AI-Evolved Modular Robots Refuse to Die - Northwestern's Metamachines Adapt to Any Terrain

Northwestern University researchers have built modular robots that use AI-driven evolution to survive damage and keep moving. They call them metamachines - Lego-like modules that snap together into legged configurations and adapt on the fly when things go wrong. Each module is a complete robot with its own motor, battery, and computer. When assembled into three, four, or five-legged designs, they can run, jump, spin, and right themselves across gravel, grass, mud, and uneven terrain. If a leg breaks, the damaged part doesn't become dead weight - it detaches and keeps rolling or crawling independently. The AI behind this compressed billions of years of evolution into seconds, testing thousands of configurations to find the ones that move best. The result is robots that don't need retraining or manual setup to handle new environments. They just adapt. The research was led by Sam Kriegman at Northwestern's McCormick School of Engineering, and the study has been published in Nature. A robot that breaks apart and keeps going as smaller robots - that's not just resilience, it's a completely different design philosophy. One where failure isn't an endpoint, it's a reconfiguration.