AI Based Holistic Co-Design of Legged Robots

Research Project, 2025 – 2030

Legged robots provide superior mobility but are significantly more challenging to design and control. It is not trivial to predict how to design the best platform to perform a given set of tasks. This is aggravated by the fact that system design, motion generation and control are usually considered sequentially, while in reality, they are deeply interconnected. Splitting them into subsequent phases leads to a slow and expensive process in which the design is modified and tested multiple times before reaching an adequate performance. To address this, the optimization of the robot hardware along with its motion and controls is required. This concurrent-design optimization takes the name of co-design. The goal of my inter-disciplinary research group enabled by this SSF Future Research Leaders grant will be to develop an AI-based interactive and generative co-design framework that can automatically and systematically design legged robots with robust locomotion behaviors and provide safety guarantees of their stability in the presence of unexpected disturbances. The scientific approach will bring together design optimization, trajectory optimization and trajectory stabilization methods in a unified and holistic framework and combine it with generative AI to allow human feedback for making the design process easier for the end-users.