Robots That Redesign Themselves through Kinematic Self-Destruction
Chen Yu, Sam Kriegman
AI summary
Problem
Most robots are statically predesigned and cannot modify their own structure during operation. This work addresses how a robot can dynamically adapt its morphology in real-time to overcome inefficient or novel body plans.
Approach
The authors train a universal autoregressive transformer policy using reinforcement learning and real-world trajectories to identify redundant body modules, execute controlled self-destruction via proprioception, and adapt locomotion control to the reduced structure.
Key results
- Universal transformer controller learns selective module detachment and adaptive locomotion
- Significantly higher locomotion speeds on 100 unseen simulated morphologies compared to non-destructive baselines
- Successful sim-to-real transfer on both in-distribution and out-of-distribution physical robots
- Prompt reset mechanism prevents policy freezing and improves out-of-distribution generalization
Why it matters
Demonstrates that irreversible, self-directed morphological adaptation can enhance robot resilience and performance, offering a new paradigm for adaptive, self-designing machines.
Abstract
Every robot built to date was predesigned by an external process, prior to deployment. Here we show a robot that actively participates in its own design during its lifetime. Starting from a randomly assembled body, and using only proprioceptive feedback, the robot dynamically “sculpts” itself into a new design through kinematic self-destruction: identifying redundant links within its body that inhibit its locomotion, and then thrashing those links against the surface until they break at the joint and fall off the body. It does so using a single autoregressive sequence model, a universal controller that learns in simulation when and how to simplify a robot’s body through self-destruction and then adaptively controls the reduced morphology. The optimized policy successfully transfers to reality and generalizes to pre- viously unseen kinematic trees, generating forward locomotion that is more effective than otherwise equivalent policies that randomly remove links or cannot remove any. This suggests that self-designing robots may be more successful than predesigned robots in some cases, and that kinematic self-destruction, though reductive and irreversible, could provide a general adaptive strategy for a wide range of robots.