent safety level to the robot system.

In this project, we work towards force estimation and control of this novel robotic arm, by taking into account the full robot model (down to the details of tendon mechanism and actuator models/control strategies). The goal is to deploy the models and controllers on a real robot to perform contact rich interaction tasks.

This project is a collaboration with Embodied AI, the spinoff from CREATE Lab.

Requirements — Strong control background, programming experience (python), real robot experience (actuation, sensing, control, etc)

Contact — Kai Junge, k[email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Robotic vacuum cleaning using a continuum robot arm

Description — Assistive in-house care is one of the grand application areas for robotics. This application poses a unique challenge where the environment remains largely unstructured, leading to specialized automation tools to be ineffective while highly generalized solutions are decades behind.

Consider the task of vacuum cleaning. Although mobile vacuum robots have become prevalent in households, they struggle in corners and are only limited to the floor.

In this project, we explore the task of combining continuum soft robots (https://www.nature.com/articles/s44182-023-00004-7) with a vacuum cleaner hose. By actuating a hose with a soft arm, surfaces in 3D can be cleaned by leveraging the manipulability of the soft arm and the ability to make safe interactions with the environment.

The work packages include robotic prototyping of this system, feasibility testing and analysis, and experimentation of different controllers. This project is a collaboration with Embodied AI, the spinoff from CREATE Lab.

Requirements — practical robotic skills (3D printing, CAD), programming experience (python), motivated for field testing, strong communication skills

Contact — Kai Junge, k[email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Open Parametric Hand design for manufacturing and accessibility

Description — Robotic manipulation has seen great strides  in actuation, sensing, and control. However, outside of biomimetic hands there has been little focus on developing the underlying mechanics and physical behaviours of the hand which play a critical role in manipulation performance, particularly for efficient control and exploitation without a human brain or muscle performance. The goal of the OPH (https://github.com/kg398/100_fingers)  project is to explore the link between robot hand morphology and manipulation performance, something that can only be accurately captured in the real world. This is where an accessible, highly customisable, and rapidly manufacturable hand is essential. 

The OPH has seen success so far, but particularly the actuation system is in need of upgrading for accessibility and potentially for scaling up manufacturing. The focus of this project is redeveloping the OPH and surrounding systems for performing large-scale experiments, and if the project goes well, performing some of those large-scale experiments and translating the work for reproduction. 70% design and prototyping, 20% testing, 10% outreach.

Requirements — strong background in CAD (fusion 360/solidworks, openSCAD (bonus)), interest in dexterous robotic manipulation and mechanical design

Contact — Kieran Gilday, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Open Parametric Hand field testing and actuation development

Description — Robotic manipulation is poised to take over the next generation of automation tasks requiring manipulations more complex than pick and place and robustness to significant environmental uncertainties. The OPH (https://github.com/kg398/100_fingers) is one potentially capable manipulation platform which focuses on high degree of customizability to tailor dexterous behaviours to a given application rather than making one general purpose hand not optimised to any single task. 

This project focuses on finding and solving real world manipulation problems such as robotic harvesting, industrial food handling, or flexible warehouse/factory tasks. This involves collaborations with industry and academic partners, bespoke system development and problem solving, field tests, and liaising with end users. One major work package will be developing customization packages for the actuation system, where prediction and optimisation of tendon actuation strategies is an unsolved problem. 30% robotic system development, 30% actuation/modeling, 40% case studies/outreach.

Requirements — practical robotic skills (3D printing, CAD), programming experience (python, matlab), motivated for field testing, strong communication skills

Contact — Kieran Gilday, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Open Parametric Hand design optimization

Description — Robotic manipulation has seen great strides  in actuation, sensing, and control. However, outside of biomimetic hands there has been little focus on developing the underlying mechanics and physical behaviours of the hand which play a critical role in manipulation performance, particularly for efficient control and exploitation without a human brain or muscle performance. The goal of the OPH (https://github.com/kg398/100_fingers)  project is to explore the link between robot hand morphology and manipulation performance, something that can only be accurately captured in the real world.

With up to 150 parameters to optimise, there is a huge possible design space of our robotic hand. The investigations we have performed so far only look at small parcels of this, but already have revealed very interesting results on finger design such as optimising for behavioural ranges converges on human-like design. The project has three major parts, the first is refining our models or developing new ones for extracting virtual hand performance metrics. The second is investigating potential optimization methods, especially towards diversity-based optimization. The final is investigating optimizations for real world problems. 30% virtual hand modeling, 30% optimization methods, 40% case studies and robotic hand development.

Requirements — strong programming background (python, matlab), knowledge of optimization algorithms is a plus, practical robotic experience is a plus

Contact — Kieran Gilday, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Open Parametric Hand large-scale tactile sensing (EIT)

Description — Large-scale tactile sensing plays a huge role in dexterous human and robotic manipulation, providing feedback for adapting to real world unknowns but also an essential mechanism for how we experience and learn from the world around us. Compared to the 17,000 mechanoreceptors in human hands, robotic hands rarely contain more than 100. Electro-impedance tomography is a potential solution to scale up tactile sensing capabilities using medical imaging techniques to generate impedance maps of a conductive skin sensitive to tactile forces. 

This project will be about developing the EIT technology for small form factors, pcb development, testing, interpreting and leveraging tactile images to solve manipulation problems. Using our existing OPH (https://github.com/kg398/100_fingers) as a basis, we can explore the efficacy of EIT for gaining tactile information, with final deliverables in demonstrations exploiting tactile information to improve manipulation performance and/or using the information to improve learning systems. 30% electronics development, 30% robotic system integration, 40% data processing and control system development.

Requirements — electronics experience (pcb development ideal), knowledge of microcontrollers, solid programming skills (python, matlab), knowledge of control systems, vision systems, and machine learning is a plus

Contact — Kieran Gilday, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Open Parametric Hand small-scale tactile sensing and force feedback (haptics)

Description — Tactile sensing and force feedback to users is a strong requirement in many robotic hand applications, enabling improved performance and robustness to uncertainties. In particular for prosthetic applications haptic feedback can greatly improve user experience. The OPH (https://github.com/kg398/100_fingers) offers a strong foundation for accessible and functional prosthetic hands, however, integration of tactile sensing and force feedback is an unsolved problem. 

This project is about developing simple but robust force feedback systems for prosthetic hand users. There is scope for exploration of different sensing and haptic technologies from purely electrical (e.g. with custom developed soft piezoresistive skins and vibration motor haptic patches), purely mechanical (e.g. transmitting tactile pressure via fluid channels to inflatable skin patches), or a mixture. Rapid prototyping, experiment design, and user testing will be significant parts of the project. 60% haptic system development, 30% experimentation and user testing, 10% robotic hand development.

Requirements — electronics and mechanical prototyping skills (mcu, pcb, cad, 3d printing are a plus), interest in robotic/prosthetic hands or the science of human-robot interactions 

Contact — Kieran Gilday, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Open Parametric Hand learning and control with foundation models

Description — The emergence of foundation models has deep implications for dexterous robotic manipulation. In particular, exploitation of diverse hand-environment interactions can provide a step up in more naturalistic and autonomous manipulation performance. Our technology, the OPH (https://github.com/kg398/100_fingers), is an ideal platform to apply recent high performing tools such as diffusion models and visual language action models, where current control methods are unable to efficiently learn for diverse tasks with sequential interactions. 

This project is about researching and applying advances in foundational models for robotic manipulation to our unique platform to achieve generalisable and adaptive performance. This involves developing systems for multimodal data capture, training models, and experiment design. 50% data systems, capture and experiment design, 40% model training and exploration, 10% robotic hand development.

Requirements — strong programming experience (e.g. python, matlab), machine learning experience, knowledge of ml-based manipulation controllers is a plus (e.g. reinforcement learning, imitation learning, diffusion models, vlas), practical robot development skills is a plus

Contact — Kieran Gilday, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis / Semester Project) Stretchable Optical Fiber Sensor for Sensing Glove

Description – The stretchable optical fiber sensor is a thin optical fiber capable of distributed sensing of bending and stretching. This sensor uses light as a signal and colored coating as a spectral filter, making it possible to multiplex the signals. As a result, it is possible to measure not only the stimuli but also the location of the stimuli. It offers several advantages such as speed and compactness, addressing the challenging problem of distributed sensing of multimodal stimuli in the field of robotics. The goal of this project is to optimize the design of this sensor and embed it into a glove in order to measure finger joint angles for wearable sensing applications.

This project involves a literature survey (20%), fabrication of sensors and the sensing glove (40%), and conducting experiments (40%).

Requirements – The project requires fabrication skills, 3D CAD experience, and basic data analysis skills.

Contact – Haewon Jeong, [email protected]
To apply, please include your CV and academic transcripts.

[CLOSED] (Master Thesis/Semester) Identification of cable duct geometric parameters in cluttered and constrained environments

Description — As the network of underground cables continues to grow, so do the difficulties associated with maintenance. Often cable ducts contain multiple sizes and types of power/communication lines (copper cables can regularly be >10cm in diameter), leading to challenges if a specific cable has to be replaced and damage to others needs to be avoided. One issue is being able to identify a cable of interest within a bundle, or identifying upcoming geometry changes in the duct itself.

This project will be focused on these types of tasks, using either vision or other sensing modalities to identify which cable should be manipulated or cut, for the back-end control to function accordingly. Similarly, obtaining basic information about the environment (like identifying cracks or changes in pipe diameter) are also key points of interest. 

This project involves 10% Literature study, 50% Software/Firmware Implementation, 40% Experimentation and Evaluation (rough percentages)

Requirements — strong programming skills (Python/C++/MATLAB), experience with computer vision/other sensing modalities and electronics skills.

Contact — Kyle Walker ([email protected])

To apply please include a copy of your CV and transcript along with any other relevant information.  

[CLOSED] (Master Thesis) Design of an in-pipe robot for navigation and intervention in underground cable ducts

Description — As the network of underground cables continues to grow, so do the difficulties associated with maintenance. Often cable ducts contain multiple sizes and types of power/communication lines (copper cables can regularly be >10cm in diameter), leading to challenges if a specific cable has to be replaced and damage to others needs to be avoided.

This project will focus on the mechanical development of a robotic platform which can successfully navigate these environments, where duct size often varies and can be occupied with cables up to 50% (cross section). The platform should be able to operate in these various conditions, undergoing some mechanical deformation or transformation to achieve this (i.e. morph in some way). This will be a very hands-on project and be heavily design focused.

This project involves 10% Literature Search, 50% Design and prototyping, 40% Experimentation (rough percentages)

Requirements — 3D CAD Experience, strong mechanical/mechatronic background, solid electronics knowledge and are interested in novel mechanism design. Programming skills are also beneficial but not a hard requirement.

Contact — Kyle Walker ([email protected])

To apply please include a copy of your CV and transcript along with any other relevant information. 

(Master Thesis) Autonomous manipulation and cutting for semi-autonomous extraction of underground cabling

Description — As the network of underground cables continues to grow, so do the difficulties associated with maintenance. Often cable ducts contain multiple sizes and types of power/communication lines (copper cables can regularly be >10cm in diameter), leading to challenges if a specific cable has to be replaced and damage to others needs to be avoided.

This project will focus on continuing the development of a method to cut these cables that is both lightweight and compact, with a focus on integration to a small-scale robot. Similarly, efforts to automate the cable manipulation, grasping and finally cutting process will be key parts of the work.. 

This project involves 10% Literature Search, 50% Design and prototyping, 40% Experimentation (rough percentages)

Requirements — 3D CAD Experience, strong mechanical/mechatronic background, solid electronics knowledge and programming skills. Experience in manipulation of objects is a plus.

Contact — Kyle Walker ([email protected])

To apply please include a copy of your CV and transcript along with any other relevant information. 

[CLOSED] (Master Thesis/Semester) Data-Enabled Model Predictive Control for Compliant Manipulators

Description — Dynamic control of compliant manipulators is still an open challenge, with many of the current issues arising from the high degree of nonlinearities in system dynamics. Whilst the use of Model Predictive Control (MPC) has been increasing in conjunction with computing capabilities, the uptake in the space of compliant manipulators has been constrained by the inability to efficiently and effectively solve the optimisation problem online using methods suitable for real-world deployment, coupled with the uncertainty attached to estimating the dynamic parameters in the system model.

This project will investigate these factors and aim to develop a data-driven, reliable and robust MPC framework for compliant manipulators, from parameter identification through to real-world implementation. The overarching aim is to be able to modify the system dynamics online when accounting for factors such as operating environment (water/air) and manipulator stiffness changes. 

This project involves 20% Literature Search, 40% Simulation, 40% Experimentation (rough percentages)

Requirements — strong programming skills (MATLAB/Python/C++), knowledge of nonlinear system dynamics, experience with model predictive control (more widely model-based control) and optimisation-based control strategies/nonlinear solvers in general. Prior experience and confidence working with real experimental platforms is a bonus but not a must.  

Contact — Kyle Walker ([email protected])

To apply please include a copy of your CV and transcript along with any other relevant information. 

[CLOSED] (Master Thesis) 3D Variable Stiffness Manipulator Design

Description — In nature, when organisms change environments or perform different tasks they often have the ability to control their internal parameters (for example joint/body stiffness) to operate effectively. This project will explore how this phenomena can be implemented in a robotic system to design a manipulator which can either be extremely flexible or extremely rigid (and varied in-between) depending on the task at hand or the environment it is operating in. 

Specifically, the project will focus on the 3D case and include implementation of a control methodology for identifying when a stiffness change is required, subsequently applying the control signal to instigate this. The target scenario for this is transitioning from air to water, where a stiffness change can improve the controllability and efficiency of the system.

This project involves 10% Literature Search, 40% Mechanical Design and prototyping, 20% Control and electronics implementation, 30% real-world experimentation (rough percentages)

Requirements — 3D CAD Experience, strong background in mechanism design, are confident working with electronics and have solid programming skills. Experience with control algorithms (particularly for compliant/nonlinear systems) is a bonus but not a must. 

Contact — Kyle Walker ([email protected])

To apply please include a copy of your CV and transcript along with any other relevant information. 

[CLOSED] (Summer project) Wearable devices to enhance sensations for robotic hands

Description — Wearable devices and smart gloves are revolutionizing human-machine interaction and robotic control. This project aims to develop wearable devices to enhance the sensing capability of robotic hands. Our goal is to gain in-depth understanding of human behaviour and human-machine-environment interfaces. The focus will be on innovative sensor integration (force, bending, temperature, proprioception, etc.), comfortable ergonomic design (wearables like gloves for both human and robot), and real-time data transmission and interaction with robotic hands. 20% Sensor/actuator design and fabrication, 30% Wearable device integration and development, and 50% Software and data analysis.

Requirements — programming skills (python, MATLAB), knowledge of control systems, computer vision systems, and machine learning.

Contact — Benhui Dai ([email protected])

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Summer project/Master thesis/Semester project) Tactile e-skin sensing with EIT

Description — Tactile sensing plays an esstential role in dexterous human and robotic manipulation, providing feedback for adapting to the real world. Electrical Impedance Tomography (EIT) is an imaging technique that measures the conductivity, permittivity, and impedance of an object. It works by attaching electrodes to the surface of the object and then using the electrodes to either inject current or measure the resulting voltages. Interpolating the raw signals then results in an image of the object’s internal conductivity. 

This project will be about developing the EIT technology for e-skin, pcb development, testing, interpreting and leveraging tactile images to solve manipulation problems. 20% electronics development, 30% robotic system integration, 50% data processing and control system development.

Requirements — electronics experience (pcb development ideal), knowledge of microcontrollers, solid programming skills (python, matlab), knowledge of control systems, and machine learning is a plus

Contact — Benhui Dai, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master thesis/Semester project) Soft robotic tongue for speech

Description — The tongue is crucial for articulation in speech. It moves in various ways—up, down, forward, backward, can curl, make contact with different parts of the mouth (like the alveolar ridge, hard palate, teeth), and its shape changes to modulate airflow, creating different phonemes. Our goal is to design a soft robotic tongue for speech that involves replicating the human tongue’s flexibility, precision, and adaptability to produce articulate sounds. 30% Tongue design and fabrication, 40% Actuators and sensors integration, and 30% Software and data analysis.

Requirements — 3D CAD experience (fusion, SolidWorks), fabrication skills for soft robotics, electronics skills (motor & pump control), programming skills (python, MATLAB)

Contact — Benhui Dai ([email protected])

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Summer project/Master thesis/Semester project) Chewing robot with sensory-motor control

Summary: The process of chewing food is achieved through the up-and-down movement of the jaw and the shearing movement of the teeth. We have developed a 6 DOF chewing robotic platform, including the mouth, jaw, and tooth. The goal is to simulate the mechanical movements of humans chewing by robots and monitor this process by force sensors and cameras. What’s more, artificial saliva and a soft robotic tongue will be embedded into the system to provide a biomimetic oral environment for better food-robot interaction.
Workload: 50% control, 30% data processing and analysis, 20% hardware design and fabrication
Requirements — programming skills (python, MATLAB), knowledge of control systems (PID control), and better machine learning.
Contact — Benhui Dai ([email protected])
To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master thesis/Semester project) Robotic Stomach Design

Description — Digestive malfunctions are linked to a variety of health conditions, including metabolic disorders, brain lesions, and gastrointestinal diseases. Current research in digestive systems lacks a reliable, controlled method to replicate the physical and mechanical processes of the human stomach. This thesis aims to design and prototype a robotic stomach that mimics the motions and environmental conditions of human digestion. By reviewing existing artificial stomach designs and integrating advancements in soft robotics, the research will create a minimalistic model suitable for replicating the mechanical digestive motions. 30% literature review, 50% prototype design and actuation,20% tests.

