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.
To limit the list to the projects matching a given keyword, click on it.
The CAN (Controller Area Network) bus is widely used in many fields, as it provides interesting features and reliability. The aim of this project is to implement a reliable communication protocol using the CAN bus on a PIC18F2580 microcontroller (which has embedded CAN hardware).
This project aims at realizing a miniature board with a powerful embedded processor, for the control of the movement of a salamander robot.
The board will be the central processor for the robot. It will have to be compact enough to be placed inside the head element of the robot. It will handle the locomotion and the vision of the robot (using two embedded cameras), as well as the radio communication with an external computer.
This project is a collaboration between BioRob and Ren Beuchat of the LAP.
This project is aimed at developing and testing simple electronic circuits to be used on board of the AmphiBot and Salamandra robotica robots. Depending on the interests of the student and the current needs of the lab, different types of circuits could be realized, especially different types of sensors (e.g., visible light or IR distance sensors, pressure sensors, light sensors, cameras, etc.) and of communication (e.g., radio, infrared, visible light, ultrasound, electric fields in water, etc.).
The goal of this project is to continue the development of a wireless remote control that communicates with CPG-controlled robot. The remote control should be able to configure a small number of locomotion parameters on the robot, and to upload a new firmware from a SD card to the robot.
This project aims at adding artificial vision to an amphibious salamander robot. The goals of the project are to design algorithms for stimulus tracking and obstacle avoidance for the robot moving in cluttered environments. The algorithms should first be designed and tested in a simulated world (i.e. with pixels coming from virtual cameras in a 3D environment). They should then be tested on the real robot in engineered environments of increasing complexity.
The project will be carried out in the framework of a collaboration between the Computer Vision Lab and the Biologically Inspired Robotics Group. The student should have good programming skills (C or C++), a strong mathematical background and an interest in computer vision and robotics.
In a collaboration with Prof. Barry Trimmer at Tufts University, BIOROB participates to the study of locomotion control in soft-bodied animals such as caterpillars, and its application to soft-bodied robots.
The goal of this project is to design adaptive locomotion controllers based on the concept of central pattern generators for controlling the locomotion of a simulated soft-bodied animal. The work also implies developping a simple spring-mass simulator, inspired by the Soda applet (see Sodaconstructor.com).
The goal of this project is to develop the hardware necessary for adding an electric sense to a swimming snake robot. Such a sense is found in some fish species and is very useful to detect obstacles and other animals (preys/predators) in murky water where vision is of limited use. See here for details.
The project will be part of a collaboration between CVLab, which has proposed a theoretical model of how this sense might work, and BioRob, which has developed Amphibot, a swimming snake robot.
The project will involve (1) the design of a snake prototype with regularly spaced electrodes and of the electronic boards necessary for processing the voltage measurements, and (2) the characterization of the type of signals that can be extracted in various conditions, for example by putting various objects/obstacles in the water.
If time allows, the system will be installed on board of the functioning robot and tested for its usefulness to avoid obstacles during swimming.
This project aims at characterizing the locomotion characteristics of the AmphiBot snake robot, or the salamander robot Salamandra robotica. This characterization may include the influence of the parameters of the PD controllers on the generated movements, the study of different types of locomotion (for walking, crawling, or swimming) and their influence on the robot's speed, etc.
The aim of this project is to implement different behaviours on the snake robot AmphiBot and/or the salamander robot Salamandra robotica. This can include for example obstacle avoidance using infrared sensors or phototaxis (i.e., following a possibly moving light source).
Nature shows a great variety of geometries and mechanical properties in the caudal fins (tails) of swimming animals. These properties have a big impact on the efficiency and speed of locomotion in water. The goal of this project is to design and investigate the effect of a passive caudal fin on a swimming snake robot, taking inspiration from biology.
