PhysEvo improves robot control and self-updating skills in physical tasks
PhysEvo: Astra Can Act, Let It
RoboticsArtificial Intelligence
Summary
Robots need good ways to watch and act on their environment to do tasks well. The authors introduce PhysEvo, a system where a fixed robot brain learns by trying tasks, checking what went wrong, fixing its methods, and testing again. Over many tests, PhysEvo improved robot skills without changing the brain's core, and it worked better than previous approaches both in simulations and on real robots. This lets a robot get better over time just by learning from its own actions and corrections.
What this means in practice
- •For industrial robot operators: Deploy robots that can autonomously refine and improve their manipulation skills over time without retraining their core control model.$Commercial implications: Enables sale of adaptive robotic systems that improve reliability and effectiveness in complex factory tasks without costly retraining.
- •For robotics software developers: Create robotic software that uses self-diagnosis and skill revision to improve task success rates in varied environments.
Authors
Wenqing Tian, Zeyu Zhang, Zhaocheng Liu, Fengwei Liu, Qiang Liu, Liang Wang
Abstract
Astra can act, yet reliable manipulation depends on the system through which it observes and controls the world. We introduce PhysEvo, a framework for physical recursive self-improvement (RSI) around a single frozen model. A task agent executes robot tasks; a meta-agent uses the resulting trajectories to diagnose failures, revise tools and skills, and test corrections. The meta-agent can also improve its own diagnostic tools, so retained revisions support both later action and later self-improvement. This process develops joint-level control, evidence-seeking observation, and reusable manipulation skills without model-weight updates or a separately trained action policy. Across 42 RoboDojo tasks, held-out-layout evaluation of retained task-specific deployment versions yields a five-dimension average score of 68.14/100 and 62.00% success, compared with 47.17% for RoboDawn's one-shot Astra agent, the strongest published reference in our comparison. On eight manipulation tasks challenging direct Astra, PhysEvo achieves 55.00% success, compared with 1.25% for the direct-Astra reference. Deploying the simulation-evolved harness on AgileX PiPER and continuing skill revision yields 90.60/100 average score and 84.00% success across 25 trials on five real-world tasks. PhysEvo turns the consequences of action into persistent, testable changes to how a frozen model acts and improves.