Requirements — 3D CAD experience (fusion, SolidWorks), fabrication skills for soft robotics, electronics skills (motor & pump control), programming skills (python, MATLAB)

Contact — Benhui Dai ([email protected]), Arnaud Klipfel ([email protected]).

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

[CLOSED] (Master Thesis/Semester project) Design and Development of a Robotic Upper Limb 

Description — The human shoulder is composed of multiple bones, including the humerus, scapula, and clavicle, and consists of various joints such as the glenohumeral joint, scapulothoracic joint, and acromioclavicular joint. These anatomical components collectively enable the complex and versatile movements of the shoulder. Inspired by this structure, the objective of this research is to design and develop a robotic upper limb capable of replicating similar shoulder motions. The system will be constructed using flexible fibers and actuated via a tendon-driven mechanism, with the goal of emulating the biomechanical characteristics of the human shoulder.

This project involves literature study (10%), design•fabrication skills (40%), actuating (30%), and conducting experiments (20%).

Requirements —  strong background in CAD, and programming skills (e.g. Python, Matlab)

Contact — Sudong Lee, [email protected]

To apply, please include a copy of your CV and transcripts.

(Master Thesis/Semester project) Development of an Adhesive Robotic Hand Utilizing Electrostatic

Description — There have been studies that utilize electrostatic principles to generate adhesion forces for gripping applications. This mechanism is particularly promising for delicate and adaptable grasping. The aim of this research is to apply such an electrostatic adhesion mechanism to a robotic hand, enabling effective and controllable object manipulation. The objective is to develop a high-voltage (low-power) generation circuit and an electrode design for electrostatic adhesion hand.

This project comprises circuit and electrode design (50%), system design for the hand (30%), and experimental validation (20%).

Requirements —  strong background in electronics (PCB development is a plus), and fabrication•design skills

Contact — Sudong Lee, [email protected]

To apply, please include a copy of your CV and transcripts.

Master Thesis Proposal: Humanoid Soft Bimanual Arm Robot with Trimmed Helicoid Structures  

Note: This topic is part of a broader initiative to advance human-robot collaboration and soft robotics. It is linked to projects on adaptive control systems, biomimetic materials, and safe human-robot interaction.  

Description — Traditional rigid robotic arms, while precise, face limitations in safe interaction with humans and adaptability in dynamic environments. Applications in healthcare, assistive technologies, and collaborative workspaces demand robots that are inherently compliant, lightweight, and capable of delicate, human-like manipulation. This project will focus on designing and prototyping a humanoid soft bimanual arm robot using trimmed helicoid structures, a novel geometric approach enabling lightweight construction, inherent compliance, and dynamic motion. The arms will integrate soft actuators and sensors to safely interact with humans while performing tasks requiring precision and adaptability (e.g., object manipulation, collaborative assembly).  

Key objectives:  

1. Design and Fabrication: Optimize the trimmed helicoid architecture for dual-arm coordination, balancing structural integrity, flexibility, and weight.  
2. Actuation and Sensing: Develop soft pneumatic/muscle-inspired actuators embedded with stretchable sensors for real-time force and motion feedback.  
3. Control System: Implement adaptive algorithms for bimanual tasks (e.g., grasping, lifting) and dynamic movements, ensuring safety during human interaction.  
4. Experimentation: Validate safety, compliance, and performance in scenarios involving human proximity and variable payloads.  

Project Breakdown:  

– 15% Literature Review: Survey soft robotics, humanoid arm design, and control strategies for bimanual systems.  

– 50% Design & Prototyping: CAD modeling of helical structures, actuator integration, and sensor embedding.  

– 35% Experimentation: Testing safety (force thresholds, collision response), dynamic motion (speed, precision), and task performance.  

Requirements —  

– Proficiency in 3D CAD (e.g., Fusion 360, SolidWorks) for complex geometric designs.  

– Strong mechatronics/mechanical engineering background (soft actuator experience preferred).  

– Knowledge of materials science (flexible composites, elastomers) and electronics (sensor integration, microcontrollers).  

– Programming skills (Python/C++ for control systems; ROS/Simulink experience a plus).  

– Creativity in problem-solving and interest in human-robot interaction.  

Contact — [Your Name] ([Your Email]), [Advisor’s Name] ([Advisor’s Email])  

To Apply: Submit a CV, academic transcript, and portfolio (e.g., CAD designs, robotics projects) demonstrating relevant skills.  

—  

Key Innovations:  

– Trimmed Helicoid Geometry: Combines helical and hyperbolic traits for tunable stiffness and compact, lightweight form.  

– Dual-Arm Synergy: Enables human-like coordination (e.g., twisting, lifting) while maintaining passive safety.  

– Applications: Assistive care, rehabilitation, and industrial co-robotics where human safety is paramount.  

Expected Challenges:  

– Balancing compliance and precision in soft actuators.  

– Synchronizing bimanual movements under sensor feedback delays.  

– Scaling fabrication for complex helicoid patterns.  

This project aims to bridge the gap between rigid industrial robots and biologically inspired systems, paving the way for safer, more versatile human-robot collaboration.

(Master Thesis) Bio-inspired Soft Snake Robot Mechanical Design and Gait Control

In recent years, a number of robots have been developed by using inspirations from animals due to their advanced and robust locomotion capabilities. One class of such animals are snakes, which are capable of moving in most of the complex environments on earth. The slim body structure of the snakes enables them to travel in narrow, constrained environments. Snake-like robots are thus promising for applications such as pipeline inspection, search and rescue in rubbles, and minimally invasive surgeries. Existing snake robots have predominantly adopted rigid modules. One classic snake robot utilized multiple motors to deform the rigid body, which demonstrated several different locomotion modes in traversing a complex terrain, climbing, and inspecting inside a pipeline. Another multi-section snake robot achieved crawling on land and swimming in water. Despite these progresses, there are some limitations for rigid snake robots. In particular, the size of a rigid snake robot is limited by the motors and complex mechanisms, which increases its difficulty in traversing a narrow environment.

In this project, we aim to design a multi-segment soft snake robot that achieves robust locomotion in contact-rich environments through gait patterns generated by section bending.

Workpackages: 60% mechanical design, 30% sensing integration, 10% gait control

Requirement: 1. Mechanical design experience, 2. Solidworks or Fusion experience, 3. FEM experience (Abaqus or Ansys), Matlab or python programming.

Contact: [email protected] (Guanran Pei)

To Apply: please send email with CV, transcripts.

(Master Thesis/Semester Project) Sensor Development and System Integration for a Slender Endoscopic Surgical Robot

Surgical robots are typically characterized by their slender profile and high degrees of freedom, enabling flexible movement within confined spaces to perform complex surgical and diagnostic tasks. However, due to their compact size, most surgical continuum robots can only accommodate actuation modules internally, lacking integrated proprioceptive sensing. As a result, these robots typically rely on teleoperation for task execution. IMU (Inertial Measurement Unit) sensors, benefiting from their modular design and miniaturization potential, can be leveraged to provide surgical continuum robots with proprioceptive sensing capabilities, thereby enhancing their autonomy.

This project aims to achieve sensor integration in slender surgical robots through the customization and miniaturization of IMU sensors, endowing surgical continuum robots with autonomous perception and decision-making capabilities.

Workpackages: 20% tendon driven slender surgical robot design, 50% IMU sensors miniaturization, 30% sensing system integration

Requirement: 1. Solid PCB design experience (important), 2. KiCAD, eagle or any other PCB EDA experience, 3. Solidworks and Fusion experience.

Contact:  [email protected] (Guanran Pei)

To Apply: please send email with CV, transcripts.

(Master Thesis) Lightweight and Compact Leg and Spine Design for a Biomimetic Mouse Robot

Quadrupedal mammals exhibit remarkable locomotion capabilities in unstructured terrains, maintaining energy-efficient stable gaits that are crucial for survival behaviors like prey pursuit or predator evasion. These evolutionary advantages have inspired significant robotics research in quadrupedal systems. Practical applications are already being demonstrated across multiple domains, including industrial inspection, emergency response, security operations, and logistics delivery.

Current quadruped robot designs primarily focus on medium to large-scale platforms. However, in extreme and confined environments, miniaturized quadrupeds demonstrate significant potential due to their superior maneuverability and compact form factor. Yet their small size imposes substantial challenges in system integration and lightweight leg-spine mechanism design.

This project aims to develop a biomimetic mouse-inspired robot with dimensions under 10cm (L×W×H). The research will focus on: compact lightweight leg-spine integrated design and high-density actuation system integration.

Workpackages: 70% leg and spine design, 20% system integration, 10% gait control

Requirement: 1. Mechanical design experience, 2. Solidworks or Fusion experience, 3. FEM experience (Abaqus or Ansys), Matlab or python programming.

Contact:  [email protected] (Guanran Pei)

To Apply: please send email with CV, transcripts.

(Master Thesis/Summer Project) Bio-hybrid robot design with collected nature part from food-waste

Description — We are exploring a new design paradigm—foraging robotics—that repurposes biological remains, particularly joints from food waste (e.g., shellfish), as high-performance mechanical components in robots. These bio-joints offer lightweight, low-friction, and mechanically efficient structures that are difficult to replicate synthetically. This project involves collecting and processing bio-joints, characterizing their mechanical properties, and integrating them into modular robotic systems. Candidates should be motivated to work at the intersection of biology and robotics, with an interest in hands-on prototyping, mechanical modeling, and bio-inspired design.

Requirements — Keen interest in working with bio-joints and food-waste-based materials, Willingness to conduct hands-on experiments with non-conventional setups, Creative and open mindset to biology/ bio-inspired design. CAD experience is required. 

Contact — If you’re interested, please contact Sareum([email protected]) and include a short motivation statement for this project, along with your CV and transcripts.

(Master Thesis/Summer Project) Modularization of an Eversion Flexible Track Robot for Constrained Space(In-pipe) Navigation

(Due to patent restrictions, the actual robot design and video are available upon request.)

Description — Navigation in unstructured terrain is challenging for robot locomotion, especially in constrained environments such as inside pipes. Flexible track robots provide a scalable, rapid-prototyping platform with adaptable navigation capabilities for such terrains. Using the eversion method, these robots roll rather than slide against their environment, which exploits the constrained space effectively. By leveraging 3D printing, the size and mechanical properties of the track can be easily customized for target pipe diameters. Currently, the robot can navigate pipes with diameters between 5 cm and 10 cm by changing the track. For each diameter, the robot can adapt to slight variations using a compressible track for passive adaptation.

The goal of this project is to perform experiments at diverse scales and modularize the driving track design to enable LEGO-like assembly based on specific needs. The project workload is approximately 50% design and prototyping, 50% testing and analysis.

Requirements — Keen interest in hands-on experiments and design. Enjoy working with robots. CAD experience is required. Basic skills in rapid prototyping (laser cutting, 3D printing) are necessary. Kinematics modeling skills or coding ability (Python/Matlab) are a plus. 

Contact — If you’re interested, please contact Sareum([email protected]) and include a short motivation statement for this project, along with your CV and transcripts.

(Master Thesis) Computationally-Designed 3D-Printed Foam Sensors for Smart Helmets

Description — This project explores the development of novel foam-based pressure sensors that can be directly embedded into 3D-printed helmet liners. Building on recent innovations in soft sensorized open-cell foams, the student will contribute to the design, fabrication, and testing of multi-layer, multi-material foam structures capable of sensing impact forces while providing superior mechanical protection.

The work involves computational design of foam geometries—such as body-centered cubic (BCC) and Kelvin structures—optimized for specific mechanical and sensing responses. Students will apply topology optimization methods, explore strategies for co-designing mechanical and electrical properties, and fabricate prototypes using state-of-the-art multi-material 3D printing techniques. The project workload is approximately 50% design and prototyping, 50% testing and analysis.

Requirements — Keen interest in hands-on experiments and design. CAD experience is required. Basic skills in rapid prototyping, especially 3D printing are necessary. Kinematics modeling skills, simulation(FEA) or coding ability (Python/Matlab) are a plus. 

Contact — If you’re interested, please contact Sareum([email protected]) and include a short motivation statement for this project, along with your CV and transcripts.

[CLOSED] Real-World Reinforcement Learning gym for alpine robots (semester project/Thesis)

We have a robot that is using built-in compliance in the frame made of fiberglass rods and tendons to adapt its shape to the terrain it locomotes on. Tendons can be used to change the general shape of the robot. Due to its compliance and use of tendons, simulated models are expensive, which hinders the training efficiency and even makes it intractable. In addition, since this robot is meant to evolve on Alpine terrains we want to explore how learning in the real world can directly help learning adapted policies. To this end, like parallel environments in a simulated gym environment, we want to build multiple versions of our robot and deploy RL algorithms on the fleet of robots. You will work closely with a PhD student on building a platform that is easy to replicate, and if time allows deploying distributed RL learning algorithms like DPPO.

Requirements:

• A basic understanding of Arduino, CAD, raspberry would be useful for hardware building
• High motivation
• RL experience (a plus)
• Good python skills

Contacts: Arnaud Klipfel ([email protected])

Morphology estimation of a compliant Alpine robot (semester project)

We have a robot that is using built-in compliance in the frame made of fiberglass rods and tendons to adapt its shape to the terrain it locomotes on. Tendons can be used to change the general shape of the robot. In these different shape configurations the robot has to be controlled using a different policy.

We are looking for a motivated student to help on the hardware building side and on prototyping the recognition of the shape based on the driving dynamics through data driven methods. For this project, a wheeled robot configuration will be used.

Requirements:

• a basic understanding of Arduino, CAD, raspberry would be useful
• High motivation
• Field robotics experience is a plus
• Machine Learning, deep learning experience

Contacts: Arnaud Klipfel ([email protected])

Learning control policies for autonomy in case of hardware failure (semester project)

We have a robot that is using built-in compliance in the frame made of fiberglass rods and tendons to adapt its shape to the terrain it locomotes on. Tendons can be used to change the general shape of the robot. Compliance makes the robot robust, but can also produce changes overtime in the frame dynamics. In addition, in a long field mission, failures on the hardware and software side are expected.

We are looking for a motivated student to work on designing a controller that could handle hardware issues, changes and failures, such as for instance a broken motor or a broken part of the frame, or even operating under low battery, which would cause the dynamics of the platform to change.

Requirements:

• a basic understanding of Arduino, CAD, raspberry would be useful
• High motivation
• Field robotics experience is a plus

Contacts: Arnaud Klipfel ([email protected])

[CLOSED] Design of an alpine climbing robot (semester project)

Similar to a pipe climbing robot, we want to design a robot that can go up steep mountain faces or buildings by leveraging crags or excavations (irregular open pipes). We are looking for a master student with a good experience in design to propose a robotic platform. You will closely work with a PhD student who has experience in control and field robotics. Field robotics experiments are expected at the end of the project.

Requirements:

• a basic understanding of Arduino, raspberry
• CAD design, 3D printing
• High motivation
• Field robotics experience is a plus

Contacts: Arnaud Klipfel ([email protected],main contact), Kyle Walker ([email protected])

Energy production from biomass to extend the autonomy of robotic field operations (Thesis)

A bioreactor/methanisation transforms biomass into energy.This project aims at helping on the automation and optimization of a bioreactor to produce as much energy as possible while minimizing the energy consumption. You will closely work with a Ph.D. student to optimize the energy production of the bioreactor for its embedding on a mobile robot platform, i.e. you will work on developing the first artificial stomach embedded on a mobile robot. The goal is to have a robot that is capable of consuming biomass for long-term operations.

Requirements

• Programming
• Arduino, esp32 experience
• Experience in wet lab or biology skills (a plus but much appreciated)

Contacts: Arnaud Klipfel ([email protected])

Autonomous Robotic  Environmental Monitoring in the Alps (Master Thesis)

Climate change is causing faster and more erratic shifts in weather conditions, leading to significant changes in the distribution of insect, bird, and animal species across regions. Using a robot designed to navigate alpine terrain, we aim to develop a system that can collect environmental data and autonomously choose the next location to sample, based on information gain.

Your role will focus on helping with the control of the robotic platform to maximize coverage of a target area. You will also assist with maintenance and monitoring of the robot and its systems.

