Biorobotics Laboratory BioRob

Project Database

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. 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.

Amphibious robotics
Computational Neuroscience
Dynamical systems
Human-exoskeleton dynamics and control
Humanoid robotics
Miscellaneous
Mobile robotics
Modular robotics
Neuro-muscular modelling
Quadruped robotics


Amphibious robotics

627 – Waterproof camera suite for Krock 2 Amphibious Robot
Category:semester project
Keywords:C#, Linux, Mechanical Construction, Programming, Python, Robotics
Type:60% hardware, 40% software
Responsible: (MED 1 1023, phone: 31367)
Description:The Krock 2 amphibious robot is capable of sprawling-gait locomotion over land as well as swimming along the surface of water. The robot has been used for field experiments in terrestrial settings, and we are now expanding the use-cases to include aqueous and amphibious operation. As such, the cameras need to be ported to work with the dry-suit that is currently obscuring their vantage point. We envision an external attachment to go onto the robot. This project would entail mounting an external wide angle and thermal cameras that can support wireless communication over a local network as well as wired communication through water-proof connectors. The project would include fabricating a water-proof enclosure for the cameras to be mounted on the Krock 2 robot.

Last edited: 26/04/2020

Human-exoskeleton dynamics and control

625 – Development of a simulation environment for the Autonomyo exoskeleton for neuro-mucular controller simulation
Category:
Keywords:Bio-inspiration, Biped Locomotion, Control, Reflexes, Robotics, Simulator
Type:30% theory, 70% software
Responsible: (MED 1 1611, phone: 34793)
Description:The idea is to design a bio-inspired exoskeleton controller for a lower limb device called Autonomyo. The project will involve A) the development of a mechanical model of the human+exoskeleton in Opensim, the optimization of the new model and B) the development of bio-inspired control algorithm based on the strategy developed and described in https://www.frontiersin.org/articles/10.3389/fnbot.2017.00030, https://ieeexplore.ieee.org/abstract/document/7523694/, https://link.springer.com/chapter/10.1007/978-3-319-46532-6_27. The project will first involve mechanical development of the device in OpenSim followed by the training of a walking controller involving the human + the exoskeleton in transparent mode (0 torque output). The second part of the project will involve the development of a neuro muscular controller (NMC) similar to the one described in ​ https://www.frontiersin.org/articles/10.3389/fnbot.2017.00030 (a thorough description of the concepts behind NMC can be found in https://infoscience.epfl.ch/record/264212​ ) used to study the interaction of the exoskeleton and the human in loop with increasing level of weaknesses/assistance in the joint actuated by the autonmyo

Last edited: 03/03/2020

Miscellaneous

583 – Simulation and control of a robotized wheelchair able to cross difficult terrains
Category:semester project, master project (full-time)
Keywords:Control, Dynamics Model, Locomotion, Optimization, Robotics
Type:20% theory, 80% software
Responsible: (MED 1 1226, phone: 32658)
Description:This project aims at extending a dynamic simulation and locomotion controllers for a robotized wheechair able to handle difficult terrains including stairs. The project is done in collaboration with Mr Christophe Cazali who invented and patented the wheelchair. The wheelchair has 4 legs with two 1-axis motorized joints + 1 motorized wheel. This mechanism enables the wheelchair to cross obstacles like stairs while keeping the seat always horizontal and statically stable. The movement is driven by a controller using sensors like 3D camera and some position sensors. One semester project and one master project already worked on setting up the simulation model and controller in Webots/Python environment. Now the main challenges are : 1) to improve the safety of the current controller in various configurations such to be able to handle hazardous events or sensor errors, 2) to develop a 3D vision system for obstacle detection and terrain estimation (choice of sensors and development of algorithms), 3) and to finalize and optimize the mechanical design in order to maximize crossing capacity with affordable technology (optimize wheelchair geometry and choose motorization technology). The project will be done using the Webots simulator and could tackle one or more of these challenges. In the medium term, this study should lead to the creation of a prototype that could participate to some robotic competition like the Cybathlon challenge, and in the long term to a product for persons with reduced mobility. The project will therefore involve modeling work, development of locomotion controllers, and iterative exploration of the best combination of sensors, actuators, mechanical geometry, and control algorithms.