The goal of the project is to apply on the motion planning problem of a snake-like robot (here the Amphibot III), the classical tools of control theory. The first step of the project will consist in reducing the inverse dynamic model of Amphibot III. This numerical model is a recursive algorithm based on the Newton-Euler equations (cf. [1]). It computes the head accelerations and the joint torques as a function of the joint positions, velocities and accelerations. As far as the reduced model is concerned, it will be obtained by systematic tests.
The idea is to design a map of the forward mean velocity of the robot head as function of the input parameters of the propulsion wave moving along the body (i.e. frequency, amplitude, phase lag). Then in a second step, in order to apply the control laws, the student will design a kinematic model of the snake-like robot. This model will be of n-trailers type. Finally, on the base on the kinematic model, the
last step of this project will consist in applying control laws based on the flat output theory [2].
[1] W. Khalil, G. Gallot and F. Boyer, "Dynamic modeling and
simulation of 3-D serial eel-like robot", IEEE Trans. on Systems, Man, and Cybernetics - Part C: Applications and reviews, vol. 37, no 6, November 2007
[2] P. Rouchon, M. Fliess, J. Levine and P. Martin, "Flatness, motion planning and trailers systems", Proceeding of the 32nd C.D.C, San Antonio, Texas, December 1993
Our salamander and lamprey robot are good anguilliform swimmers. However they sometimes show instability due to unwanted rolling. Taking inspiration from biology[1], the student should design and implement feedback controller capable of stabilizing the roll during swimming.
Tasks include:
- adding an inertial sensor in the head of the robot
- possibly designing an asymmetrical caudal fin to allow for roll control
- designing a feedback controller in simulation
- implementing the controller in the robot microcontroller
- testing the implementation with the real robot swimming in a pool with perturbations
Reference:
[1] Kozlov et al., Modeling postural control in the lamprey, Biological Cybernetics, 2001
The aim of this project is to evaluate different radio communication possibilities for use in a future version of the amphibious snake/salamander robots. This might include looking for different radio chips/modules, realizing simple circuits with a microcontroller and a radio chip (e.g., nRF24L01 ) or module (e.g., MRF24J40MB), evaluating different antennas, measuring maximal distance and throughput, etc.
Most of the existing robots do not use fingers as part of their feet and good traction and terrain adaptation are mostly ignored. Adaptability and compliance at the foot level could yield increased speed and efficiency of locomotion. The aim of this project is to study the importance of different parts of the real salamander's limbs and transfer these features to the limbs of the Salamandra Robotica II.
*This project is a continuation of a previous semester project: http://biorob.epfl.ch/page-53933-en.html
10% theory, 10% hardware, 60% software, 20% other activities
Responsible:
(INN 233, phone: 36714)
Description:
Tail is found in most of the animals in nature. In this project the aim is to study the effect of the tail using a salamander robot.
Dynamical systems
The field of complex systems studies the nonlinear dynamics of systems with multiple components such as neural networks, ecosystems, the internet, ... In particular, it is interested in phenomena such as self-organization, synchronization, emergence, and other multiscale phenomena.
The goal of this project is to implement nonlinear oscillators for the locomotion control of modular robots on a network of ARM microcontrollers interconnected by a global CAN bus and possibly by neighbor-to-neighbor serial connections. The oscillators will be coupled together to implement distributed locomotion controllers (central pattern generators).
This projects aims at investigating the use of chaotic systems for escape reactions and exploration in robots.
It is known that nonlinear dynamical systems are an interesting tool to generate movement and locomotion in robots. Usually, the systems are used in a periodic or discrete regime. However, another aspect of nonlinear dynamical systems is that often they can be brought into a chaotic regime by the change of a single parameter.
The goal is to investigate the use of the chaotic regime for useful behavior of the robot such as escape reaction or exploration behavior. The work will be done mostly in simulation.
We have developed mathematical models of adaptive frequency oscillators (AFOs) that can adapt their parameters to learn the frequency of any periodic input signal. This mechanism goes beyond mere synchronization because the new frequency stays encoded in the system even if the teaching signal disappears and it works for any initial conditions. See http://biorob.epfl.ch/page38914.html
for more information.