Requirements

• Control skills for robotics
• Microcontroller experience : Arduino, esp32
• Robotics skills

Contacts: Arnaud Klipfel ([email protected])

Hand Pose Reconstruction from Video for Dexterous Robotic Manipulation (Master Thesis / Semester / Summer Project)

Outline: This project aims to develop a method to capture hand pose from human demonstrationsin video to learn robotic manipulation skills.

Background: Learning a policy from human demonstrations can effectively leverage human knowledge to teach robots complex manipulation tasks. However, collecting tele-operated demonstrations is often limited by the substantial physical and cognitive demands placed on the human expert. As a result, learning from video demonstrations offers a promising alternative.

This project focuses on dexterous manipulation using a multi-fingered robotic hand. The proposed method involves two main steps: (1) capturing the hand trajectories and object interactions from demonstration videos, and (2) mapping the estimated human hand poses to the corresponding robot hand actions, accounting for differences in their action spaces. This approach aims to simplify the data collection process for learning from demonstration.

Work Package: In this project, we will first use a hand pose estimation method to extract the human expert’s hand poses from video. Next, we will construct a mapping that translates human hand actions into robot hand actions. If time and progress permit, we will use the collected demonstrations to train an imitation learning model for a specific manipulation task.

literature review about most recent related work 10%, implementation of open-source project 45%, algorithms development 45%

Requirements — strong programming experience (python), machine learning experience, knowledge of computer vision is a plus, knowledge of ml-based manipulation controllers is a plus (e.g. reinforcement learning, imitation learning)

Contact — Cheng Pan, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

Simulated Tactile Feedback for Learning Contact-Rich Robotic Manipulation (Master Thesis)

Outline: This project aims to simulate the tactile feedback that can be leveraged to improve any manipulation task using Reinforcement Learning or or Imitation Learning.

Background: Reinforcement Learning or Imitation Learning methods have been widely applied for robotic manipulations via sim-to-real transfer, typically with proprioceptive and visual information. However in fine-granded and contact-rich robotic manipulation tasks, tactile sensing is critical for the policy performance. However, how to simulate tactile feedback during the interaction between manipulator and objects is a challenge. Furthermore, if using the physics engine’s contact data (from MuJoCo, Isaac Gym, etc.) to approximate tactile feedback. The difference of fidelity and characteristics to real tactile sensor, and noise characteristics of the real sensor can affect the performance of learned policy. To deal with issue, we need to build a robust mapping between the tactile feedback estimated by simulator contact information and real tactile sensor reading.

Work Package:
In this project, we will first simulate the tatile feedback in a simulator. Then with the tactile input, we train a policy using RL/IL in simulation. By building a mapping between the tactile readings in simulation and real-world and apply domain randomization for the model parameters, we enable successfully transfer of policies from the simulation to the real robotic manipulator.

literature review about most recent related work 10%, simulation for tactile sensing 45%, algorithms development and implementation 45%.

Requirements — strong programming experience (python), machine learning experience, knowledge of tactile sensing is a plus, knowledge of ml-based manipulation controllers is a plus (e.g. reinforcement learning, imitation learning)

Contact — Cheng Pan, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

Sim-to-Real Transfer for Robotic Manipulation via Policy Distillation and Domain Randomization (Master Thesis)

Outline: This project aims to develop a robust and generalizable policy for robotic manipulation using Reinforcement Learning (RL), transferable from simulation to the real world.

Background:

For robotic manipulation, policies trained using RL can result in more general and robust dexterous skills. However, transferring these policies from simulation to the real world—known as the sim-to-real gap—presents challenges:

State Representation Mismatch: In simulation, full state information (e.g., object and manipulator poses) is available, whereas in the real world, only partial observations such as RGB images and robot proprioception are accessible. To bridge this gap, we need to distill the learned policy into one that can operate using real-world observations.

Behavioral Mismatch: The behavior of the robot in simulation often differs from that in the real world. Domain randomization is necessary during training to account for these discrepancies and improve policy transferability.

Work Package:
In this project, we will first train a policy using RL for a robotic manipulation task, leveraging privileged information during training. Domain randomization techniques will be applied to mitigate the sim-to-real gap. Once the policy is trained, we will distill it into a version that relies only on RGB images and proprioception. Finally, we will evaluate the policy on a real-world robotic system.

literature review about most recent related work 10%, simulation for robotic manipulation 40%, algorithms development and implementation 50%.

Requirements — strong programming experience (python), machine learning experience, knowledge of simulation for robot, knowledge of ml-based manipulation controllers is a plus (e.g. reinforcement learning, imitation learning)

Contact — Cheng Pan, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

Improving Policy via Expert Feedback for Robotic Manipulation Tasks (Master Thesis)

Outline: This project aims to improve robotic manipulation policies by incorporating expert feedback or corrective suggestions when the policy makes mistakes.

Background: Learning robust policies for complex manipulation tasks remains a challenge, as learned policies often fail in subtle or safety-critical ways. Rather than refining reward functions (in Reinforcement Learning) or randomly collecting more demonstrations (in Imitation Learning), a more effective approach is to query experts for feedback on failure cases. Such feedback can be evaluative (e.g., good/bad), preference-based (e.g., better than another), or corrective (e.g., in the state-action space). While corrective demonstrations provide the most informative guidance, they demand more effort. In contrast, evaluative and preference feedback are less burdensome to provide but may require larger quantities for equivalent performance gains. Exploring this tradeoff between feedback informativeness and expert burden is a key question. Moreover, effectively aligning a pre-trained policy with different forms of expert feedback remains a non-trivial and open challenge.

Work Package:
In this project, we will investigate policy update methods that incorporate different types of expert feedbacks. We aim to evaluate the effectiveness of each feedback type in improving policy performance, with particular attention to the tradeoff between feedback informativeness and the intervention effort required from the expert. The initial phase will focus on implementing and testing these methods in simulated environments. If time and progress permit, we will extend our experiments to a real-world robotic manipulation platform.

literature review about most recent related work 10%, simulation for robotic manipulation 40%, algorithms development and implementation 50%.

Requirements — strong programming experience (python), machine learning experience, knowledge of simulation for robot is a plus, knowledge of ml-based manipulation controllers is a plus (e.g. reinforcement learning, imitation learning)

Contact — Cheng Pan, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

Data Augmentation for Robust Policy Learning for Robotic Manipulation (Master Thesis)

Outline: This project aims to improve the generalization capability of robotic manipulation policies learned from expert demonstrations by leveraging data augmentation techniques.

Backgroud: Learning robotic manipulation policies from tele-operated demonstrations is often constrained by limited data, leading to poor generalization and out-of-distribution problems. To mitigate this issue, data augmentation is employed to enhance policy robustness by generating synthetic variations of the original data. These augmentations span object spatial poses, object types, camera viewpoints, environmental appearances, and different robot embodiments, thereby enriching the training set and improving the policy’s adaptability to unseen scenarios.

Work package: In this project, we will begin by reimplementing recent data augmentation algorithms within a simulated environment. Building upon these foundations, we will then explore the development of novel augmentation methods. If time and progress allow, we will optionally extend our work to a real-world robotic setup.

literature review about most recent related work 10%, simulation for robotic manipulation 40%, algorithms development and implementation 50%.

Requirements — strong programming experience (python), machine learning experience, knowledge of simulation for robot and data augmentation is a plus, knowledge of ml-based manipulation controllers is a plus (e.g. reinforcement learning, imitation learning)

Contact — Cheng Pan, [email protected]

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

(Master Thesis) Vision-based state estimation for Soft Continuum Robot in the unstructured environment

Soft continuum robots exhibit exceptional motion flexibility and reachability due to their inherent compliance, but their theoretically infinite degrees of freedom pose significant challenges for accurate pose estimation. While vision-based methods enable high-resolution state estimation, their performance degrades in unstructured environments where occlusions—particularly those arising from human-robot interaction—obstruct visual feedback. This project aims to overcome this limitation by integrating an inertial measurement unit (IMU)-based proprioceptive system to capture the robot’s internal state with high precision, even during complex interactions. The proprioceptive data will serve as ground-truth labels to train and validate a vision-based deep learning model, enabling robust 3D pose estimation under partial visual occlusion. The outcome will advance the deployment of soft continuum robots in collaborative settings, ensuring reliable operation in occlusion-prone scenarios.

This project aims to develop a learning-based visual estimation algorithm that utilizes dual-view cameras to achieve state estimation for soft continuum robots, ultimately enabling robust performance in partially occluded human-robot interaction scenarios.

Workpackages: 10% camera-based sensing system design, 20% soft arm kinematics modelling, 70% learning-based state estimation algorithm design.

Requirement: 1. deep learning algorithm background (ResNet, LSTM and Transformer), 2. robotic system dynamics background, 3. matlab and python programming experience.

Contact: [email protected] (Guanran Pei)

To Apply: please send email with CV, transcripts.

Online Learning of Dynamic Control for Soft Manipulators (Master Thesis)

Advisor(s): Qinghua Guan, Hao Ma, Cheng Pan

Supervisor(s): Prof. Josie Hughes, Prof. Melanie Zeilinger, Michael Muehlebach

Introduction

Soft robots offer notable advantages over rigid robots in terms of flexibility and compliance, which facilitate safe and robust interactions with environments. These characteristics make them ideal for diverse applications, such as medical contexts and bionic robotics. However, due to the absence of precise models and the time-varying nature of their system dynamics, achieving optimal control performance on soft robots remains a significant challenge. With the rapid development of machine learning, we now have various learning-based methods to achieve precise pose control of soft robots. Nevertheless, in practice, we have found that the (off-line) well-trained control policies often become ineffective upon deployment due to time-varying system dynamics, which lead us to rethink the traditional “sampling-training-deployment” paradigm in machine learning.

The rise of online learning can effectively enable models to adapt to changing system dynamics. Recent studies have shown that in online learning, even when using an approximate (or even linear) model, it is possible to ensure a sublinear convergence of the regret. Additionally, online learning demonstrates high learning efficiency and stability in both simulations and complex real-world systems.

In this project, we will adopt gradient-based stochastic online learning to achieve precise pose control of a soft manipulator. The cable-driven soft robot arm comprises three independent modules (about 0.2m for each module), and each module is actuated independently with three cables. More details of the robot setup, such as structural and mechanical description, can be found in our previous work. We will first establish an approximate model of the system, then learn both feedforward and feedback controllers in an online manner, which enable the soft arm to perform a  series of acrobatic demonstrations. Finally, we will compare our approach with some existing offline control algorithms to evaluate the performance in terms of learning efficiency, handling distribution shifts, and tracking accuracy.

Requirements

• Highly motivated for the topic
• Programming experience (Python, PyTorch, Matlab)
• Knowledge of machine learning, convex optimization and dynamic systems
• Knowledge of reinforcement learning is a plus
• Practical experience in robotics applications is a plus
• Experience in vision systems is a plus

Contact (please cc all emails)

Qinghua Guan, [email protected], EPFL 

Hao Ma, [email protected], ETHZ

Cheng Pan, [email protected], EPFL

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

Imitation Learning / Inverse Reinforcement Learning for Dexterous Robotic Manipulation (Master Thesis)

Outline

The goal of this project is to learn control policies for robotic hand manipulation tasks from a set of human expert demonstrations using inverse reinforcement learning and the IsaacSim simulation environment.

Motivation

Dexterous robotic manipulation tasks, like grasping different objects, are challenging control tasks due to the many degrees of freedom of a robotic hand and the variability of materials involved. Reinforcement learning with GPU-accelerated simulators like IsaacSim offers a promising solution, but sparse reward functions (e.g., only rewarding task completion) hinder efficient optimization. Hand-crafting dense rewards is difficult and often leads to suboptimal behavior. On the other hand, learning from human expert demonstrations via imitation learning or inverse reinforcement learning (IRL) can help to effectively guide policy optimization. While offline imitation learning approaches, such as behavioral cloning, are simple to implement, they often suffer from poor generalization due to covariate shift. The key problem is that the imitating policy performs poorly on states that were not encountered in the training data. In contrast, IRL allows to effectively incorporate new samples collected in simulation, which leads to improved generalizability. Moreover, IRL enables learning from state-only observations, potentially eliminating the need for teleoperation and simplifying the process of collecting expert data.

Milestones

• Simulation for robot hand based on IsaacSim.
• Extract human hand pose using a VR headset, and collect demonstration data.
• Develop imitation learning algorithm to learn a policy for the manipulation task.
• Add domain randomization to simulation, if necessary.
• Implement whole pipeline in real-world tests using a robot arm and multi-finger robot

hand.

• Combine imitation learning with RL to improve upon experts’ performance.

Requirements
We look for motivated students with a strong background in machine learning and coding. Furthermore, experience with ROS and robotic simulation is a plus.

References:

[1] Pan, Cheng, Kai Junge, and Josie Hughes. “Vision-Language-Action Model and Diffusion Policy Switching Enables Dexterous Control of an Anthropomorphic Hand.” arXiv preprint arXiv:2410.14022 (2024). https://vla-diffu-switch.github.io/

[2] Eze, Chrisantus, and Christopher Crick. “Learning by Watching: A Review of Video-based Learning Approaches for Robot Manipulation.” arXiv preprint arXiv:2402.07127 (2024).

[3] Hejna, Joey, and Dorsa Sadigh. “Inverse preference learning: Preference-based rl without a reward function.” Advances in Neural Information Processing Systems 36 (2024).

[4] Goecks, Vinicius G., et al. “Integrating behavior cloning and reinforcement learning for im-proved performance in dense and sparse reward environments.” arXiv preprint arXiv:1910.04281 (2019).

This project will be supervised by Prof. Josie Hughes, Prof. Maryam Kamgarpour, Cheng Pan ([email protected]), and Andreas Schlaginhaufen ([email protected]).

Contact: [email protected], [email protected], 

Advancing proprioceptive sensing for squid-like swimming robotics (Thesis)

Summary: Advancing soft robotic swimmers requires integrating proprioception to sense motion without external aids, enhancing their adaptability and autonomy. Previous work presents an approach embedding pressure sensors in soft underwater structures, paired with learning-based methods to reconstruct shape and assess swimming performance. Robust proprioceptive sensing is demonstrated on a soft robotic squid. Extending this, we propose a fully sensorized robotic swimmer, incorporating embedded sensing across its tentacles and body to enable self-awareness of movement, improving control and performance without reliance on external systems like cameras. We also consider the possibility of integrating decentralized control for adaptive control maneuvers when the morphology changes (i.e. structure and shape of robot changes through damage).

References: 

https://ieeexplore.ieee.org/abstract/document/10121999

https://www.nature.com/articles/s41467-024-54327-6

Workload: Programming (Python), Machine learning, Real-world robot testing, Hardware integration

Contact:[email protected], [email protected]

Adding steering to a parametric robotic swimming (Thesis)

Summary: A scalable, parametric design for bio-inspired robotic swimming fish enables adaptation across sizes and fluid environments. By leveraging minimal actuation and flexible but strong materials, the design stores and releases energy for bio-inspired swimming behavior. We would like to add a steering mechanism and integrate electronics into the fish. Experimental validation includes parametric design optimization, frequency response analysis, and field tests, showcasing adaptability to environments from small rivers to large lakes. This approach provides a versatile framework for designing swimming robots for applications such as environmental monitoring and ecological research, emphasizing efficiency and scalability across diverse contexts.

Workload: Programming (Python), Real-world robot testing, Robotic hardware design & integration

Contact: [email protected]

LLM-FoldIt, a Large Language Driven approach to robotic fabric folding 

Summary: Deformable object manipulation (DOM) for robots has a wide range of applications in various fields such as industrial, service and health care sectors. However, compared to manipulation of rigid objects, DOM poses significant challenges for robotic perception, modeling and manipulation, due to the infinite dimensionality of the state space of deformable objects and the complexity of their dynamics. In this thesis, you will leverage the embedded physics intuition of Large Language models to create a policy for fabric handling and folding. The thesis aims at publication in top robotics journals. 

References: 

https://arxiv.org/abs/2208.10552 

https://www.linkedin.com/posts/nextgenaimedia_robotics-ai-automation-activity-7267593943616278529-fSCm?utm_source=share&utm_medium=member_desktop 

Workload: 50% Software development (computer vision), 50% Real world robotic testing and deployment

Contact: [email protected]

Visual Language Action models for Safe Collaborative Robots

Summary: Robotics has long been promising to come close to humans and support them in their daily living tasks (DLT). However, to do so, we require non-trained individuals to be able to give instructions to robots, and robots that can interact safely with humans. Visual Language action models provide a powerful tool to instruct robots through language but do not guarantee the safety of the generated policy. To ensure a safe deployment in close contact with people, we need to be able to deploy these algorithms on inherently safe robotic systems. In this thesis, you will train and deploy a Visual Langage Action Model on the Helix soft continuum manipulator, and demonstrate its capabilities on assistive, close contact tasks. The thesis aims at publication in top robotics journals. 