Last edited: 04/08/2020

Neuro-muscular modelling

630 – Reflex based model simulating running behavior in human locomotion
Category:semester project, master project (full-time)
Keywords:Biped Locomotion, Locomotion, Optimization, Reflexes
Type:30% theory, 70% software
Responsible: (MED 1 1611, phone: 36714)
Description:The EPFL Biorobotics laboratory is using neuromuscular simulation to better understand the mechanisms behind the neural control of human locomotion. These simulations have shown that a simple sensory driven controller is able to replicate healthy gaits stimulating reflexes based on stretch and force receptors and gained by specific constant parameters. However, CPGs based feedforward control are also able to replicate human behavior with a higher flexibility in gait modulation. More precisely, Aoi (2019) was able to switch from walking to running behavior with the anticipation of one locomotor pattern in the gait cycle. This project aims to simulate the human running behavior in a purely reflex based model applying modification in the currently used reflex based controller from Ong (2019). Previous experiences with non-convex optimizations is recommended. Experiences with C++ programming is a plus.

Last edited: 18/05/2020
629 – Shared feedback and feedforward control in human locomotion
Category:semester project, master project (full-time)
Keywords:Bio-inspiration, Biped Locomotion, Feedback, Locomotion, Optimization, Reflexes
Type:30% theory, 70% software
Responsible: (MED 1 1611, phone: 36714)
Description:The EPFL Biorobotics laboratory is using neuromuscular simulation to better understand the mechanisms behind the neural control of human locomotion. These simulations have shown that a simple sensory driven controller is able to replicate healthy gaits stimulating reflexes based on stretch and force receptors and gained by specific constant parameters. Yet, the integration of feedforward rhythmic signals through central pattern generators (CPGs) have shown a great flexibility in gait modulation. This project aims to analyse the amount of shared control between feedback and feedforward commands with a co-optimization of reflexes and CPGs parameters highlighting the differences of the shared control between distal and proximal muscles and between flexors and extensors. Previous experiences with non-convex optimizations is recommended. Experiences with C++ programming is a plus.

Last edited: 18/05/2020
628 – Reflex based controller with non-linear stretch reflex response for human locomotion
Category:semester project, master project (full-time)
Keywords:Bio-inspiration, Biped Locomotion, Control, Feedback, Locomotion, Optimization, Reflexes
Type:30% theory, 70% software
Responsible: (MED 1 1611, phone: 36714)
Description:The EPFL Biorobotics laboratory is using neuromuscular simulation to better understand the mechanisms behind the neural control of human locomotion. These simulations have shown that a simple sensory driven controller is able to replicate healthy gaits stimulating reflexes based on stretch and force receptors and gained by specific constant parameters. Current reflex based model used to replicate human locomotion are based on a linear correlation between the sensed muscle changes and the reflex stimulation provided. Although this approach was able to reproduce faithfully kinematic, dynamic and muscle activation of human walking, it is not coherent with experimental observation. This project aims to integrate the non-linear spindle response modeled by Prochazka in the reflex based model, optimize the parameters and check the efficiency of the predictive behavior for human locomotion. Previous experiences with non-convex optimizations is recommended. Experiences with C++ programming is a plus.

Last edited: 18/05/2020
626 – Personalized musculoskeletal model from 3D reconstruction
Category:
Keywords:3D, Kinematics Model, Motion Capture, Optimization, Programming, Vision
Type:30% theory, 70% software
Responsibles: (MED 1 1215, phone: 32602)
(MED 1 1611, phone: 34793)
Description:he recent developments in computer graphics and deep learning lead to high quality 3D models of objects and the human body. Among all the possible approaches for the 3D reconstruction, the most attractive method so far has been image-based modelling, either single, multiple 2D images or 3D scanning [1,2]. These approaches use 2D images as the only input of the algorithms and calibrate the object recognition and reconstruction. From the human 3D reconstruction perspective, recent advances in deep neural networks and large datasets of manually labeled images have resulted in rapid progress in 2D human “pose” estimation. The goal of the project will be to first use, test and potentially improve and existing 3D reconstruction model of human body based on 2D images [3]. In the second part of the project the goal will be to - given the pose estimation of a person - a scaling of musculoskeletal model using the information of the 3D reconstruction model will be studied. This approach requires calibrating the opensim human model while considering the findings of the 3D model, as well as the muscle attachment points and parameter settings. At the end, the validation of the method requires motion capture data of the same person to adjust the muscle dynamics and parameter settings.