When multiple AFOs are coupled together they can perform a kind of dynamic Fourier transform of arbitrary, possibly time-varying, signals. The goal of this project is to analyze the properties of the system and to adapt it for applications in signal processing, such as compression and frequency analysis.
Several stability criteria exists in the literature (ZMP-CoP, GCoM, FRI,..). This project aim at reviewing and comparing those criteria for quadruped and biped locomotion.
A selection of stability criteria will be implemented in Webots on the robot iCub and Aibo.
For a master project, a strategy could then be designed to ensure stability of the robot during locomotion, with the possibility of implementing it on the real robot Aibo.
In this project we would like to use reservoir computing approaches for solving stable movement generation problem for a soccer player humanoid robot. Some basic actions such as walking, running, and kicking beside more complex combined actions such as dribbling will be investigated.
Reservoir Computing methods are new, simple and powerful approaches for training of recurrent neural networks which are currently becoming more and more popular. The idea is to use a random recurrent neural network, known as reservoir, which is left almost untrained, and using only adaptation of the weights of the output layer, known as readout. Different dynamical systems can be utilized as transfer function of the neurons that lead to emergence of qualitative different swarm behavior.
In particular, the project can be divided into two parts:
- A dynamical system will be designed which is used as the transfer function of the neurons that makes the reservoir able to produce periodic, discrete and rhythmic behavior by means of a few adjustable global parameters.
- A control strategy for stable locomotion will be developed which will consist in two readouts: (a) a rhythmic readout generating the pattern of locomotion, and (b) a discrete readout applying discrete corrections to the rhythmic movement to ensure stability (according to the center of mass information).
Utility of distributed mobile robotic platforms, constituted for instance of a group of mobile agents, oftentimes hinges upon the capacity of the group members to conform to a given formation or group configuration. While there exists in the literature powerful theoretical tools allowing to address the formation maintenance problem, their application to the case of swimming robots raises interesting issues. Of particular interest is the interaction between (lower level, CPG-based) motion controllers and (higher level) formation maintenance control algorithms.
The project will begin with a (short) literature review on formation maintenance. Following selection (and possible adjustment) of an appropriate algorithm, numerical simulations will be used to assess the efficacy of the chosen solution. Implementation on actual swimming robots will then be pursued (time allowing).
This project (master or semester project) aims at the implementation, or the support for the implementation of a force treadmill for the gait analysis of robots (biped, quadruped, centipede, modular) at the Biorobotics laboratory at EPFL, Switzerland. The work can be extended by preliminary tests and experiments of a legged quadruped on the resulting force treadmill (to be provided: Cheetah, small quadruped compliant robot).
Depending on the background of the student, projects with an emphasis both on mechanics or electrics and control can be pursued. Preliminary mechanical sketches, and extensive reading material on force plates, load cells, and gait analysis are available. Any design, whether mechanical, electrical or in software, should be heavily based on off-the-shelf
parts, components, and systems.
Extensive project description:
PDF
The goal of this project is to implement learning algorithm in a real humanoid robot (Hoap2) so it can learn how to perform movements we show to it. The learning algorithms will be based on programmable CPG controllers (refer to the Dynamical movement primitives page or the Programmable CPG page for more information)
The goal of this project is to study the control of biped locomotion for a humanoid robot. You will have to design a controller for the robot so it can walk at different speed, turn or go backwards. The controller will have to be robust against perturbations, it means it should not fall if the ground is not flat, or if someone pushes it a little.
The integration of feedback will be important to deal with unpredictable environments. The approach to design such controllers will be CPG based controllers.
The first experiments will be done on a simulated robot and then they will be implemented on the Hoap2 humanoid robot if the time permits.
