Semester/Thesis Projects

Semester/Thesis Projects

The Bio-Inspired Robotics Lab (BIRL) offers research projects for both bachelor and master students. The research topics and contents are essentially open to the students as long as they are challenging and within the scope of BIRL research. The research interests of the Bio-Inspired Robotics Lab lie at the intersection of robotics and biology. Toward our deeper understanding of animals' adaptivity and autonomy, we abstract design principles of biological systems, and develop core competences which are the design and control of dynamic mechatronics systems, bionic sensor technologies, and computational optimization techniques. From this perspective, we offer semester/thesis projects on both theoretical and experimental robotics research. For more detailed information, please contact Prof. Fumiya Iida (fumiya.iida -at- mavt.ethz.ch), or visit our homepage at: http://www.birl.ethz.ch/

Research Directions and Project Examples

"Design and experiment of Bio-Inspired Robots"

Motor control in biological systems are known to exploit a number of smart mechanical designs that, on the one hand, simplify the control problem of complex behaviors, and on the other, increase the stability in uncertain and noisy environment. The animals' muscle-tendon systems have, for example, a specific force-length curve profile, which is known to significantly increase the locomotion stability of legged animals. The goal of this research area is to develop novel mechanical designs of actuation and body structures inspired by the design principles in nature, and demonstrate "mechanical intelligence" in a simple robotic platforms such as a legged robot or a manipulator. While the design details will be dependent on the resources and the expertise of students, some potential ideas include:
    •    Development of walking, running, and climbing robots
    •    Development of novel actuators (e.g. series-elastic actuators, variable impedance actuators)
    •    Robot design with unconventional materials (e.g. carbon fibers, glass-reinforced fibers)
    •    Design and development of modular reconfigurable robots
    •    Development of unconventional rapid prototyping systems

"Modeling and Control of Bio-Inspired Robots"

System identification is one of the most critical procedures in control and optimization of the real-world robotic systems. It is not a trivial problem that engineers often put a lot of efforts to develop precise and effective models that match to the real-world systems. From this perspective, the challenge addressed in this project is to develop a learning architecture that automatically generates physics models from sensory-motor data of the real-world robotic platform. Through the trials-and-errors in the robot, the architecture corrects sensory-motor data and estimates equations of motion, for example. Starting from a relatively small problem such as a simple pendulum, this project tackle with more complex models of coupled pendula, the SLIP model, or the Compass Walker model, for example.
    •    System identification of underactuated robots (e.g. walking and running robots)
    •    Simulation of biological muscles and reflex based robot control
    •    Control optimization of underactuated robots in unstructured environments
    •    Modeling and simulation of soft robots

"Motor Learning and Self-Modeling of Autonomous Robots"

If compared to biological systems, conventional robotic systems still suffer from motion control in uncertain and dynamic environments. One of the key challenges in robotics, therefore, still lies in the motor learning mechanisms that function in the real world. In this research area, we investigate theories and techniques to deal with high-dimensional optimization problems of motor control in underactuated robots.
    •    Motor learning of CPG-based (or reflex-based) robot controllers
    •    Automated system identification of underactuated robots
    •    Reinforcement learning of real-world dynamic robots
    •    Visual navigation and learning of dynamic legged robot
    •    Development of a « Neural Machine» with the ability of resonance based reasoning.