Legged metamachines keep moving after injury

This is a strong robotics paper because the AI is shaping the body plan, not just tuning a controller. The bigger idea is graceful degradation: robots that break, reconfigure, and still complete the job are far more useful than brittle demo bots.

• –The key advance is co-designing morphology from simple modules, letting evolution discover strange but effective forms humans would be unlikely to sketch first
• –Real outdoor tests on gravel, mud, roots, sand, and bricks make this much more credible than simulation-only robot evolution papers
• –Each module is its own robot with compute, power, and motion, so damage does not instantly turn the full system into dead weight
• –It is still early-stage research rather than a deployable platform, but the failure-tolerance angle is highly relevant for embodied AI and field robotics