Central Pattern Generators (CPGs) offer an interesting solution for the control of locomotion of legged robots, especially bipeds. This concept comes from biology, where CPG neural networks located in the spine of animals are supposed to be responsible for the generation of complex control signals for the coordination of the limbs during periodic movement (e.g. locomotion). These CPGs are activated and modulated by simple control signals coming from higher part of the brain and are tightly coupled to sensory information (force and position sensing).
In robotics, CPGs are often modeled as systems of coupled oscillators. The interest of such systems is their intrinsic stability properties that make them suitable for feedback integration (limit cycle behavior), their synchronization properties that allow entrainment between the robot and the controller. They also reduce the dimensionality of the control problem, since only parameters such as frequency, amplitude and coupling between the oscillators have to be chosen to generate high dimensional control policies. However a major drawback of these approaches is the difficulty to integrate specific constraints, such as for example to fulfill dynamic equilibrium criteria.
We have found recently that the dynamic equilibrium of a walking motion can result from very simple rules such as minimizing the norm of derivatives of the future motion of the Center of Mass. The idea is to continuously look ahead in time for a couple of seconds and predict the outcome of the motions we are going to realize. It can be shown then mathematically that continuously choosing the motion with minimal norm of derivatives will lead to a stable motion when possible. The idea that we propose to follow here is therefore to model the signals that usually come from the higher part of the brain and modulate the CPGs with such a simple rule: continuously adapting the frequency, amplitude and coupling parameters of the CPGs for obtaining a minimal norm of the derivatives of the future motion of the Center of Mass. The objective then is to obtain a robust dynamically stable CPG based walking motion. This research project will be realized in collaboration between the http://biorob.epfl.ch/ (EPFL) and the BIPOP (INRIA) teams.
(PSE C 330, phone: 38624)
(PSE C 330, phone: 38624)
Description:
Nao is a small humanoid robot for entertainment currently under development at Aldebaran Robotics.
The Nao robot was selected by the organizing committee of the RoboCup as the successor of the Aibo robot for the Standard Platform League. The simulation needs to precisely model Nao's dimensions, servomotors (axes, torques, etc.), mass distributions, etc. (We hope to be able to get a hardware sample of the Nao in order to facilitate the modeling process).
The first part of this project is to improve (or redesign) a physically and visually accurate model of Nao for the Webots simulator.
Then a robot controller must be programmed (in C/C++) in order to prepare a demo for the RoboCup (walk, shoot the ball, etc.) or a demo of wresting robots (make the other robot fall, etc).
Find more info there: Aldebaran Robotics, Cyberbotics and Robocup.
The goal of the project is to develop a controller for biped locomotion and its stabilization based on studies on monkeys.
A central question to be addressed is the minimal information - both sensory and cerebral - needed by the spinal cord circuits to generate stable locomotion,
Results will be confronted to real biped walking data of monkeys.
This project is done in collaboration with Solaiman Shokur from LSRO.
Crawling is the main type of locomotion that infants use. The main concept of this project is to improve the CPG-based crawling controller that is implemented. The key points in this project are:
1- Study the effect of contact sensor feedback on crawling, and try to improve it.
2- Examine the closed-loop controller in different situations to see the robustness of the controller.
This project will be done on the iCub simulator in Webots.
The goal of this project is to design locomotion controllers for modular robotic systems, e.g. with our Roombots platform. It is challenging to create controllers that can learn how to cope with arbitrary and changing robotic structures, e.g. after the addition/ the removal of modules and limbs. Successful student projects in this topic have been done by
Simon Lepine,
Sandra Wieser,
Jerome Maye,
Michel Yerly,
Yvan Bourquin and
Daniel Marbach.
This project is part of the Roombots project that aims at using modular robots for adaptive and self-reconfigurable furniture. The project should participate to the design of a database of interesting furniture configurations, and of algorithms for planning reconfigurations from one robotic structure, e.g. a chair, to another, e.g. a table. The system should be able to decide when prior reconfiguration plans can be re-used and when new ones need to be learned. At least two reconfiguration-strategies are open to exploration: centralized and decentralized approaches.