References: https://openvla.github.io/  

Research path: 

1. Define a family of trajectories for DLT and collect an Image-Action-Goal dataset. 
2. Train a Visual Language Action model
3. Test and characterize the behavior on the real system 
4. Perform a demonstration of the acquired robot capabilities on the DLT

Workload: 70% Software development, 30% Real world robot deployment 

Contact: [email protected], [email protected] 

Bio-inspired Robotic Flower: A Lightweight Environmental Information Collection Unit Summary: Global environmental change, particularly rising temperatures, is causing alpine glaciers to retreat, severely impacting ecosystems and human societies. Understanding and monitoring biodiversity changes in glacial ecosystems is essential, especially regarding the relationship between insects and plants, which serve as key indicators of biodiversity and environmental change. Currently, data collection is conducted manually, limiting the number and scope of samples. There is an urgent need for new methods to improve data collection, analysis, and responses to ecosystem changes. Sensors applied in robotics, such as cameras, barometers, temperature sensors, and acoustic sensors are considered to hold great potential for environmental monitoring and data collection. Based on these sensors, we aim to design a lightweight workstation incorporating a biomimetic flower. This station is intended to attract and capture insects while simultaneously collecting environmental sounds, images of animals, and basic environmental information such as temperature and humidity, all for assessing biodiversity in the ecosystem.

Research path: 1. Biomimetic flower design for attracting and capturing insects.

2. Raspberry Pi shield design for multi-sensor integration.
3. Code development for biomimetic flower and multi-sensor platform.
4. Outdoor experimental testing of the robotic station’s performance and environmental data collection.
5. Processing and analyzing outdoor experimental data using AI tools to assess environmental species diversity.

Workload: 20% Literature review, 50% Hardware design and testing, 30% Programming and data processing

Contact: [email protected]

Wearable devices to enhance sensations for robotic hands (Semester/Thesis)

Summary: Wearable devices and smart gloves are revolutionizing human-machine interaction and robotic control. This project aims to develop wearable devices to enhance the sensing capability of robotic hands. Our goal is to gain in-depth understanding of human behaviour and human-machine-environment interfaces. The focus will be on innovative sensor integration (force, bending, temperature, proprioception, etc.), comfortable ergonomic design (wearables like gloves for both human and robot), and real-time data transmission and interaction with robotic hands.

Workload: 20% Sensor/actuator design and fabrication, 30% Wearable device integration and development, and 50% Software and data analysis. Required experience in computer vision (CV).

Contact: Benhui Dai ([email protected])

Chewing robot with mechanism design, sensing, and control (Semester/Thesis)

Summary: The process of chewing food is achieved through the up-and-down movement of the jaw and the shearing movement of the teeth. Develop a 6 DOF chewing robotic platform, including mouth, jaw, and tooth. The goal is to simulate the mechanical movements of humans chewing by robots and monitor this process by tactile sensors and cameras. What’s more, artificial saliva and a soft robotic tongue will be embedded into the system to provide a biomimetic oral environment for better food-robot interaction.

Workload: 50% design and fabrication, 50% control

Contact: Benhui Dai ([email protected])

Robotic cutting mechanism for semi-autonomous extraction of underground cabling (Thesis)

Summary: As the network of underground cables continues to grow, so do the difficulties associated with maintenance. Often cable ducts contain multiple sizes and types of power/communication lines (copper cables can regularly be >10cm in diameter), leading to challenges if a specific cable has to be replaced and damage to others needs to be avoided. This project will focus on developing a method to cut these cables that is both lightweight and compact, which can be integrated onto a small-scale robot. The goal is to transform typical, manual cutting methods into a portable, teleoperated or autonomous form, so that cutting can occur inside the duct rather than only at manhole entry points. Please apply if you have a strong mechanical/mechatronic background and are interested in mechanism design.

Workload: 10% Literature Search, 60% Design and prototyping, 30% Experimental validation (rough percentages)

Contact: Kyle Walker ([email protected])

Multi-modal compliant manipulators for unmanned marine robots (Thesis)

Summary: This project will investigate the use of naturally compliant manipulators for mobile, floating-base robots, which can be used for both classic manipulation of objects or to anchor the floating-base to a nearby structure by means of a “mode” (stiffness) change. The target use case will be for marine surface vehicles, however the design should be applicable to subsea vehicles as well. Both the physical manipulator design and programming teleoperated/autonomous control algorithms will be key aspects of the work. The end goal is to realise a working prototype and demonstrate this in a set of experiments – previous experience working with robots in a marine setting is beneficial but not crucial.

Workload: 10% Literature Search, 30% Design and prototyping, 30% Control, 30% Experimental validation (rough percentages)

Contact: Kyle Walker ([email protected])

Image-based terrain sensing for FLEXIV (Thesis)

Summary: FLEXIV (FLEXible Inspection Vehicle) is a two-track mobile robot with adaptable features, leveraging track flexibility and a real-time adaptive controller to navigate various terrains. It excels at adaptive slope climbing, stepping down, and recovery, guided by an IMU-based feedback control system. This project aims to develop an image-based controller for FLEXIV, utilizing onboard camera imagery for environment detection for better driving adaptation. It involves integrating the camera, ensuring data transfer and processing, and conducting thorough robot testing. 

Workload: 50% Software development, 50% Robot testing and data collection 

Contact: Sareum Kim ([email protected])

Image-based terrain sensing for FLEXIV (Thesis)

Summary: FLEXIV (FLEXible Inspection Vehicle) is a two-track mobile robot with adaptable features, leveraging track flexibility and a real-time adaptive controller to navigate various terrains. It excels at adaptive slope climbing, stepping down, and recovery, guided by an IMU-based feedback control system. This project aims to develop an image-based controller for FLEXIV, utilizing onboard camera imagery for environment detection for better driving adaptation. It involves integrating the camera, ensuring data transfer and processing, and conducting thorough robot testing. 

Workload: 50% Software development, 50% Robot testing and data collection 

Contact: Sareum Kim ([email protected])

Machine learning approaches for reliable multimodal sensory interfaces (Thesis)

The human skin is capable of detecting multiple sensory stimuli in a selective manner. Artificial sensory interfaces attempt to mimic the multisensory features of our human skin, but classifying responses when multiple stimuli are present remains a challenge. With 3D printing methods, conductive sensory patterns can be integrated into inert substrates, mimicking the way sensory receptors are positioned into human skin. The final goal of the project is to exploit data-driven machine learning based approaches, to enable classification and localization of an applied sensory stimulus upon a 3D printed E-skin interface, even when multiple stimuli are simultaneously present.

Workload: 10% Literature search, 50% Prototyping, 30% Programming and data curation, 10% Thesis Writing

Contact: Antonia Georgopoulou ([email protected])

Design and data analysis of soft optical fiber sensor for proprioceptive sensing (Thesis/Semester)

Proprioception is our body’s ability to sense its position and movement in space. We’re working on developing a soft sensor that uses light to detect bending motion and location within a single flexible fiber. The setup consists of a light emitter on one end and a color sensor on the other. As the fiber bends or deforms, the color sensor detects changes in the transmitted light, providing data about the movement. Our goal is to optimize the design of this soft optical fiber sensor to deliver the most accurate and responsive data. By analyzing the collected data with machine learning, we aim to enhance the sensor’s capabilities for proprioceptive sensing applications.

Workload: 40% Design and fabrication 40% Data analysis 10% Literature search 10% Writting

Contact: Haewon Jeong ([email protected])

Study and Implementation of Autonomous Energy Solutions for Mobile Robotic Applications (thesis)

Energy autonomy is a critical challenge in mobile robotics, limiting operation time and effectiveness in prolonged tasks, particularly in remote or resource-constrained environments. This thesis aims to study, analyze, and implement energy harvesting and storage systems to enable self-sustained robotic operations. By leveraging renewable energy sources such as solar, wind, or biomass, the goal is to design and demonstrate a practical energy autonomy solution for low-tech mobile robot platforms. The research will provide insights into the feasibility, efficiency, and integration of various energy solutions in robotic systems.

Possible sustainable energy sources to study more extensively are : 

• Digestion/fermentation of biomass
• Digestion/fermentation in aqueous environments
• Wind
• Solar
• Kinetic energy storage from robot motion

Workload: 30% literature review and comparison of different energy solutions, 30% prototype design, 30% tests.

References

• Towards enduring autonomous robots via embodied energy | Nature 
• “Gastrobots”—Benefits and Challenges of Microbial Fuel Cells in FoodPowered Robot Applications 
• Row-bot: An energetically autonomous artificial water boatman | IEEE Conference Publication
• Power solutions for autonomous mobile robots: A survey – ScienceDirect

Contact: Arnaud Klipfel ( [email protected] )

Robotic Stomach Design (thesis)

Digestive malfunctions are linked to a variety of health conditions, including metabolic disorders, brain lesions, and gastrointestinal diseases. Current research in digestive systems lacks a reliable, controlled method to replicate the physical and mechanical processes of the human stomach. This thesis aims to design and prototype a robotic stomach that mimics the motions and environmental conditions of human digestion. By reviewing existing artificial stomach designs and integrating advancements in soft robotics, the research will create a minimalistic model suitable for replicating the mechanical digestive motions.

1. State of the art review
2. Mechanical design and fabrication
3. Actuation
4. Tests

Workload: 30% literature review, 30% prototype design, 30% tests.

Contact: Benhui Dai ([email protected]), Arnaud Klipfel ([email protected]) 

Sensitivity improvement and calibration of tactile sensor with hall-effect sensors

Summary: We have developed a soft tactile sensor by using a magnet and a hall-effect sensor. This sensor detects external stimuli by measuring changes in magnetic flux caused by the displacement of the magnet due to external forces. In this project, we aim to enhance the sensor’s sensitivity through design modifications or changes in the soft substrate material. The second objective is to investigate and calibrate the correlation between the improved sensor’s signal and the applied force. Through this project, you will learn the mechanism of soft sensors and magnet-based tactile sensors. In addition, you can acquire basic data processing skills and experience in sensor calibration.

Workload: 5% literature study, 45% design and fabrication (CAD, 3D printing, and silicone fabrication), 35% programming and data post-processing (Python and Arduino), 10% Experimental setup

Contact: Sudong Lee [sudong.lee (at) epfl.ch]

Feedback control of the cable-driven spherical joint

Summary: We have developed a cable-driven spherical joint with a wide workspace and 3 degrees of freedom (rotation). Currently, this joint is capable of open-loop position control based on an analytic model or a model trained using machine learning (supervised). In this project, we aim to embed a sensor such as IMU into the joint and complete a closed-loop system to enhance position control accuracy. Through this project, you will gain knowledge of the fundamentals of motor control and basic closed-loop control.

Workload: 10% literature study, 45% programming and data-processing (Python and Matlab), 40% control, 5% Experimental setup

Contact: Sudong Lee [sudong.lee (at) epfl.ch]

Vision-based Configurations Estimation of Soft Continuum Manipulators (Thesis)

Summary: Soft continuum manipulators, like those modeled after elephant trunks, offer greater safety and flexibility because they can deform to absorb impacts and have virtually unlimited movement possibilities. These features make them ideal for interactions in unstructured settings. However, controlling these manipulators precisely is challenging due to their extensive deformability. The piecewise constant curvature model helps in effectively sensing and managing their movement.
For configuring soft manipulators, various sensing systems including external motion capture and internal IMU-based systems are used, which show good accuracy but have limited sensing capabilities. Vision-based systems, potentially offering unlimited sensing capabilities, are expected to provide more accurate configuration reconstructions. Deep learning techniques are useful for identifying object shapes, and this study aims to develop a neural network to define and reconstruct the configuration of a specific soft robot named “Helix v1” based on these identified shapes, even in visually obstructed, unstructured environments.
Link to proposal: https://drive.google.com/file/d/1-bAx0HVkEPJ4X6GTsJZVfePo3m5Jr6XJ/view?usp=drive_link
Link to previous soft sensing research: https://drive.google.com/file/d/1crSa5Z6a9zsTq8MjV8qEY5D2P5GE0Q0i/view?usp=drive_link
Workload: 30% hardware, 70% software
Contact: Guanran Pei ([email protected])

[filled] Control / Learning / (Haptic) teleoperation on a dexterous robotic hand (Thesis Project)

TL;DR: 
A thesis project to perform some control/learning/teleoperation using the ADAPT Hand v2. Details and directions up to discussion. 

Summary:
The ADAPT Hand v2 is an anthropomorphic robotic hand developed in this lab. However, the control of such a hand for robotic tasks are non-trivial given the complexity of manipulation and the robot itself. 
Recently the hand has acquired tactile sensors in its fingers and thumb, and a teleoperation system using the Apple Vision Pro headset (See video). 
We have several projects lined up with all potential to roll them out as master thesis projects. 
1. Shared control using tactile feedback with the sensorized robotic hand to assist kinematics based teleoperation. 
2. (Potentially bimanual) Dexterous manipulation leveraging immitation learning through real world data collection. 
3. Using haptic feedback to control not only the position but also the force of the interaction of the robot hand. 

If you also have ideas related to this we can discuss!

Skills involved / would be nice if you are confident or touched upon even a few:
– Data collection, processing
– Robotic systems integration (not really to integrate systems but have experience and intuition on working with complex systems with multiple processes and parts etc)
– Robotic hardware (in the sense that you will be working with a physical robot)
– Programming with Python

Contact: Kai Junge ([email protected])

Whegs for our GOAT (Thesis or Semester Project)

Summary: Our GOAT(Good Over All Terrain) robot would be delighted to obtain a set of new compliant wheel-legs to increase its operational range, robustness and efficiency. Your project involves evaluating existing compliant wheel-leg mechanisms and designing and prototyping a compliant wheel-leg mechanism. Please apply if you have prior experience in mechanism design, prototyping with non-rigid, compliant materials and are fascinated about mobile, outdoor robots.

Workload: 20% Literature research, 50% Prototyping and design, 30% Experimental validation ( these are arbitrary percentages 😉 )

Contact: Max Polzin ([email protected])

Parachute for our GOAT (Thesis or Semester Project)

Summary: Our GOAT(Good Over All Terrain) robot would be delighted to augment its flying capabilities with gliding as gliding offers increased efficiency for controlled aerial descend. For this, it is your task to adapt and equip our GOAT with an RC parachute. Please apply if you have prior experience in rapid prototyping and servo control and are fascinated by flying/gliding robots.

Workload: 30% Prototyping and design, 30% Control, 40% Experimental validation ( these are arbitrary percentages 😉 )

Contact: Max Polzin ([email protected])

Drowning our GOAT (Thesis or Semester Project)

Summary: Our GOAT (Good Over All Terrain) robot can swim, fly and drive. It is time for it to venture into new realms and go diving. Your task will be to waterproof the GOAT’s electronics, design a prototype buoyancy system and showcase the feasibility of underwater locomotion. Please apply if you have prior experience working with mechanical prototyping and working with water and electronics.

Workload: 15% Literature research, 55% Prototyping and design, 10% Control, 20% Experimental validation ( these are arbitrary percentages 😉 )

Contact: Max Polzin ([email protected])

[filled] Tendon-Driven Lightweight Robotic Arm for Agricultural Applications (Semester) 

Summary: Develop a tendon-driven 7-DoF robotic arm with two-to-three end effectors used in harvesting, such as scissors and grippers. The goal is to minimize joint weight and inertia generated during fast motion. It can be based on the current robotic arm we have and modify it accordingly. Workload: 90% Design, 10% Control

Contact: Paul Cheng ([email protected])

[filled] Motion Stabilization of a floating robot (Semester)

Summary: Analyze and mitigate vibrations caused by a robot moving on a rope. Develop a control approach and mechanical damper to reduce oscillation or explore other methods. The closest scenario is similar to [ https://doi.org/10.1016/j.engstruct.2022.115527] or cable cart, but on a smaller scale with a robot involved. Simulation with an appropriate control method may suffice. The main goal is to stabilize the real robot in its desired position and orientation.

Workload: 60% Modeling and Control, 40% Hardware Design.”

Contact: Paul Cheng ([email protected])

[filled] Analyze and cancel out vibrations caused by the robot moving on a rope.

Develop a control approach and mechanical damper to reduce oscillation (or explore other methods). The situation is similar to but on a much smaller scale and with a robot on it. Simulation with a proper control method might be sufficient. The final goal this the stabilize the real robot in the desired position and orientation.

Workload: 60% Modeling and Control, 40% Hardware Design

Contact: Paul Cheng ([email protected])

[filled] Vision and Motion Training for Harvesting Robotic Arm (Thesis)

Summary: We have a mobile robotic arm designed for harvesting strawberries. We aim to implement a learning-based approach to navigate the robot arm to reach the target position and perform harvesting. Motion data will be captured through teleoperation. The goal is to explore ways to expedite the harvesting process.

Workload: 70% Programming, 30% Hardware Design

Contact: Paul Cheng ([email protected])

Control a soft arm with its twin (Semester/Thesis)

Summary: The inverse kinematics of a soft arm are complex and costly for real-time control. However, the physical world acts as an unparalleled simulator in terms of speed and realism. This is why we use a reduced-scale physical twin to control a larger-scale soft arm robot. Naturally, there are some discrepancies between the two arms, which is precisely what we aim to resolve.

