Computational Robot Design: Differentiable Simulation and Control

This lecture covers topics such as optimization-driven design, generative and learning-based design, robot control, fabrication, and interactive workflows in computational design. It delves into differentiable models, simulation steps as neural networks, the material point method, closed-loop and open-loop control strategies, trajectory optimization, and hybrid control schemes. The instructor discusses the challenges and opportunities in computational co-design, the development of computational robot design algorithms, and the evolution of virtual creatures and soft machines. The lecture also explores bilevel optimization, evolutionary algorithms, and the process of evolving living creatures.