Modular robotics are based on the assembly of many small, independent modules into larger structures capable of complex tasks (e.g., walking). Finding a viable structure for the assembled modules is a hard task. In this project, the student will use developmental algorithms (i.e., algorithms inspired by the growth process of living organisms) to direct the assembly of the modules. Artificial evolution is a possible tool to achieve this goal.
The goal of this semester/master project is to develop a graphical
user interface (GUI) which provides users the
ability to build different structures by connecting the Roombots modules as building blocks.
Background: Modular Robotics is an approach to building robots for various complex
tasks. Instead of designing a new and different mechanical robot for
each task, many copies of one simple module are built. The module
can't do much by itself, but when being connected many of them
together assemble a system that can do complex tasks. In fact, a
modular robot can even reconfigure itself change its shape by
moving its modules around to meet the demands of different tasks or
different working environments. Our group has recently
developed new prototypes for modular robot units, called Roombots. A model of
these modules in Webots
simulator is also available.
The approach must be general such that adding different modules to the
collection of the building blocks is possible. No physics simulation
is required, except for collision detection. You must be
proficient in C/C , GUI, and 3D programming.
We have developed modular robot units: Roombots. The aim of the Roombots project is to create robot units that can rapidly attach to each other in order to create arbitrary multi-robot structures. The units have on-board wireless communication. The goal of this semester project is to develop a graphical user interface on the iPad to remotely monitor and control multiple Roombots units. The GUI is supposed to enable monitoring the value of sensors, encoders, and camera. It is supposed also to enable the user to control the units by sending high or low-level commands to one or a group of them. To carry out this project you need knowledge in objective C.
Modular robots present an interesting platform to explore locomotion strategies
for robotics. However, the large variety in robot configurations and having many degrees of freedom make it problematic for the user to design the suitable controller to achieve the desired locomotion behavior and also to re-design (adaptation) of the controller for the new morphologies. The idea in this project is to use morphological information in order to reduce the computational efforts for evolving the controller by providing the robot with some basic knowledge about itself. We will use and improve some developed tools for extracting these sets of information and investigate the effect of using them on the online-learning capability of the robot.
Modular robots with their network of microcontrollers and distributed computational units are a perfect playground for distributed control and dynamic reprogramming of individual computational units. In this project we will design and implement a bootloader that allows dynamic reprogramming of individual microcontrollers over a RS485 network. Demonstrations and testing will be performed on our Roombots platform. A perfect candidate should have good skills in C-programming as well as in embedded systems.
In order to ease the building of Roombots structure we would like to create an easy to use interface, especially targeting naive users that would like to explore the capabilities of the Roombots. To do so we come up with the idea of using Play-doh as a tangible interface for the creation of RB structure. The user would create a 3D structure using Play-doh, scan it with a 3D scanner and a software will then be used to reconstruct the corresponding RB structure.
The main questions that we would like to address in this project are the following:
1. Given the volume corresponding to the Play-doh structure, how should the RB modules be placed to maximize the occupied space ? In other words, what are the different servo angles and connection types allowing such a structure to be built ?
2. It seems relatively clear that the solution to the filling problem is not unique: it will depends on the granularity of resulting RB structure, i.e. if we increase the number of modules (without considering the required scaling factor) we will increase the quality of the fit in the volume, but this will require many modules. So the goal would be to optimize this number of modules such that the filling of the volume is above a given threshold (for example).
3. Supposing that the entire process of scanning and converting the volume into modules is fast enough, we could explore the possibility of using the Play-doh structure (as well as the mockups) to control the real modules, e.g. to slowly move them from one pose to the other.
This project will be co-supervised by the Laboratoire d'Informatique Graphique et Geometrique (LGG) of EPFL.