Workload: 90% Programming, 10% Hardware Design, 20% Experimental test, 60% Programming for control, 20% Modeling

Contact: Qinghua Guan ([email protected])

3D printed sensor inspired by the human skin (Semester/Thesis)

Background: Human Skin Sensing

Human skin sensing relies on an intricate network of receptors and neurons embedded within the skin. These receptors vary in their functions, structures, arrangements, and receptive fields. The neural structure not only transmits sensory signals to our brain but also preprocesses this information. It can even influence human behavior through low-level conditioned reflexes. Together, these elements shape our perception of the environment.

Project Goal:

The aim of this project is to develop a multi-functional sensor system that integrates multiple receptors and reconfigurable neurons. This system will investigate the synergistic effects of different sensors and neurons working in concert.

Workload: 70% Hardware Design, 20% Programming for control, 10% Modeling

Contact: Qinghua Guan ([email protected])

Simulation or physics modeling for underwater swimming soft robots using (Thesis)

Summary: Simulating underwater soft structures is challenging. The field has not been explored as much compared to computational fluid dynamics analysis of rigid structures. We have a squid-inspired swimming robot that swims using its 4 soft tentacle arms. We would like to close the simulation-to-reality gap for this robot. Please apply if you have strong simulation or physics-modeling skills. 

Workload:  Simulation 80%, Analysis 20%

Contact: Nana ([email protected])

[filled] Exploring the embodiment of a soft sensor (Semester/Thesis)

Summary: We want to explore how the controllability of robots change if we have different embodiments of a soft sensor. This project will involve fabrication of a soft strain sensor and testing. Please apply if you have strong mechatronics and embedded systems skills. 

Workload: Sensor fabrication 30%, Experiments and system integration 70%

Contact: Nana ([email protected])

Developing a buoyancy bladder for soft swimming robots (Semester/Thesis)

Summary: We have a number of soft swimming robots that need buoyancy control. Currently they are attached to a big floating platform. Your job would be to develop a buoyancy bladder that can be integrated to an existing swimming robot. If you are good at hardware design and fabrication and have previous underwater mechatronics experience, please apply.

Workload: Hardware design and testing

Contact: Nana ([email protected])

[filled] Arts and Soft Robotics (Semester/Thesis)

Summary: This is an open-ended project. We want you to design and build an interactive arts display using soft robotic components or ideas. The underlying aim is to delve into the potential of soft robots in the entertainment field or the potential for them to positively affect human life. Once the piece is built, we will conduct user studies on the human-robot interaction. Creative individuals with prior mechatronics and building experience are welcome. 

Workload: Design and Fabrication 70%, User studies and analysis 30%

Contact: Nana ([email protected]) Please have a rough idea of your arts display before contacting!

[filled] Chewing robot mechanism design and control

Summary:  The process of chewing food is achieved through the up and down movement of the jaw and the shearing movement of the teeth. Develop a simple chewing robotic system, including mouth, jaw and tooth. The goal is to simulate the mechanical movements of humans chewing by robot, and monitor this process by sensors and camera.

Workload: 50% design and fabrication, 50% control

Contact: Benhui Dai ([email protected])

[filled] Pneumatic soft robotic actuators for healthcare (Semester/Thesis)

Summary: Soft robots are made from highly flexible materials that mimic the properties of living organisms, which allows them to adapt to complex environments and friendly interactions. In this project, our goal is to learn the principles of soft robotics, focusing on the structural design of novel actuators, their fabrication with curing and 3D printing, and their control with pneumatic devices.

Workload: 40% Design and fabrication, 30% Hardware, and 30% Control

Contact: Benhui Dai ([email protected])

Anthropomorphic robotic hand augmentation and design space exploration (Thesis)

Summary: With our approach for parametric design and modular actuation, we can create skeletal robotic hands rapidly (<1 day) with high functionality and dynamic capabilities. We want to explore how the hand behaviours evolve with the morphology by replicating and generating and testing diverse hand designs, including primate hands. Additionally, the project would involve augmenting the hand with naturalistic features, such as a soft skin and tactile sensors. You project would involve mechanical design, 3D printing/rapid prototyping, experiment design and analysis. Please apply if you are interested in robotic manipulation and developing integrated systems.

Workload: 50% design and fabrication, 20% modelling and control, 30% experiment design and analysis

Contact: Kieran Gilday ([email protected])

[filled] Implementation and test of a morphology agnostic controller for legged robots using Reinforcement Learning and Generative AI (semester)

Summary : The goal of this project is to adapt and transfer a controller from one morphology to another. Supposing that we have a controller for a given design and task, how can we generate a controller for another design? Current approaches merely retrain the controller in model-free settings or use a different model corresponding to the new design. This is very data expensive and does not re-use anything from what we already have. A main issue in the controllers that are designed in Robotics now is that they are always very specific, single-use, disposable, and are not reusing prior knowledge efficiently and exploiting other controllers that we have on hand.

The proposed approach is to use Reinforcement Learning to adapt a controller using a Generative Model, more specifically a GAN. Conditioned on the morphology and joint locations, the GAN could generate adapted reference trajectories for the target design.

Apply if you are interested in learning about RL, using RL gyms, generative models. You should have coding experience in Python, and some basic understanding of Deep Learning.

workload: 80% implementation of python code and tests, 20% literature review, research for modifications

Contacts: Arnaud Klipfel ([email protected])

[filled] Data collection and processing of plant electro-physiological signals (summer project/semester)

Summary : Plants respond to external stimuli in different ways, these responses can be measured or detected using electro-physiological sensors. If proper signals are collected and suitable models are built, it would be possible to use plants as sensors. Using electrophysiological sensors your goal will be to set up an experimental platform to record these signals and collect a dataset in different light conditions. If time allows, we will build a generative model to predict the light condition changes based on the plant internal state.

Workload : 30% set up of sensor, and plants, 30% data collection, 40% result analysis

Contacts: Arnaud Klipfel ([email protected])

[filled] Wearable devices to enhance sensations for robotic hand (Semester/Thesis)

Summary: Wearable devices and smart gloves are revolutionizing human-machine interaction and robotic control. This project aims to develop wearable devices to enhance the sensing capability of robotic hands. Our goal is to gain in-depth understanding of human behaviour and human-machine-environment interfaces. The focus will be on innovative sensor integration (force, bending, temperature, proprioception, etc.), comfortable ergonomic design (wearables like gloves for both human and robot), and real-time data transmission and interaction with robotic hands.

Workload: 40% Sensor/actuator design and fabrication, 30% wearable device integration and development, and 30% Software and data analysis

Contact: Benhui Dai ([email protected])

Building Adaptive Controller for Rolling Robot (Summer Project)

Summary: The Rolling Robot project focuses on the development of a flexible and soft 1-D locomotion platform designed to navigate complex terrain using adhesive locomotion. To enhance adaptivity in locomotion, we aim to implement sensors and controllers capable of detecting upcoming terrain geometry and ensuring stable locomotion. Your role will involve modifying the design of the existing rolling bot, designing controllers for adaptable locomotion, and conducting robotic experiments to validate performance.

Workload: 20% Prototyping and design, 50% Experimental validation and data collection, 20% Modeling and Simulation

Contact: Sareum Kim ([email protected])

Cartesian Robots in a wet lab: tracking bacteria growth (Thesis/Semester)

Summary:  To lower the access barrier to lab automation we developed a cartesian robot that can pick and place laboratory labware such as petri dishes and, by using computer vision, can assess how different bacteria colonies grow. The robot is currently deployed in a real wet laboratory at UNIL. The candidate will implement new computer vision strategies to track bacteria growth and will improve software and hardware components.

Workload: 40% hardware, 60% Software

Resources: https://www.epfl.ch/labs/create/create-lab/lab-spin-offs/  + more upon request

Contact:  Vincenzo ([email protected]) & Francesco (francesco.stella@epfl@ch)

Automating light scattering (Semester)

Summary: Light scattering is a technique used to examine the unique properties of protein solutions that don’t behave as expected. To do this effectively, scientists need to test the proteins at various concentrations. However, manually adjusting these concentrations can be a long and error-prone process. This project aims to simplify this task by automating the process of mixing protein samples using a dual syringe system. The system will be built in the course of the project using existing technology that is already used in the lab.

More info will be given upon request.

Workload: 25% Mechanical design, 25% Fabrication, 50% Experimental 

Contact: Vincenzo ([email protected])

Autonomous Navigation Integration in Cleaning-Bot for Efficient Path Cleaning (Semester)

Summary:
In the framework of the Swiss CAT+ project, we have developed a concept for the transfer of chemical samples based on a fleet (or swarm) of mobile mini-robots developed locally with ROS and designed around the typical shape of the sample. They are equipped with Ultrasonic sensor to measure their position in the lab. We plan to have about 10 of these mini-robots running in a track suspended above the scientific instruments. The track and the robots already physically exist. The transfer between the scientific instruments and the mini-robots in their track will be handled by 6-axis cobots (Universal Robots 5e or 10e) assisted by a set of industrial cameras for precise positioning.
The objective of this project is to implement an already developed navigation algorithm, originally designed for a similar robot, into the Cleaning-Bot. This integration aims to utilize the Cleaning- Bot’s existing sensor systems to achieve efficient and autonomous cleaning along a predefined track. The challenge lies in adapting and optimizing this existing navigation algorithm to fit the specific hardware configurations and cleaning requirements of the Cleaning-Bot.

OSS. Previous ROS knowledge is highly desirable

Deliverables:

• a) Algorithm Adaptation: Study the existing navigation algorithm and understand its compatibility with the Cleaning-Bot’s sensor systems.
• b) Software Integration: Implement the navigation algorithm into the Cleaning-Bot’s software framework, ensuring seamless integration with its hardware.
• c) Customization and Optimization: Customize the algorithm parameters to optimize the Cleaning-Bot’s path planning, obstacle avoidance, and area coverage capabilities.
• d) Testing and Validation: Conduct extensive testing in various environments to validate the performance of the Cleaning-Bot with the integrated navigation system. Adjust and finetune the algorithm based on real-world feedback and performance data.

Contact:
Edy ([email protected]) & Vincenzo ([email protected])

Development of a Battery Exchange System Using a 6-Axis Robot(Semester)

Summary:
In the framework of the Swiss CAT+ project, we have developed a concept for the transfer of chemical samples based on a fleet (or swarm) of mobile mini-robots developed locally with ROS and designed around the typical shape of the sample. They are equipped with Ultrasonic sensor to measure their position in the lab. We plan to have about 10 of these mini-robots running in a track suspended above the scientific instruments. The track and the robots already physically exist. The transfer between the scientific instruments and the mini-robots in their track will be handled by 6- axis cobots (Universal Robots 5e or 10e) assisted by a set of industrial cameras for precise positioning.
This project aims to develop a prototype system that automates the battery exchange process for a mobile robot using a 6-axis cobot from Universal Robots. The current mobile robot uses a fixed battery system that requires manual replacement upon discharge. An automated system for battery exchange would significantly enhance the robot’s operational efficiency.

OSS. Previous ROS knowledge is highly desirable

Deliverables:

• a) Design of Removable Battery Module: Develop a modular battery system similar to those used in hand drills, which can be easily detached and attached to the robot.
• b) Development of Exchange Mechanism: Utilize a 6-axis Universal Robot to create a mechanism that can autonomously remove a discharged battery from the mobile robot and replace it with a charged one.
• c) Programming and Automation: Program the Universal Robot for precise movements required for the battery exchange process.
• d) Testing and Refinement: Test the prototype under various scenarios to ensure reliability and safety. Refine the system based on feedback.

Contact:
Edy ([email protected]) & Vincenzo ([email protected])

6-month paid robotics internship opportunity
More information here: Internship In Wind Turbine Control
Contact: [email protected]

[closed] Designing a modular robot for simulating cell growth and division in fluid environments (Thesis)

Summary: This project aims to develop modular robots that simulate cell growth and division using a combination of soft materials and/or polymers. The focus will be on designing and fabricating these modular robotic units, intending to replicate the fundamental processes of cell development. Operating in a fluid environment, such as water, the robots will emulate the dynamic nature of real cells. The project requires significant mechanical design and fabrication skills to ensure the successful construction of these robotic units. The research outcomes have the potential in offering new insights into cellular behavior and inspiring innovative technological applications.

Workload: 50% Mechanical design, 50% Fabrication

Contact: Nana ([email protected])

Examining collaboration between multiple Cartesian robotic arms (Thesis) 

Summary: To accomplish diverse manipulation tasks in environments like laboratories or kitchens, the use of dual robotic arms becomes essential. While programming and synchronizing 6-DOF articulated robotic arms pose challenges, employing simpler yet versatile systems like Cartesian platforms often suffices for numerous tasks. The lab’s current two-arm Cartesian setup has demonstrated high reliability in automating lab tasks such as unscrewing caps, pipetting, and dispensing. 

This project aims to explore collaboration beyond two cartesian robotic arms. Investigating horizontal or vertical stacking methods can prove viability of a ‘continuous manufacturing line’, while examining how multiple robotic arms (3, 4, or more) can share the same workspace can prove beneficial for execution of highly complex tasks. 

Your responsibilities will involve testing various stacking concepts, with real-life demonstration of the best configuration across a series of tasks. This project will require design and building of additional arms and end-effectors but will also require robot-arm programming and exploration of viable control methods. This project has high potential for publication in robotics journals. 

Link to previous research: https://drive.google.com/file/d/1E-8BRuCfi-gVaQx2MU9WtCLMqsCOznSF/view?usp=sharing 

Link to videos:  

https://drive.google.com/file/d/1E-fPgmKiFfA_UAA22IgYzuZ2xN6rB_tC/view?usp=sharing 

https://drive.google.com/file/d/1E-zj12UVV84gPPo1B9oSZM7rDngKlSxB/view?usp=sharing 

Workload: 50% hardware, 50% software 

Contact: Stefan ([email protected]) 

Autonomous Tethered TriPod for Enhanced Biomass Quantification in Dense Forests (Thesis)

Summary: Accurate quantification of forest biomass is essential for understanding carbon sequestration and ecosystem health. Traditional methods, however, are often labor-intensive and subject to human error. This project aims to design and implement an autonomous motorized tripod system capable of traversing along a tether and utilizing an onboard laser scanner for precise biomass measurements. The system will feature ground anchoring for stability during data acquisition and a roller mechanism for seamless repositioning. You will be responsible for developing the mechanical design of the tripod, integrating a laser scanner, and programming the movement and data processing algorithms. The project will culminate in field trials to demonstrate the system’s efficacy and potential publications in leading environmental and robotics research outlets.

Workload: 70% Mechanical design and integration, 30% Field testing

Contact: Max ([email protected])

Enhancing passive walking in the PAWS quadruped with embodied contact sensing (Thesis)

Summary: Biological systems display an incredible diversity in behaviors and motion capabilities. It has been shown that for most tasks, only a subspace of the possible poses is used. We propose to exploit minimal feedbacks from contacts at the paws to develop bio-inspired controllers that exploit the embodied passive properties of the PAWS robot. In this thesis you will start by integrating contact sensors in the PAWS quadruped robot, and then link the sensor readings into minimal control inputs to achieve diverse yet stable locomotion strategies. In the thesis we will continue optimizing the design, manufacturing and control of a biomimetic cheetah with a reduced set of actuators. Potential publication of results in top Robotics venues.

Link to prior research: https://www.researchsquare.com/article/rs-3195331/v1 

Workload: 70% Mechatronics design, 30% Control design

Contact: Francesco ([email protected]) & Sudong ([email protected])

Exploiting dynamics for fast soft robots (Thesis)

Summary: Soft robots are celebrated for their ability to efficiently store energy in their bodies, and efficiently releasing them during repetitive motions . We propose to exploit this capability to generate the most energy efficient manipulator, able to perform highly dynamic movements with minimal energy consumption. In this thesis you will start by estimating the eigenfrequecies of the Helix manipulator thanks to the embedded proprioceptive sensing.  You will then merge the eigenfrequency excitations into a higher level controller to perform efficient periodic motions. Potential publication of results in top Robotics venues.

Link to prior research: https://www.nature.com/articles/s44182-023-00004-7 

Workload: 80% control, 20% Mechatronics design

Contact: Francesco ([email protected]) & Qinghua ([email protected])

Soft arm with high load capability: designing, building and modeling (Thesis)

Summary: Leveraging the designability of architectured material/structure, develop a soft manipulator with inertia compliance and also high load capability

-3D Printing of architected arm structure with deliberate mechanical properties.

-Build a tendon-driven robot with soft structure and a rotating basement

-Build the open-loop control of the robot based on mechanical modeling

Workload: 60% hardware, 40% Software

Contact: Qinghua ([email protected])

Soft sensor with multi-mode and programmable sensing (Thesis)

Summary: Develop  a sensor with multi-mode sensing(shearing/compression) ability based architectured lattice material.  Lattice materials have various mechanical properties. By involving different mechanical responses of lattice materials, the sensor can achieve different sensitivity to different . 

