Robot Morphology-Control Co-Design
Overview
Robot morphology-control co-design studies how a robot’s shape, materials, actuators, and control policy can be designed together, since each constrains what the others can achieve. Common approaches include evolutionary and Bayesian search, bilevel reinforcement learning, and differentiable simulation or learned latent spaces that make design exploration more efficient. These methods are used across soft and legged robots, hands and tools, and aerial systems, with evaluations ranging from simulated locomotion benchmarks to fabricated robots and real-world tasks. A central challenge is finding designs that remain robust and manufacturable beyond their training conditions, while making comparisons fair across different controllers, search budgets, and benchmarks.
Latest papers
60 papers11 papers in the last four weeks, up 10% on the four weeks before. 0.1% of all new papers.