Last edited: 03/03/2020
624 – Predictive model of human musculoskeletal locomotion
Category:semester project
Keywords:Biped Locomotion, Computational Neuroscience, Control, Learning, Reflexes, Simulator, Synchronization
Type:30% theory, 70% software
Responsibles: (MED 1 1611, phone: 36714)
(MED 1 1611, phone: 34793)
Description:Understanding the behavior of dynamical systems is one of the challenging topics in many fields such as engineering, finance, economics and sociology. The governing rules behind these systems often have certain structures, and follow the laws of physics. Evolution of such systems are described with ordinary or partial differential equations, with dependent or independent variables within the system. Dissecting the role of each variable for an unknown dynamical system is gaining attention due to the increase in computational power with the aim of improving the quality/robustness of control system. Data dependent methods, such as machine learning, aims at capturing the internal dynamics of such a system directly from the samples generated by the system itself. Hence, the more data generated by the system, the better the prediction. Given the importance of capturing the uncertainty of the predicted model (in order to be able to assess the quality of the prediction), probabilistic machine learning methods are promising to be useful framework. One of them called Gaussian processes will be used in this project. One of the advantages of the Gaussian processes is its applicability on medium size data sets, such as in robotics [1,2]. In this work, we will be studying the application of Gaussian processes to Biped locomotion. We will use different pre-trained biped gait patterns and also different cost functions settings to construct the predicted dynamical system of a human musculoskeletal walking patterns.

Last edited: 03/03/2020

Mobile robotics

632 – Locomotion tests on A1 Unitree quadrupedal robot
Category:semester project, master project (full-time)
Keywords:Architecture, Control, Quadruped Locomotion, Robotics
Type:20% theory, 40% hardware, 40% software
Responsibles: (MED 1 1024, phone: 37506)
(MED 1 1023, phone: 31367)
Description:Advanced project on selecting, simulating and implementing a CPG-based controller for the quadrupedal A1 robot (link below) from Unitree. The project is intended to verify the developer tools shipped with the new robot by simulating and porting a custom controller to the platform. Due to the varied tools on the platform, strong experience with Python programming, Linux platforms, and past projects with hardware is required. Experience with ROS is a strong plus. Contact supervisor Matthew Estrada with examples of past projects and a CV if interested! http://www.unitree.cc/e/action/ShowInfo.php?classid=6&id=366

Last edited: 21/07/2020
631 – Chassis design for a planetary rover (EPFL Xplore)
Category:semester project
Keywords:Locomotion, Mechanical Construction, Robotics
Type:100% hardware
Responsible: (MED 1 1226, phone: 32658)
Description:EPFL Xplore is an interdisciplinary project whose aim is to design and develop a Rover to participate in two international competitions: the University Rover Challenge and the European Rover Challenge. The chassis is one of the most important part of the Rover: it supports all the different sub-systems and components of the Rover. Therefore, the chassis has to be stable and reliable. The students will choose and design a chassis as light as possible that allow the Rover to navigate challenging terrains. It shall not be bigger than 1.2m x 1.2m and it will face some weight constraints. The Rover shall be easily maneuverable and shall not be stuck. The validation of the design will be based on various simulations and tests. The students are quite free in the design but the chassis shall be as safe as those that already exist, i.e. rocker bogie or shrimp model. Contact persons: Thomas.Manteaux@epfl.ch, Auke.Ijspeert@epfl.ch

Last edited: 26/06/2020

10 projects found.