AI Gives Robots a Bounce

MIT’s CSAIL team taught generative AI (specifically, diffusion models) to finetune robot parts. By testing 500 initial designs and refining the top performers over five rounds, the AI “blobbed” out a winning shape—thick, drumstick-like linkages that store energy efficiently. The result? A robot that hops about 2 feet high, a 41% improvement over the human-designed version, while landing falls plummet by 84% thanks to AI-tuned feet.

“We wanted to make our machine jump higher, so we figured we could just make the links connecting its parts as thin as possible to make them light,” says co-lead author and CSAIL postdoc Byungchul Kim  . “However, such a thin structure can easily break if we just use 3D printed material. Our diffusion model came up with a better idea by suggesting a unique shape that allowed the robot to store more energy before it jumped, without making the links too thin. This creativity helped us learn about the machine’s underlying physics.”