This page contains the database of possible research projects for master and bachelor students in the Biorobotics Laboratory (BioRob). Visiting students are also welcome to join BioRob, but it should be noted that no funding is offered for those projects (see https://biorob.epfl.ch/students/ for instructions). To enroll for a project, please directly contact one of the assistants (directly in his/her office, by phone or by mail). Spontaneous propositions for projects are also welcome, if they are related to the research topics of BioRob, see the BioRob Research pages and the results of previous student projects.
To limit the list to the projects matching a given keyword, click on it.
Efficient swimmers rely on sensing the local changes in surrounding waters and use them to their advantage. For example, fishes swimming in water can sense local deformations generated by the vortices generate by surrounding fishes and swim in school formations to reduce the energetic cost. What are the key components of these behaviors? In this project, you will study this problem in simulation. Previous simulations of movement of body in fluid consider overly simplified fluid models, that does not capture the fluid dynamics, or simplified body models. We recently developed a new fluid-body simulator that can simulate the dynamics of complex rigid body geometries, similar to that of a real fishes and robots, and the dynamics of the fluid. This allows the study of collective behaviors like schooling and the incorporation of water sensing. The main goal of this project is to continue the development of the fluid solver in PyTorch, and test the ability of the model to generate self-propelled swimming. The goals can be divided in four subgoals (in order of priority): 1. Implement an interpolation method for the body fitted meshes to compute the velocities of the bodies in the fluid solver. 2. Improve the simulator's performance by porting part of the fluid solver in C++/CUDA by writing a PyTorch extension. 3. Validate the solver based on traditional benchmark tests and particle image velocimetry data from a swimming robot, and against simpler drag based fluid models. 4. Test and refine the implementation of the forces acting from the fluid to the body.
Many quadruped robots use simple ball feet while animals usually have complex foot structures. Some studies have tried designing more complex actuated or adaptive feet for quadruped robots. However, few have systematically investigated the benefits of such feet when they are integrated into the robot, especially for the sprawling type quadrupeds. The lack of understanding also exists in animal locomotion because of the complexity and small dimensions of the structure. To start understanding the role of biomimetic foot structures, this project aims to systematically compare the performance of a salamander robot equipped with ball feet and with passive adaptive feet in both simulation and hardware experiments. A semester project student can choose either one to work on while a master project student needs to do both parts. For the simulation experiments, the student will: (1) build simplified models of the feet in our Mujoco-based simulation framework, FARMS, (2) optimize the design parameters using optimization or learning algorithms, and (3) compare the results with those using models with ball feet and with data collected in animal experiments. The student is thus required to be familiar with Python programming and optimization/learning algorithms. Students who have taken the Computational Motor Control course would also be preferred. For the hardware experiments, the student will: (1) design and manufacture the feet based on previous studies, (2) integrate the feet into our salamander robot, and (3) perform systematic tests in different environments. The student is expected to be experienced in mechanical design and manufacturing and have basic knowledge of the mechanics of materials. Students who are interested in this project shall send the following materials to the assistants: (1) resume, (2) transcript showing relevant courses and grades, and (3) other materials that can demonstrate your skills and project experience (such as videos, slides, Git repositories, etc.).
Exoskeletons have experienced an unprecedented growth in recent years and hip-targeting active devices have demonstrated their potential in assisting walking activities. Portable exoskeletons are designed to provide assistive torques while taking off the added weight, with the overall goal of increasing the endurance, reducing the energetic expenditure and increase the performance during walking. The design of exoskeletons involves the development of the sensing, the actuation, the control, and the human-robot interface. In our lab, a hip-joint active hip orthosis (“eWalk”) has been prototyped and tested in recent years. Currently, multiple projects are available to address open research questions. Does the exoskeleton reduce the effort while walking? How can we model human-exoskeleton interaction? How can we design effective controls? How can we optimize the interfaces and the control? Which movements can we assist with exoskeletons? To address these challenges, the field necessitates knowledge in biology, mechanics, electronics, physiology, informatics (programming, learning algorithms), and human-robot interaction. If you are interested in collaborating in one of these topics, please send an email to giulia.ramella@epfl.ch with (1) your CV, (2) your previous experiences that could be relevant to the project, and (3) what interests you the most about this research topic (to be discussed during the interview).
Last edited: 05/11/2024
Quadruped robotics
A small excerpt of possible projects is listed here. Highly interested students may also propose projects, or continue an existing topic.
During terrestrial locomotion, some frog species display both out-of-phase walking or in-phase hopping limb movements. It has been suggested that changes in these gaits arise to minimize energy consumptions. In this project we will explore this hypothesis by simulating the frog terrestrial locomotion using reinforcement learning. We will use a biomechanical model of the frog adopted with artificial muscles to investigate the optimal gaits for different terrain conditions (low-medium-high ground stiffness). The plantaris longus tendon has been associated with a crucial ability of the frog to store elastic energy during frog jumping. We will test this hypothesis in simulation. The goals can be divided in these subgoals (in order of priority/time): 1. Compute the inertial properties of the frog and URDF file creation 2. Train a neural network controller using reinforcement learning and design of the cost function 3. Testing the ability of the model to walk and hop in simplified scenarios
There are several quadruped robot projects available related to locomotion, jumping, and human-robot interaction, with methodologies including deep reinforcement learning, imitation learning, optimal control, and computer vision. Students who already have experience with deep learning, C++, vision, and who have worked with hardware are especially encouraged to apply. Please send Guillaume your CV, transcript, and explain your motivation on what kind of topics you would be interested in working on (more details = better!).
Recent years have shown impressive locomotion control of dynamic systems through a variety of methods, for example with optimal control (MPC), machine learning (deep reinforcement learning), and bio-inspired approaches (CPGs). Given a system for which two or more of these methods exist: how should we choose which to use at run time? Should this depend on environmental factors, i.e. the expected value of a given state? Can this help with explainability of what exactly our deep reinforcement learning policy has learned? In this project, the student will use machine learning to answer these questions, as well as integrate CPGs and MPC into the deep reinforcement learning framework. The methods will be validated on systems including quadrupeds and model cars first in simulation, with the goal of transferring the method to hardware. To apply, please email Guillaume with your motivation, CV, and briefly describe your relevant experience (i.e. with machine learning, software engineering, etc.).
Controlling vehicles at their limits of handling has significant implications from both safety and autonomous racing perspectives. For example, in icy conditions, skidding may occur unintentionally, making it desirable to safely control the vehicle back to its nominal working conditions. From a racing perspective, drivers of rally cars drift around turns while maintaining high speeds on loose gravel or dirt tracks. In this project, the student will compare several approaches for high speed, dynamic vehicle maneuvers, including NMPC with a standard dynamic bicycle model, NMPC with a dynamic bicycle model + GP residuals, NMPC with learned dynamics (i.e. a NN), and lastly a pure model-free reinforcement learning approach. All approaches will be tested in both simulation as well as on a scaled vehicle hardware platform. To apply, please email Guillaume with your motivation, CV, and briefly describe your relevant experience (i.e. with machine learning, software engineering, etc.).