Master thesis on learning control for self-assembled magnetic miniature robots. Collaboration between Physical Intelligence Department and Intelligent Control Systems Group.
Physical and computational adaptability is a sought after skill in autonomous robots. Especially in miniature robots, adaptation emerges from learning how to use the limited number of body components and their various configurations to generate the necessary functions.
Our research is based on the combination of physical experimentation, rigorous theory and mathematical analysis. In this project, we want to design miniature magnetic robot components which can self-assemble in different configurations. For every configuration, we want the robots to identify their morphology and learn how to walk on a surface (see project description for some examples). This project involves the exploration of component design space, modeling and analysis of physical interactions, applying machine learning methods, and experimental validation with fabricated robot components.
See project description for more information.