In this project we evaluate the use of Zigbee communication nodes for locatization in modular robots. It has been shown that based on the power consumption of the Zigbee communication we can estimate distances. In this project we will evaluate energy efficient strategies and algorithms to reliably estimate distances between Roombots modules and for building localization maps in unknown environments. The project will be partially carried out in simulation and will include many experiments with the real hardware. This project is a collaboration with the startup Domo Safety.
The current version of the Roombots module has very limited sensory capabilities including position and torque sensors for the different degrees of freedom. In order to have a consistent representation of the state of the Roombots structure, we would like to incorporate different kind of sensors like infrared distance sensors, accelerometers, gyroscopes, strain gauges.
The goal of this project is to design an optimal sensor layer for Roombots to support self-modelling of their modular structures. More precisely, the project can be decomposed into two main parts:
1. Sensor layer optimization: in order to have precise knowledge of the current state of the structure we aim at finding the best distribution of the sensors in the Roombots structure in terms of number of sensors and type so that it is possible to reconstruct the structure from the sensory data. Several case studies should show to what precision the structure can be determined if sensors fail.
2. Self-modelling: using the sensory information we would like to be able to directly create a consistent simulated model of the robot, integrating the effect of the gravity (bending effect) or of the motors backlash for example. This process should be fully automatized and applicable to both static structures and dynamic ones. The noise at the level of the sensor measurement should be included in the model and can e.g. be represented as probabilistic uncertainty.
Our colleagues in Denmark developed a versatile robotics tool kit called Locokit. Locokit is designed for exploring the locomotion capabilities of a variety of robots with changing morphologies. The goal of this project is to use online optimization to learn optimal gaits for a quadruped robot composed of Locokit elements. Experiments will be performed on real hardware using a motion capture system. If the project is successful the student will be invited to present her/his results at the international robotic competition at AMAM 2013. The ideal candidate should be very motivated and have experience with programming in C/C
and interest in AI/learning.
Quadruped robotics
A small exert of possible projects is listed here. Highly interested students may also propose projects, or continue an existing topic (contact people).
Recent research shows that passive body dynamics, such as elasticity
(compliance) is an important part of locomotion. It is however not
clear yet how to control such systems in an efficient manner.
The latest addition in our robot zoo is a
robotic
cat with tri-segmented, springy legs,
thus very pronounced passive body dynamics. The goal of this project
is to explore locomotion control strategies for such a quadruped
mammalian robot. Further details:
[Publication][Video walking gait][Video pacing gait].
We recently developed the mechanical framework for a new cat-like quadruped robot (see the semester project of S. Rutishauser, further details: [Publication][Video]).
Task of this project is to integrate newly purchased controller boards, and/or to develop a self-sufficient power, actuation and sensor system based on RC servo motors or, DC motors/gearboxes.
We are interested in enhancing the existing design of our compliant quadruped robot platform. This includes a) spring/damper/pantograph leg, b) motor/actuator, c) compliant foot, d) adaptive and active spine, and e) scapula design. Ideally a state-of-the-art survey is carried out, and the mechanical/mechatronical design and implementation is backed up with appropriate modeling, e.g. modeling of the robot's dynamics in simplified setups (Sagittal layer). Possible projects can have one of the above sub-tasks as topic, but also combined approaches. Interest in bio-inspired robot designing is required.
A variation of the project can be carried out entirely in simulation, i.e. in Webots.
Animals can adapt their locomotion gait when dealing with some injuries and in turn changes in their bodies. That is an interesting research question that which strategies they use to adapt their motor control skills for the new morphology. In this project, we have links and collaborations with biologists to extract insights from the experimental results on animals. We will use extracted data on several three-legged dogs to investigate natures strategies to deal with this problem and we will use them for our simulated quadrupedal robot to improve its robustness against the damages in its body.
Our colleagues in Denmark developed a versatile robotics tool kit called Locokit. Locokit is designed for exploring the locomotion capabilities of a variety of robots with changing morphologies. The goal of this project is to use online optimization to learn optimal gaits for a quadruped robot composed of Locokit elements. Experiments will be performed on real hardware using a motion capture system. If the project is successful the student will be invited to present her/his results at the international robotic competition at AMAM 2013. The ideal candidate should be very motivated and have experience with programming in C/C
and interest in AI/learning.