-Design and 3D Printing of architected lattice foams

-Design and build capacitive (or resistance/optical ) sensors （including the PCB design for the signal collecting） with programmed lattice structures to achieve different sensing property and modes in different areas.

-Design and build capacitive  (or resistance/optical ) sensors （including the PCB design for the signal collecting） with multi-mode sensing with shearing, twisting and compression based on multi-layer lattice structures. 

Workload: 80% hardware, 20% Software

Contact: Qinghua ([email protected])

Electrostatic adhesion sheet and Characterization (Thesis)

Summary: Fabricate a centimeter-scale electrode and develop a circuit to generate adhesion force (for grasping applications) via electrostatic.
– Make the circuit to generate high voltage and optimize the circuit in compact
– Design patterns for the electrode
– Characterize normal and shear forces according to the input voltage

Workload: Fabrication (20%), Electronics (50%), and Experiment (30%)

Contact: Sudong ([email protected])

Design soft skin with special patterns for adhesive force (Thesis)

Summary: Design soft skin for a rigid robot hand having electrostatic adhesion sheet. We are going to develop patterns and structures of the skin to amplify or optimize electrostatic effects for manipulation.
– Design and make soft skin with silicone
– Embed the electrostatic adhesion sheet
– Test and verify the effects of the patterns and electrostatic forces in grasping applications

Workload:  Design (35%), Fabrication (35%), and Experiment (30%)

Contact: Sudong Lee ([email protected])

(Master thesis) Surgeon test bed : Self-healing phantom for suture practice
In collaboration with CyRIS Lab in Max Planck Institutes (https://is.mpg.de/cyris)

Description — Suture training is essential in medical training of surgeons. However, it is hard to monitor and assess suture quality using the current suturing pad which is normally made of silicone. Since quantitative measuring of suture placement, damage and recovery time is absent. Thus developing a robotic phantom that is capable of self-healing and sensing is needed so that  it can measure damage, suture placement and recovery time. By using self-healing hydrogel with electrical impedance tomography, we are aiming to measure the resistance change of the full surface and reconstruct the image of damage such as cutting, suture placement, etc and also monitoring the recovery time of hydrogel. At the end of the project, the developed phantom will be tested by surgeons and surgical robots in Max Planck Institutes for user study. 

30% Design and fabrication of phantom, 20% Programming of microcontrollers, and 20% Image reconstruction of damage, 20% User study with surgeons

Requirements — Prototyping skills (3D printing, molding), electronics skills (microcontroller), programming skills (python, MATLAB)

Contact — Haewon Jeong ([email protected])

To apply, please include a short motivation for this project, as well as a copy of your CV and transcripts.

Design and control of miniature continuum robots for neurosurgery (Thesis)

Summary: Soft continuum robots can reduce surgery invasiveness with safer approaches and low force interactions. Following conversations with practicing neurosurgeons, we want to develop a novel robot combining concentric tube robots and tendon-driven continuum robots for a simple to control system with guaranteed safety during operations. This project involves: 

• Developing scaled prototypes using 3d printing and other fabrication techniques
• Developing position/tendon actuation and control systems
• Characterising the system using our sensorised brain phantom

Workload:  Design/fabrication (50%), Control systems (40%), and Experiment (10%)

Contact: Kieran ([email protected])

[closed] Scaling up underwater swimming soft robots (Thesis)

Summary: Sea creatures exhibit amazing maneuverability underwater. Many soft fish-like robots exist that are capable of navigating underwater complex environments. However, scaling these robots up to larger sizes pose challenges and have been relatively unexplored. Using materials such as carbon fiber rods, the goal of this thesis is to create a large (>2m), light-weight, soft swimming robot and then to analyze performance capabilities.

Workload:  Design 20%, Fabrication 40%, and Experiment/Analysis 40%

Contact: Nana ([email protected])

[closed] Simulation for underwater swimming soft robots using DiffPD (Thesis)

Summary: DiffPD is a fast differentiable simulator for soft-body learning and control applications. In this thesis, you will create a simulation for soft underwater swimming robots using DiffPD which will help close the simulation-to-reality gap with our robot hardware.

See here for more information about DiffPD: http://diffpd.csail.mit.edu/

Workload:  Simulation 80%, Analysis 20%

Contact: Nana ([email protected])

[closed] Large-scale data capture and online optimization of structures in water (Thesis)

Summary: Previously, we have used a robot scientist to automatically optimize a paper airplane shape so it flies a specified distance when thrown. Robot scientists are particularly useful in that they can collect a lot of data autonomously. In a similar way, we want to see if we can create a closed-loop platform where we gather feedback from fluid dynamics using PIV to optimize a shape underwater autonomously. Your thesis will involve hands-on setup creation, experiments, and optimization.

See a related paper here: https://www.nature.com/articles/s41598-023-31395-0

Workload:  Design/Fabrication 40%, Experiment 40%, Optimization 20%

Contact: Nana ([email protected])

Concentric tube robot for neurosurgery (Thesis)

Summary: Developing a miniaturised concentric tube robot that has additional tendon driven steering at the tip for various neurosurgery applications.

Workload: 50% Design and fabrication, 50% Control and Modelling

Contact: Kieran Gilday ([email protected])

Hand Evolution Analysis (Thesis)

Summary: Using our single print approach for hand manufacture, we want to explore how the passive properties and morphology has evolved with species/time. To do so, this project will involve creating benchmarking methods and analysis, and exploring different hand designs and structures.

Workload: 50% Experiment design and analysis, 50% Fabrication and design

Contact: Kieran Gilday ([email protected]) & Cheng Pan ([email protected])

[Temporarily closed] RoboCup @Home MAKE Project — Software Engineering (Semester)

Summary: 
RoboCup @Home is a MAKE Project where we aim to build a home service robot to participate in the international robotics competition: RoboCup. The aim of this project is to work on the software systems of the robot. Software tasks will be around navigation and behavioural tree of the robot.

In both cases, the student should have some level prior experience in relevant fields, as the project requires practical knowledge to integrate state of the art technologies. Specific project directions will be discussed with each student based on their interest and past experience. 

Workload: 
– 80% technical development / robotic integration 
– 20% decision making and planning of the project with other team members

Contact:
Max Polzin ([email protected])

Synergy Based Quadruped (Thesis)

Summary: In previous projects, we have used the concept of motor synergies to create an underactuated quadrupedal robot that exploits passive behaviours for robustness.  In this project we aim to improve the mechatronics to improve the walking speed, and also investigate further the actuation and emergent properties.

Workload: 50% Mechatronics and design, 50% Control

Contact: Francesco Stella ([email protected])

Closed-loop Control of Soft Robotic Manipulators (Thesis)

Summary: Developing closed-loop control algorithms for soft robotic manipulators using IMU data.  This will leverage our 1m long compliant robot arm that has a large workspace and range of options.

Workload: 80% Control. 20% Hardware

Contact: Francesco Stella ([email protected])

Electrostatic adhesion finger (Thesis)

Summary: Develop the electronics and electrodes to integrate electroadhesion into robot fingers. The project will then explore the characterization of this, and how it can be combined with other mechanical properties (e.g. locking/jamming), and how it can be used to aid control and interactions with objects.

Workload: 50% Electronics, 50% Mechanism design & implementation

Contact: Sudong Lee ([email protected])

Elastic Morphing Wheel to Operate in Extreme Environments (Thesis/Semester Project)

Summary: Design and implement an elastic morphing wheel that improves our GOAT (Great Over All Terrain) robot’s ability to safely operate in challenging natural terrains in extreme environments.

Workload:70% Design and fabrication, 30% Experiments

Contact: Max Polzin ([email protected])

Exploiting Bistability to Operate in Extreme Environments (Thesis/Semester Project)

Summary: Design and implement a mechanism that allows our GOAT (Great Over All Terrain) robot to morph its shape to safely operate in challenging natural terrains in extreme environments.

Workload:70% Design and fabrication, 30% Experiments

Contact: Max Polzin ([email protected])

[closed] Developing distributed tactile sensing on a robot hand (Thesis)

Summary: Ditributed large are tactile sensing is essential for robust manipulation.  Starting from our anthropomorphic robot hand, this project will explore the mechatronic integration of tactile sensors receptors into the hand, and investigate how this can be used for sensory motor control.

The project will begin from this custom dexterous hand, and modify the design to integrate hall effect sensors repurposed for tactile sensing. This work will lead to one of the world’s first dexterous hand with integrated distributed sensing.

The student should have a good understanding and intution/experience on:

• Mechanical design using FDM 3D printing 
• Electrical design and IC components
• Simple PCB design
• Low level signal processing

Workload: 80% Mechatronics integration, 20% Sensory Motor Control

Contact: Kai Junge ([email protected])

[closed] Cartesian Robots in a wet lab: tracking bacteria growth (Semester)

Summary:  To lower the access barrier to lab automation we developed a cartesian robot that can pick and place laboratory labware such as petri dishes and, by using computer vision, can assess how different bacteria colonies grow.

The robot will be deployed in a real wet laboratory at UNIL. The candidate will finish to implement the computer vision program following the needs of the UNIL laboratory and will deploy the robot to study what can be improved software and hardware wise.

Workload: 40% hardware, 60% Software

Contact:  Vincenzo Scamarcio ([email protected]) & Francesco Stella (francesco.stella@epfl@ch)

[closed] Wireless Charging System for Autonomous Mobile Platforms (Semester)

Summary: In the framework of the Swiss CAT+ project, we have developed a concept for the transfer of
chemical samples based on a fleet (or swarm) of mobile mini-robots developed locally with ROS. The project aims to develop a charging system for these mini-robots, utilizing wireless charging
technologies, such as inductive coils. The mini-robots would be able to charge their batteries on
several stations situated on key positions on the track.

Workload: 70% Hardware, 30% Software

Contact: Vincenzo Scamarcio ([email protected]) & Edy Mariano ([email protected])

[closed] Robot Laboratory Cleaner (Semester)

Summary: In the framework of the Swiss CAT+ project, we have developed a concept for the transfer of
chemical samples based on a fleet (or swarm) of mobile mini-robots developed locally with ROS. The project aims for the enhancement and integration of our existing mini-robot to create an
autonomous cleaning robot, adept at removing dust on the dedicated track. This enhanced robot
will take advantage of its existing navigation and sensory capabilities, enabling seamless operation
without disrupting the routines of other robots.

Workload: 80% Hardware, 20% Software

Contact: Vincenzo Scamarcio ([email protected]) & Edy Mariano ([email protected])

[closed] Rescue-bot (Semester)

Summary: In the framework of the Swiss CAT+ project, we have developed a concept for the transfer of
chemical samples based on a fleet (or swarm) of mobile mini-robots developed locally with ROS. The project aims for the enhancement and integration of our existing mini-robot to create an
autonomous breakdown-robot, designed to promptly detect and relocate any malfunctioning robots
to a designated safe zone.

Workload: 80% Hardware, 20% Software

Contact: Vincenzo Scamarcio ([email protected]) & Edy Mariano ([email protected])

[closed] Hydrodynamic simulation of a swimming robot (Thesis/Semester)

Summary: Developing a hydrodynamic simulation of a soft swimming robot using commercial software or other. 

Workload: 100% Simulation

Contact: Nana Obayashi ([email protected])

[closed] Cartesian Robots in a wet lab: tumors pick & placing & analysis (Semester)

Summary: Working toward the goal of personalized healthcare, we developed a low-cost cartesian robot that can efficiently pick and place mice cancerous tissue explants, by using computer vision and innovative pneumatic grippers. 

The robot is ready to be deployed in a real wet laboratory at EPFL to assess its performance. The candidate will study the architecture of the existing robot to propose possible software and hardware changes and will evaluate its performance while executing the tissue explant workflow in a real biological laboratory.

Workload: 30% Hardware, 70% Software

Contact:  Vincenzo Scamarcio ([email protected]) & Francesco Stella (francesco.stella@epfl@ch)

Design and manufacturing of a fully-fledged soft manipulator for in-house care (Thesis)

Summary:
In this thesis you will optimize the design of the soft manipulator presented in https://arxiv.org/pdf/2211.10188.pdf , to target real-world applications, such as in-house care. The goal of the thesis is to build a bigger (around 3m long), stronger, yet safe soft manipulator, equipped with embedded sensors and actuation. We look for students passionate about rapid prototyping, CAD and eager to build the new generation of soft robots. Potential publication of results in top Robotics venues.

Workload:80% design optimization and manufacturing, 20% performance evaluation and demonstration

Contact:
Francesco Stella ([email protected])

Understanding human preferences on manipulators in human-robot interaction (Thesis)

Summary:
Robots are still largely confined to controlled environments, far from the human reach. In the last decades the research community put a strong effort to design robots able to interact safely with humans. However, limited study have evaluated the human preference when it comes to close human-robot interactions. In this thesis you will study and compare the human acceptance between traditional rigid robots, collaborative robots, articulated and continuum soft manipulators with real-world tests and user-studies. We search for students eager to answer the question: what kind of robot do we really want around?

Workload: 30% Experiment design, 40% User study, 30% Data analysis.

Contact:
Francesco Stella ([email protected])

Animal behavior to robot design: a methodological framework for computational design of robot structures from animal synergies (Thesis)

Summary:
While biological systems display an incredible diversity in behaviors and motion capabilities, it has been shown that for most tasks, only a subspace of the possible poses is used. We propose to exploit these emerging principles from animal motor control as inspiration to guide robot design. In this thesis you will start by extracting such synergistic motions from animal videos and data-bases. Hence, you will develop a computational design method to translate the emerging synergies into precise robot specifications. Finally, you will transform the simulation results into the design of a fully fledged biomimetic robot. In particular, the thesis will be focused on continuing and optimizing the design, manufacturing and control of a biomimetic cheetah with a reduced set of actuators. Potential publication of results in top Robotics venues.

For reference see link: https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8373731

Workload:
80% design optimization and manufacturing, 20% performance evaluation and demonstration

Contact:
Francesco Stella ([email protected])

Setting a robot fleet control system for the Swiss CAT+ chemical sample transfer system (Thesis/Semester)

Summary:
In the framework of the Swiss CAT+ project, we have developed a concept for the transfer of chemical samples based on a fleet (or swarm) of mobile mini-robots developed locally and designed around the typical shape of the sample. For localization, they are equipped with LIDAR IMU and wheel encoders. We plan to have about 10 of these mini-robots running in a track suspended above the scientific instruments. The track and the robots already physically exist. We also already implemented the ROS Navigation Stack. The transfer between the scientific instruments and the mini-robots in their track will be handled by 6-axis cobots (Universal Robots 5e or 10e) assisted by a set of industrial cameras for precise positioning. Each 6-axis cobot will be considered as a station. The fleet of mobile mini-robots will be controlled within the track via the ROS framework by the laboratory scheduler and should be autonomous to systematically assign a free mini-robot to a requested station or to deliver a sample.

OSS. Previous ROS knowledge is highly desirable

Deliverables:

• a. To make a scientific literature review for fleet control
• b. To develop, within the framework of the ROS, a complete control system for a fleet of 10 mobile
• mini-robots, 10 induction charging bases and 5 Universal Robots stations.
• c. To establish a full simulation (ideally in Gazebo) based on the existing track form using the
• control software developed in b.
• d. If time permits, start the actual implementation of the developed control system in the Swiss
• CAT+ environment.

Contact:
Vincenzo Scamarcio ([email protected])

Developing a complete solution for robotic chemical commercial bottles manipulations (Thesis)

Summary:
In the Swiss CAT+ project, we are developing a robotic module to standardise chemical solids from different types of commercial bottles up to 2 kg into homogeneously shaped 20 ml screw-capped plastic containers. The first step in this standardisation is the gripping of all commercial chemical bottles by a 6-axis cobot (type Universal Robot UR3e or UR5e). These bottles can have various shapes (cylindrical, round, square) and materials (glass, hard plastic, soft plastic such as PET) and weights ranging from 10 g to 2 kg depending on the density of the chemical product. Once the bottles have been seized, they must be kept stable and secured for subsequent operations such as pouring the contents into a standard container or allowing a sampler to enter the bottle and take a chemical sample.

Deliverables:

• a. To review the scientific literature in the field of robotic safe gripping of variable shapes and
• materials.
• b. To review the existing solutions available on the market.
• c. Select and acquire the most suitable tool.
• d. If no convincing solution exists, design and develop a suitable prototype gripper.
• e. Program the 6-axis UR cobot to work with the chosen gripper.
• f. Validate through a series of reproducibility tests with different commercial bottles, the quality of
• the gripping (systematicity, stability in case of robot movements with different accelerations…).

