Learning-based modeling and control of soft robots
Brief Description of Project:
Soft robots represent one significant evolution of robotic systems, since they are designed to embody safe and natural behaviors. The modeling and control of these robots is often challenging due to complex physical structures based on the soft materials. A common approach is to derive a simple, dynamic model from first-principles that, however, is very time-consuming and lacks accuracy. To overcome this issue, machine learning techniques deliver promising results in modeling and control of soft robot dynamics. Even though these models are highly expressive, the underlying physics of the system are typically neglected that results in data-hungry models and lack of trustworthiness.
The goal of this project is to 1) set-up a soft robot simulator 2) implement a physics-enhanced machine learning model and 3) use a model-based controller to steer the robot to a desired position.
- Highly motivated and independent thinker; interest on the intersection of machine learning, dynamical systems and soft robotics
- Comfortable with mechatronic systems and the basics of control systems
- Solid math background (linear algebra, differential equations)
- Experience programming in Python
Please feel free to contact the PI if you are interested in the project and have any questions about the qualifications.
Nature of Supervision:
Weekly one-to-one meetings with the PI including brief written reports about the past activities to keep track of the project process. The student can expect prompt feedback on project-related questions.
A Brief Research Plan (period is for 10 weeks):
Weeks 1-2: Setting up the soft robot simulator and learning the pipeline
Weeks 3-4: Setting up the interface between the simulator and Python
Weeks 5-8: Implementing and testing data-driven modeling and control approaches
Weeks 9-10: Analyze results and finalize report
Number of Open Slots: 1
Name: Thomas Beckers
Department: Computer Science