Rehabilitation robotics
Rehabilitation robotics is a field of intense investigations for helping people having movement disorders, e.g. due to a stroke or amputation.
Robotics for rehabilitation therapies is appealing to take over part of the burden of the therapists. The goal of a rehabilitation robot is therefore to help the patient to move, taking his/her intentions into account.
This group lists projects dealing with robots for rehabilitation, assistance, and prosthetic.
See the lab research page.
The goal of this project is to use optimization algorithms such as Genetic Algorithms, or Particle Swarm Optimization, to develop a model of a walking human in simulation in order to research the underlying principles of human locomotion and how this information can be used for effective functional electrostimulation. The project will consist of the following items:
Accurately modeling a human (skeleton) in a physics simulator, for example Webots.
Exploring different types of muscle models (in increasing complexity) that capture the most important aspects of the human muscle system.
Optimize the control input (to some degree related to functional electrostimulation) for these muscle models to generate a nominal walking gait.
A selection of the above items can be made depending on the type of project.
Many results were obtained to demonstrate that passive dynamics play an essential role in energy-efficient locomotion by humans and animals. These concepts were successfully transferred to robotics design leading to (quasi-)passive walking robots.
In this project, we would like to address the use of passive elastic tendons spanning across multiple articulations to boost walking efficiency.
This is based on a paper providing simulation results:
Exotendons for assistance of human locomotion, A.J. van den Bogert, Biomedical Engineering OnLine, 2003.
The project might cover the following items:
1) Reproduction of the paper results
2) Similar optimization over broader search space (evolution/optimization of the exoskeleton structure)
3) Validity and robustness to different gait patterns
4) Implementation of similar mechanisms to a real robot, e.g. Hoap2
Intensive rehabilitation is necessary for impaired people to recover proper locomotion, but is highly time and energy consuming for the physiotherapists. In that perspective, robot-assisted rehabilitation is a promising avenue.
In this project, the student will have to develop new rehabilitation protocols based on the concept of Central Pattern Generators. Implementations on real rehabilitation robots will be achieved, through a collaboration with the LSRO lab.
Good programming skills is a plus.
References:
Programmable Central Pattern Generators: an application to biped locomotion control, Righetti and Ijspeert, ICRA, 2006.
Lower Extremity Exoskeletons and Active Orthoses: Challenges and State-of-the-Art, Dollar and Herr, IEEE Transactions on Robotics, 2008.
The master thesis of Sarah Moussouni, available in summer 2010.
The goal of this project is to continue the design and programming of a demo experimental setup for various applications, including human augmentation. Human augmentation means that an external device provides some torque to a user in order to augment his/her strength.
This setup is based on a single degree-of-freedom structure in parallel to the elbow: the arm is inertially fixed, the forearm is attached to the robot beam, and a motor actuates the joint in parallel to the elbow.
From the existing hardware, the student will have to:
(1) Design a user interface.
(2) Reproduce existing results for human augmentation about the elbow.
(3) Study the mechanisms of resonance tuning.
Reference: the semester project of Fabien Delaloye, available in summer 2010.
The LSRO have developed a lower limb orthosis for rehabilitation purposes, the Lambda robot. This orthosis has three actuated degrees of freedom per leg and can be used to rehabilitate people with impaired movement control (due to a variety of conditions, such as muscle weakness). Although robotics have been used for some time for rehabilitation purposes, it is yet unclear what the best strategy is to facilitate optimal rehabilitation for impaired people. In this project, a new strategy for adaptation of the control of the device, based on haptic feedback will be developed, aiming to maximize the rehabilitation performance of the robot. The effectiveness of the new control strategy will be validated and analysed using human subjects.
This project is in collaboration with the LSRO lab.