Contact:
Vincenzo Scamarcio ([email protected])

[closed] Cartesian Robots in a wet lab: dip and spray (Semester)

Summary:
The limit of what is achievable in a laboratory experiment is frequently defined by the researcher’s willingness to spend time on a repetitive task. Robots are more efficient than humans in this area, therefore, we want to modify two different cartesian robots to create layer-by-layer films. The first robot will perform the ‘dipping method’ (lab-scale substrates, up to 20 layers) (Already done in this semester) the latter will automate the ‘spray method’ (up to 200 layers) (To be done in this semester, focus of the project). The final objective is to find the optimal parameters to improve the film thickness error, found by using some characterization techniques

Workload (split of type of work): 

• 40% Hardware design and construction
• 40% Control of system
• 20% Characterization + Optimization

Contact:
Vincenzo Scamarcio ([email protected])

How well can a robot handshake?  (Thesis x2 students most likely)

Summary:
The project is to investigate, to what extent can a robot fool a human in a handshaking task; i.e.: a Turing test for robotic handshaking. There are many layers and directions to this project, and the specific details of the project will be discussed based on the student’s interests expertise. Here are some directions. 

– Make a bio-inspired passive hand. By covering it with multiple layers (thin gloves, ski gloves, etc), at what point will humans be able to not detect the hand is a robot or human?

– How can we understand (and replicate in the future) human-human handshaking? With a passive hand with a large array of tactile sensors, can we obtain “sensor synergies” from the handshaking motion?

– What does the passive stiffness of the wrist play a role to convey the “human-ness” in hand-shaking tasks? 

– If we add active tendons (motorise some aspects of the hand), to what extent does twitching or slight gripping convey the “human-ness” in the hand-shaking task?

Workload:
– Fabrication of hand/sensors: 60%
– Developing test methods and experimental protocols for human tests: 10%
– Conducting the aforementioned experiments: 30%

Contact:
Kai ([email protected])

Sensorizing the environment – Developing physical twins for robotic manipulation (Thesis)

Summary:
Robotic manipulation at its heart is the interaction between the robot and the environment. To assess the quality of manipulation and train robotic manipulators in real life, a sensorized “object” which can “feel” how the robot interacts with it can be a useful tool. This project aims to lay the groundwork on creating sensorized “objects” with abilities (including but not limited to) such as: localization of contact, detection of forces, orientation and acceleration measurement, etc. Specific tasks of what this environment will be is to be discussed. Current ideas include a “cube” which the robot can manipulate in-hand and handling of tools such as kitchen tools. 

Some links for related technologies: 
https://arxiv.org/pdf/1803.00628.pdf
https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9762135
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=9706272

The project requires a multidisciplinary skillest – mechanical design, fabrication, microcontrollers, coding and analysis.

Workload:
– Exploration of technologies to develop the physical twin: 40%
– Developing and (some)characterisation of the twin: 60%

Contact:
Kai ([email protected])

(Closed) Tactile Sensor development (Thesis/Semester)

Summary: 
Sensing technologies are essential to develop intelligent robots that can interact and respond to the environment. Commercial sensors usual are engineered well to have reliable, repeatable, and often precise measurements, but can be difficult to integrate into novel hardware as they are constrained by their geometry or other properties set by the manufacturer. In this project, there are two directions. 

One is to investigate and explore a variety of sensing technologies which can be developed in house, mostly focusing on tactile sensing. This is an exploratory project, with the aim for the student to try different technologies to see their feasibilities and pros/cons for future robotic implementation. Some starting points include: EIT tactile skins, hydrogel tactile skin, re-implementation of ReSkin (tactile sensing through Hall effect sensors), developing conductive silicone, and combinations of the aforementioned.

Another is to apply existing (or possibly new) tactile sensing technologies explored in the lab to apply them on robotic hands. This would include a lot of trial and error in prototyping the correct shape and manufacturing technique for the application, but also ensuring stability and simplicity in the design such that the sensor can robustly function while the robot is in operation. Design of PCB modules to read the sensors is also another necessary step, although not the priority. 

Prior experience with mechatronics prototyping is recommended but not strictly necessary. 

Workload: 
– Exploration of options “out there”: 20%
– Testing and prototyping different sensors: 60%
– Combining (or rejecting) certain sensors to finalise on a design: 20%

Contact:
Kai ([email protected])

(Closed) RoboCup @Home MAKE Project (Semester)

Summary: 
RoboCup @Home is a MAKE Project where we aim to build a home service robot to participate in the international robotics competition: RoboCup. The aim of this project is to work on the software systems of the robot. Software tasks will be around navigation and behavioural tree of the robot.

In both cases, the student should have some level prior experience in relevant fields, as the project requires practical knowledge to integrate state of the art technologies. Specific project directions will be discussed with each student based on their interest and past experience. 

Workload: 
– 80% technical development / robotic integration 
– 20% decision making and planning of the project with other team members

Contact:
Kai ([email protected])

Worming up: Ascending a taut tether with a soft robot to capture atmospheric measurements

Summary:
A kite can be used to capture atmospheric measurements for environmental monitoring. It is of interest to capture data along its tether. Despite their simple design cable stockings can hold large loads. In our lab, we are building tendon-driven soft robot structures which can deform under load and are actuated. These soft robots follow a similar design pattern to cable stockings. The goal of this project is to modify our soft robots to ascend on a taut tether. The robot will then be evaluated in comparison to a conventional wheel based robotic line ascender. The student should have prior experience in 3D printing and rapid prototyping. 

Workload:
– 15% literature review
– 30% design of soft robotic line ascender
– 35% fabrication of soft robotic line ascender
– 20% evaluation of soft robotic line ascender

Contact:
Max Polzin ([email protected])

Towards Spider-Robot

Summary: 
Aerial robots have advantageous locomotion in free space and do not require to interact with their environment while moving. However, this freedom of motion comes at a reduced locomotion efficiency (with typical flight times in the range of minutes). Ground vehicles on the other side are continuously interacting with their environment requiring advanced perception and controllers to move stable. In this project, we explore hybrid locomotion of a tethered robot which exploits swinging dynamics to preserve energy while moving and limited interactions with its environment, e.g. when forming new attachments. The student should have prior experience in rapid, mechanical prototyping and fabrication.

Workload:
– 10 % literature review
– 20 % narrowing down scenario
– 35% prototype concepts
– 35% evaluation of implemented concepts

Contact:
Max Polzin ([email protected])

Design of a power system for tethered robots

Summary:
In our lab, we are developing robots that are intended to operate in extremely harsh conditions, e.g. on glaciers, caves or tropical forests, where conventional operation modes (connectivity, power, locomotion) face severe drawbacks. Virtually all robots are tethered at one point in their life, be it for debugging or extended duration experiments. Thus tethers are vital and therefore our robot systems are continuously connected through tethers by design. A tether can provide communication, power and safety. The student will design a tether management system which meets these criteria. The tether management system will be deployed to a dangling, rappelling robot intended to explore subglacial cavities. The student should have prior experience with power electronics, communication links and mechanical design.

Workload: 
– 30 % research on available components and component selection, e.g. power, communication
– 40% design of sensorized, actuated winch mechanism
– 30% characterization and evaluation of mechanism

Contact:
Max Polzin ([email protected])

Anchor point estimation for actuated, dangling robots

Summary:
A tethered, rappelling bicopter can be used to explore unknown, complex environments e.g. glacial crevasses or tree canopies. Its tether can provide a communication link, power and safety. Depending on its past trajectory the tether can form intermediate anchor points in the environment. These anchor points influence the dynamic behaviour of the robot. From the observed dynamic behaviour of the robot, i.e. its swinging motion, we can estimate the location of the last intermediate anchor point. With an estimate of the last intermediate anchor point, it is feasible to plan a trajectory to successfully detach from said anchor point. The student will develop a simple simulation to capture the described dynamics of the system. Further, the student develops an algorithm to estimate the tether length below the last anchor point for our tethered bicopter. The develop algorithm is evaluated in simulation and on the real robot.

Workload: 
– 30% develop simulation
– 40% develop/test algorithm in simulation
– 30% implement algorithm on real system 

Contact:
Max Polzin ([email protected])

Learning ROS with ROSbloX

Summary:
The Robot Operating System (ROS2) is becoming the de facto standard in programming complex robotic systems. However, getting started with ROS (in Python or CPlusPlus) can be challenging and overwhelming, particularly when starting with no prior knowledge and learning self-paced. In our Lab, we have developed ROSbloX to ease the entrance into the ROS world for students with no prior knowledge in ROS. In this project, the student will evaluate the quality of our available ROSbloX, e.g. for Lidars, cameras, GNSS systems, IMUs. Further, the student will define metrics to assess how ROSbloX simplify the introduction of novel students to ROS. Prior knowledge of ROS2, Docker, Linux, Single Board Computers is of advantage. You get the chance to take over responsibility for an open source project and to continuously contribute to its future development. 

Workload: 
– 20% understand problems faced by students when working with ROS
– 15% familiarise with existing ROSbloX
– 25% define metrics to assess qualtiy of ROSbloX
– 25% data collection through survey to apply developed metrics
– 15% improve/advance idea behind ROSbloX

Contact:
Max Polzin ([email protected])

Advanced control of a rappelling, tethered bicopter

Summary:
Tethering robots can be advantageous in numerous scenarios. A tether can provide a reliable, high-bandwidth communication link, safety and power. A rappelling, tethered bicopter has been designed to retrieve ice samples in a glacier. The platform is self-stabilising. However, depending on the flown path, the stabilisation can be sped-up by developing and implementing novel control algorithms. The student will analyse the system and implement a simulation to capture the observed behaviour. Afterwards, the student will come up with novel control strategies to stabilise the robot faster. The control strategies are evaluated in simulation and on the real robot. 

Workload: 
– 25% understanding dynamics of the system
– 25% implementing simulation to capture system dynamics
– 25% developing novel control algorithms
– 25% implementing novel control algorithms in simulation and on real robot

Contact:
Max Polzin ([email protected])

Visual data logging in field robotics

Summary: 
Many field robots are equipped with either stereo or monocular cameras. Often processing their image streams (rectifying, running machine learning inference, logging) poses the highest load on a robot’s onboard computer. The bandwidth of onboard computers is usually limited and enough resources must be available for mission-critical tasks. In this project, several image processing pipelines are implemented and benchmarked to select an optimal pipeline to log visual data (either onboard or remote). The image processing pipelines of particular interest are Nvidia’s hardware accelerated NITROS image processing pipeline, the default CPU-only ROS2 image processing pipeline, and finally the ZED Stereo camera’s pipeline implemented in the ZED camera’s SDK. The student should have prior experience working with ROS2, Docker, Linux, Single Board Computers (ideally Nvidia Jetson).

Workload: 
– 20% selection of (feasible) image processing pipelines
– 40% implementation of image processing/logging pipelines
– 40% implementation of benchmarking metrics 

Contact:
Max Polzin ([email protected])

Environmental Sensing with a Soft Manipulator on a Robotic Line Ascender 

Summary: 
Biodiversity is a measure of our ecosystem’s health. Particularly in mountainous regions, the migration of low-altitude species to higher altitudes is an indicator of the speed of a changing climate. Monitoring biodiversity, particularly studying smaller species, e.g. insects, and plants is meticulous work. Robots provide a means to automate these tasks partially. However, the turbulence of thrusters prevents the application of uncrewed aerial vehicles. In this project, the student builds up on previous work in our lab to expand our tethered robot platform with a soft manipulator such that it can measure biodiversity in our backyard. This is a robotics project. It involves designing the hardware, programming and testing it.

Workload: 
– 20% Exploring available concepts
– 30% Hardware design and implementation
– 30% Software and control algorithm design and implementation
– 20% Experiments

Contact:
Max Polzin ([email protected])

[closed] Optimal stiffening of a fish tail (Thesis preferred)

Summary:
Being able to actively control the stiffness of a soft structure will allow us to build more capable soft robots, in particular for bioinspired underwater robots that harvest maneuvering energy from the fluid environment. We have previously created “pouches” that we can insert into silicone structures that can stiffen them in different ways. The goal of the project would be to find the best design for these pouches so we can efficiently stiffen or soften a fishtail on demand. The tail should be implemented on robotic hardware for testing the capabilities.

Workload:
– Design and fabrication: 40%
– Analysis and characterization of stiffening: 60%

Contact: 
Nana Obayashi ([email protected])

[closed] Design and sensorization of a robotic sea turtle (Thesis/Semester) 

Summary:
Development of sensory and control systems for an underwater swimming sea-turtle robot.  This includes imaging, water quality and other sensory feedback.  We will also explore how we can use this information to enable autonomous exploration. 

Workload:
75% experimental (design, fabrication, experimental testing), 25% control/simulation

Contact: 
Nana Obayashi ([email protected])

[closed] Evolutionary and regenerative exploration of falling paper morphology (Thesis/Semester)

Summary: Robotic investigation of how passive structure shapes or morphology (such as helicopter seeds) came to be. We will look into evolutionary and regenerative algorithms to reverse engineer how these structures were optimized by nature. Large scale robotic experimentation will also be used. 

Workload:
50% experimental (design, fabrication, experimental testing), 50% optimization, learning, algorithm development

Contact: 
Nana Obayashi ([email protected])

[closed] Robotic art in the fluid environment (Thesis/Semester)

Summary:
Is there a way to emulate emotion using robots in the fluid environment? An example (although not in fluids) would be to investigate reactions from robotic arm by changing the impedance of the arm based on human pose in dancing. This is an exploratory project idea. Please contact if you are interested!

Workload:
10% Literature review, 50% experimental (design, fabrication, experimental testing), 40% optimization, learning, algorithm development

Contact: 
Nana Obayashi ([email protected])

[closed] Robotic phenotyping using Farmbot for co-optimization of crop environmental impact and nutritional value (Thesis)

Summary:
Climate change, ever-increasing population and imbalanced dietary requirements poses a significant challenge for conventional crops and agricultural methods.The aim of the project is to set up the cartesian Farmbot as robotic phenotyping system and develop full physical automation of seeding, environmental control, harvesting and the assessment of the nutritional value of the crops using the cartesian ‘farm bot’. The data generated will be used to have efficient growing methods.

The project requires multidisciplinary skills – mechanical design, fabrication, microcontrollers, coding and data analysis.

Workload:
30% Setup hardware for data acquisition
40% Software algorithms and sensor calibration
30% Data analysis

Contact:
Shiv Katiyar ([email protected])

[closed] Robotic platform for investigation of flocculation in beverages (Thesis/Semester)

Summary: 
During the beverage formulation process it is often investigated how developed food product (e.g. coffee) interacts with other beverages like milk. With variation in beverages (e.g. dairy or plant based milk) or even in their formulation, significantly different behaviour in terms of colour, flocculation, sedimentation or foaming of the resulting mixture is detected. Acquired information is used for the objective assessment of the resulting mixture but more importantly as a metric for the further optimization of the constituent ingredients (milk, coffee etc.).

Goal of this project is to develop a robotic system which is able to repeatedly combine two manually prepared beverages, mix them if necessary and keep the temperature constant. Resulting beverage will be then assessed using computer vision methods primarily for  the flocculation and phase separation. Assessment of sedimentation, foaming, color or other relevant metric will be explored if possible.

The robot design will be based on the frame of the 3D printer and with the gripper which needs to be designed, such system can repeatability and reliably perform pick, place, pour and mix tasks. Integration of other necessary equipment (e.g. temperature control or camera) has to be done within this constrained environment.

There are no requirements considering computer vision methods, but it is preferred that assessment should be performed using Python and preferably OpenCV library, but other suggestions are most welcome.

Workload: 
– 40% design and building of robotic system
– 60% implementation of image processing pipelines

Contact:
Stefan Ilic ([email protected])

Robotic Simulation Toolkit for Education (Thesis/Semester)

Summary:

Educational Robotics helps students of all ages familiarise and expand their knowledge of robotics and programming. while learning other cognitive skills.Given the physical form of the robot and its control system, it is particularly promising to make robotic systems adaptable to tasks and the environment. This project will aim to use robotic simulation for morphology and computational thinking education. 

https://gears.aposteriori.com.sg/

Workload: 

• 20% Explore education robotic elements 
• 30% Design learning activities 
• 50% Experiment and analysis data

 Contact person: Alan Wu ([email protected])

Robotic Education with AI guidance (Thesis/Semester)

Summary:

Educational Robotics helps students of all ages familiarise and expand their knowledge of robotics and programming. while learning other cognitive skills. This project will aim to use AI to give guidance and suggestions to students instead of human instructors. The task for experiment subjects will be building a robotic gripper. The AI can give advice to experiment subjects when they are facing problems. We compare the learning performance between human assistants and AI assistance.

Workload: 

• 20% Explore learning outcome evaluation methods 
• 30% Refine learning activities and define evaluation process
• 50% Experiment and analysis data

 Contact person: Alan Wu ([email protected])

Animal behavior to robot design: a methodological framework for computational design of robot structures from animal synergies

Summary: While biological systems display an incredible diversity in behaviors and motion capabilities, it has been shown that for most tasks, only a subspace of the possible poses is used. We propose to exploit these emerging principles from animal motor control as inspiration to guide robot design. In this thesis you will start by extracting such synergistic motions from animal videos and data-bases. Hence, you will develop a computational design method to translate the emerging synergies into precise robot specifications. Finally, you will transform the simulation results into the design of a fully fledged biomimetic robot. In particular, the thesis will be focused on the design, manufacturing and control of a biomimetic cheetah with a reduced set of actuators. Potential publication of results in top Robotics venues.

