This site contains information and resources related to my research project on ECoDE(Efficient Co-Designer).
Co-design involves simultaneously optimizing the controller and the agent’s phys- ical design. Its inherent bi-level optimization formulation necessitates an outer loop design optimization driven by an inner loop control optimization. This can be challenging when the design space is large and each design evaluation involves a data-intensive reinforcement learning process for control optimization. To improve sample efficiency we propose a multi-fidelity-based design explo- ration strategy in which we tie the controllers learned across the design spaces through a universal policy learner for warm-starting subsequent controller learn- ing problems. Experiments performed on a wide range of agent design problems demonstrate the superiority of our method compared to the baselines. Addi- tionally, analysis of the optimized designs shows interesting design alterations including design simplifications and non-intuitive alterations that have emerged in the biological world.
Keywords: Co-Design, Reinforcement Learning, Robotics, Machine Learning
Trained Agents (Videos)