For reference see link: https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=8373731 

Workload:

• 10% Extraction of existing synergies from videos and open-source databases of animal motion (Computer vision, Inverse Kinematic and order-reduction methods)
• 10% Development of a computational method for mechanical design specifications (Evolutionary algorithms)
• 80% Design and testing of a cheetah robot (Mechatronic design)

Contact person: Francesco Stella

IMUblock: A modular and ready-to-go solution to the pose estimation problem of soft robots

Summary: Due to the inherent compliance, soft robots are able to deform in infinite degrees of freedom. Moreover, the limited availability of physically compatible sensors, makes the shape estimation of soft robotic systems extremely challenging. In this project you will develop the modular and scalable solution – IMUblock – to this problem by leveraging IMU technology and discretized models of soft robots. The IMUblock will then be evaluated experimentally on a growing, extendable continuum soft manipulator. Potential publication of results in top Robotics venues.

For reference see link: https://link.springer.com/chapter/10.1007/978-3-030-71151-1_48 

Workload:

• 50% Design of a modular, wireless IMU sensor (Rapid Prototyping, Electronic design)
• 20% Embed a filter based on reduced orders of soft robots (PCC) in the IMUblock (C++, Python, ROS) 
• 30% Test the novel sensor solution on existing soft manipulators.  

Contact person: Francesco Stella, Max Polzin

Soft Robotic Toolkit for Education

Summary: Educational Robotics helps students of all ages familiarize and expand their knowledge of robotics and programming. while learning other cognitive skills. There are many modular educational robots in the market but there is a missing piece, soft component. The aim of this project is to develop a toolkit with several soft components such as soft actuators.

Workload:

• 20% Explore soft robotic elements
• 30% Design several modular soft robotic components
• 30% Manufacture
• 20% Testing

Contact person: Yi-Shiun Wu

Cartesian Robot with custom gripper and computer vision

Summary: Laboratories are ripe for automation as they are semi-structured.  Cartesian robots offer precision motions, yet typically rigid end effectors are used.  By combining a cartesian robot with a compliant gripper and computer vision and modular equipment, we want to explore the complexity of tasks that can be accomplished.

Workload:

• 40% Hardware design and construction
• 20% Interfacing electronics
• 40% Control of system and computer vision

Contact person: Stefan Ilic, Vincenzo Scamarcio

Cartesian Robots in a wet lab: dip and spray

Summary: The limit of what is achievable in a laboratory experiment is frequently defined by the researcher’s willingness to spend time on a repetitive task. Robots are more efficient than humans in this area, therefore, we want to modify two different cartesian robots to create layer-by-layer films. The first robot will perform the ‘dipping method’ (lab-scale substrates, up to 20 layers) the latter will automate the ‘spray method’ (up to 200 layers). The final objective is to find the optimal parameters to improve the film thickness error, found by using some characterization techniques.

Workload:

• 40% Hardware design and construction
• 40% Control of system
• 20% Characterization + Optimization

Contact person: Vincenzo Scamarcio

Fabrication and characterization of a variable stiffness soft structure

Summary: Being able to actively control the stiffness of a soft structure will allow us to build more capable soft robots, in particular for bioinspired underwater robots that harvest maneuvering energy from the fluid environment. In this project, you will explore techniques that are used to achieve variable stiffness for soft silicone structures, such as  an underwater soft tentacle. The aim for this project is to investigate mechanisms for high stiffness variation, characterizing the structural behavior, and if successful, implement on robotic hardware for maneuverability demonstrations. 

Related references: 

• https://link.springer.com/chapter/10.1007/978-3-030-25332-5_6
• https://www.liebertpub.com/doi/full/10.1089/soro.2019.0034

Workload: 20% mechanical design, 50% design iteration, exploration, and fabrication, 30% prototyping with microcontrollers

Contact person: Nana Obayashi

Exploration of emergent limit cycles of floating objects in water

Summary: To design better robots that exploit the fluidic environment such as water, it is important to study the bio-inspired interactions between morphologies. Non-contact manipulation is seen in nature, for example in tool usages by marine animals. This project involves the development of an experimental setup for manipulating objects in water using soft structures and exploring the emergent behaviors of these objects. It was previously seen that certain “limit cycles” can be observed from non-contact manipulation of floating objects using a structure similar to a soft fin. This will be a highly exploratory project on characterizing the emergent limit cycles to help us understand how we can design robots that exploit the fluid environment for a given task. 

Related references: 

• https://www.nature.com/articles/s41598-021-04204-9
• https://pubs.acs.org/doi/10.1021/acsami.1c13551

Workload: 20% mechanical design, 60% design iteration, exploration, and fabrication, 20% prototyping with microcontrollers

Contact person: Nana Obayashi

Robotic phenotyping for co-optimization of crop environmental impact and nutritional value

Summary: Food production and agriculture are facing enormous challenges due to factors such as climate change, ever-increasing population and imbalanced dietary requirements. This poses a significant challenge for conventional crops and agricultural methods. Agri-robotics opens the potential for mass direct in-field phenotyping of crops under true farm conditions. The aim of the project is to set up the cartesian farmbot as a robotic phenotyping system and develop full physical automation of seeding, environmental control, harvesting and the assessment of the nutritional value of the crops using the cartesian ‘farm bot’.

The project requires multidisciplinary skills – mechanical design, fabrication, microcontrollers, coding and data analysis.

Workload:

• 30% Setup hardware for data acquisition
• 40% Software algorithms and sensor calibration
• 30% Data analysis

Contact person: Shiv Katiyar

Topographic terrAin Modeling for Agricultural Robots (TAMAR)

Type: Semester Project
Summary: Robust rough terrain navigation of uncrewed ground vehicles enables their usage for novel agricultural applications, especially in environments where the usage of traditional farming machinery is impossible, e.g. autonomous harvesting in steep vineyards or orchards. However, testing uncrewed ground vehicles (UGV) on natural slopes is a cumbersome and potentially dangerous task. In this project, a simulation of the slopes of a Swiss vineyard is developed to evaluate novel designs for agricultural robots. The simulation a) imports topological information provided by the Swiss Federal Office of Topography and b) includes a data-driven traction model for UGVs maneuvering on steep terrain. The accuracy of the developed simulation is validated experimentally.

Keywords: Terrain simulation, data-driven modeling, mapping
Workload: 15% literature research, 30% experiments, 55% simulation development
Contact Person: Max Polzin

Topologically constrained planning for robots operating in precipitous environments (TOP-ROPE)

Type: Semester Project / Master Thesis
Summary: Tethered rappelling uncrewed ground vehicles (UGV) have potential applications in future agricultural systems, especially where the terrain prevents the usage of traditional systems, e.g. autonomous harvesting in steep vineyards or orchards. While a tether connected to a UGV expands its operating area to steep slopes, it imposes additional constraints when planning feasible paths for said UGV. In this project, novel path planning algorithms which consider the constraints imposed by a tether connected to a UGV are developed.

Keywords: Path planning, robotic navigation
Workload: 20% literature research, 35% simulation experiments, 45% thinking
Contact Person: Max Polzin

Anchor deployment and Collection for Tethered Robots Exploring Steep Slopes (ACTRESS)

Type: Semester Project / Master Thesis
Summary: Uncrewed ground vehicles (UGV) have potential applications in future agricultural systems, e.g. automatic harvesting, phenotyping, weed control and pest management. Agricultural terrain is generally challenging to maneuver in, especially when driving on slopes such as in steep vineyards and orchards. A UGV can be equipped with a winch and tether to a) expand its maneuverable space and b) to improve its recovery capabilities from catastrophic failures. The maneuverable space of a UGV which is connected to an anchor point by a tether differs from the maneuverable space of an untethered one. In this project a dynamic anchoring system is developed to be deployed and retracted autonomously by a UGV, to maximize its maneuvering capabilities.

Keywords: Mechatronics prototyping
Workload: 10% literature research, 50% mechanical design, 40% experiments
Contact Person: Max Polzin

Automated Robot Testing In Steep Terrain (ARTIST)

Type: Semester Project / Master Thesis
Summary: Improving the maneuverability of uncrewed ground vehicles (UGV) on natural slopes enables novel applications of UGVs in agriculture, e.g. autonomous harvesting in vineyards or orchards. However, testing UGVs and novel algorithms on natural slopes is a cumbersome and potentially dangerous task. Thus, a platform to simulate driving on changing slopes with different surface textures is developed. The platform alters the direction of the gravitational force acting on the systems under test. Altering the direction of the gravitational force requires lifting and pulling respectively pushing the systems under test with a constant force. In this project, a system is designed and implemented to exert a varying force on a UGV.

Keywords: Dynamic systems, motor control
Workload: 35% system design, 40% experiments, 25% control
Contact Person: Max Polzin

Fabrication and characterisation of a soft sensorized fingertip

Type: Semester Project
Summary: Tactile sensing is key to achieving robust dexterous manipulation when using robot hands. Previous methods of tactile sensing include using cameras to measure the deformation of soft material and attaching integrated PCB electronics on fingertip. In this project, you will explore methods of using 3D printable soft conductive filament to manufacture a fingertip sensor which can be used directly from the print. The aim for this project is to investigate different morphologies of 3D printed sensors (provided with a few starting points), measure/understand their response, and if successful implement it on a robot finger.

Keywords: Soft sensors, mechatronics prototyping
Workload: 15% mechanical design, 50% design iteration, exploration, and fabrication, 30% electronics
Contact Person: Kai Junge

Topology optimization for soft electronic skins

Type: Semester Project
Summary: Soft electronic skins (E-skins) capable of tactile pressure sensing have the potential to endow robotic systems with many of the same somatosensory properties of natural human skin. In this project, you will first develop computational methods to optimize the relation between topology and sensitivity for a grid of soft pressure sensors. Hence, the simulation results will be brought to real life, evaluating manufacturing possibilities and limitations arising from 3D printing and casting techniques. Finally, the design performance will be evaluated on a specific application such as safe human-robot interaction, dexterous manipulation, or guided rehabilitation.

Keywords: Soft sensors, optimal design, mechatronics prototyping
Workload: 60% simulation and optimization , 30% system design and experiments, 10% proof of concept application
Contact Person: Francesco Stella

Low-cost, sensorized, finger prosthetics for developing countries

Type: Semester Project
Summary: The World Health Organization estimates that, in the developing world, there are 40 million amputees, and only 5% of them have access to any form of prosthetic care. Several initiatives, by academic and non-profit organizations community, have already tried to leverage rapid prototyping techniques to propose a solution to this problem. In this project you will contribute to this collective knowledge, developing a parametric design of a finger prosthetic with embedded tactile sensors that can be quickly personalized and 3d-printed. The design will then be quantitatively tested and validated. Finally, if the design process is successful, the design will be sent and evaluated by amputees in Mali. 
For reference: Limbs International, “Why Limbs,” [Online]. Available: https://www.limbsinternational.org/why-limbs.php. 2015.

Keywords: Prosthetics, embedded sensors, prototyping
Workload: 80% design , 20% experimental validation
Contact Person: Francesco Stella

IMU sensor fusion for soft robotics

Type: Semester Project / Master Thesis
Summary: Due to the inherent compliance, soft robots are able to deform in infinite degrees of freedom. Moreover, the limited availability of physically compatible sensors, makes the shape estimation of soft robotic systems extremely challenging. In this project you will combine discretized models of soft robots, such as the constant curvature model, with state of the art sensor fusion algorithms. The algorithm will then be evaluated experimentally on a growing, extendable 3D printed continuum body structure with embedded IMUs.
For reference, see link. 

Keywords: Soft robotics, Kalman filtering, IMUs
Workload: 80% algorithm design , 20% experimental validation
Contact Person: Francesco Stella

Control optimization and steering of underwater swimmers

Type: Semester Project
Summary: There are several simple soft underwater swimmer platforms that are of interest. One focus of the project will be the optimization of actuation control for the underwater swimmers. We will also investigate different methods of controlling the trajectory of the swimmer (eg. drag-based, differential thrust, etc) and explore how we can use this to enable autonomous exploration. 

Keywords: Soft robotics, optimization, control, computer vision, electronics
Workload: 60% control/simulation, 40% experimental (design, fabrication, experimental testing)
Contact Person: Nana Obayashi

Underwater object manipulation with air bubbles

Type: Semester Project / Master Thesis
Summary: Development of experimental setup for manipulating objects/debris underwater leveraging air bubbles or soft structures. We take inspiration from sea creatures, such as dolphins and whales that use “bubble nets” to capture prey. We will also investigate methods for simulating these complex solid-fluid interactions.

Keywords: Manipulation, computer vision, prototyping, electronics, sensing, control, simulation
Workload: 75% experimental (design, fabrication, experimental testing), 25% control/simulation
Contact Person: Nana Obayashi

Agricultural Robotics: Robotic optimization of plant growing conditions

Type: Semester Project / Master Thesis
Summary: Using a cartesian robot (farmbot) we want to develop feedback control based on computer vision to assess plant growth and optimize conditions through running robotic growing experiments. Can we make plants grow faster, and can be predict their performance from early stage growth?

Keywords: Feedback control, optimization, agricultural robotics, computer vision
Workload: 75% experimental (design, fabrication, experimental testing), 25% control/simulation
Contact Person: Shiv Katiyar

‘Robotic Scientist’ Intelligent automation of experimental analysis (titration)

Type: Semester Project / Master Thesis
Summary: Develop the hardware, control and computer vision to intelligently automate titration experiments.  Many experiments require the use of burette to perform titration, where the colour change is used to indicate when the experiment should be terminated.  Currently this is highly manual and there are significant problems with repeatability and reliability.  Automating this process will be highly impactful, and also enable more intelligent methods to be explored.

Keywords: Mechatronics prototyping, computer vision, control
Workload: 60% design and implementation , 20% computer vision, 20% control
Contact Person: Stefan Ilic

Mechanism for Growing Robots

Type: Master Thesis
Summary: Biological systems show a significant ability to change and alter their structure through growth.  A number of mechanisms for growth have been developed for robotic structures.  Here, we want to develop mechanisms that allow for plant-inspired growth – i.e. structures that can ‘grow’ and subdivide, to allow us to create large scale structures through growth. 

Keywords: Mechatronics prototyping, control
Workload: 80% design and implementation , 20% control
Contact Person: Shiv Katiyar

Design and fabrication of sensorized bioinspired robots hands (Semester/Full Masters Project)

Summary: Starting for biological inspiration, use novel 3D printing technologies and casting to rapid fabricate robotic hands with incorporated soft sensors. Investigate the design optimization of the sensor placement, and utilize multi-body, multi-material FEA based simulators to optimize control.

Keywords: Bio-inspired robotics, soft sensors, control, design optimization.
Workload: 75% experimental (design, fabrication, experimental testing), 25% control/simulation

Agricultural Robotics: Robotic optimization of plant growing conditions (Semester/Full Masters Project)

Summary: Using a cartesian robot (farmbot) we want to develop feedback control based on computer vision to assess plant growth and optimize conditions through running robotic growing experiments. Can we make plants grow faster, and can be predict their performance from early stage growth? 

Keywords: Feedback control, optimization, agricultural robotics, computer vision
Workload: 75% experimental (design, fabrication, experimental testing), 25% control/simulation

Optimization of Robotic Sea Turtle Flipper (Morphology and Control) (Semester/Full Masters Project)

Summary: The development of soft swimming robots is particularly challenging due to the interactions of the soft body with the fluid.  Here, we want to leverage modelling techniques and data-driven methods to optimize the design of the flippers for a sea-turtle and also the controllers. 

Keywords: Manipulation, computer vision, sensing, feedback, optimization, learning
Workload: 75% experimental (design, fabrication, experimental testing), 25% control/simulation

Sensorization of a Robotic Sea Turtle (Semester/Full Masters Project)

Summary: Development of sensory and control systems for an underwater swimming sea-turtle robot.  This includes imaging, water quality and other sensory feedback.  We will also explore how we can use this information to enable autonomous exploration. 

Keywords: Sensing, control, computer vision, electronics.
Workload: 75% experimental (design, fabrication, experimental testing), 25% control/simulation

Evolutionary Design of Soft Sensors (Semester/Full Masters Project)

Summary: The design of soft sensors for tactile perception significantly affects their performance.  Using new fabrication techniques for soft sensors, we will develop evolutionary approaches to optimize the design for different tasks.

Keywords: Sensing, control, optimization, 3D printing, fabrication
Workload: 50% experimental (design, fabrication, experimental testing), 50% optimization, learning, algorithm development

Wearable Sensorized Gloves (Semester/Full Masters Project)

Summary: utilizing our novel fluidic soft sensors and knitted sensorized sensors, develop gloves which perform on-board real-time sensor processing.  In addition incorporate activ efunctionality – e.g. control adhesion, or actuation 

Keywords: Sensing, control, optimization,  fabrication
Workload: 75% experimental (design, fabrication, experimental testing), 25% optimization, learning, algorithm development