PolyMorph policy controls different legged robots without retraining
PolyMorph: Learning Locomotion Policy for Cross-Embodiment Adaptation
Robotics
Summary
Legged robots usually need special control programs made just for their body shapes, which can be tricky when robots have different numbers of legs or joints. The authors created PolyMorph, a single control method that can work with many types of legged robots by understanding their body parts and joints as a graph. It learns how to make these robots walk smoothly, even if their shape changes or if it’s a robot it hasn't seen before. This means one program can make different robots walk without needing to be rebuilt for each one.
What this means in practice
- •For robotics engineers: Control multiple legged robot designs with a single learned policy, avoiding separate training for each morphology.
- •For robotics platform developers: Deploy one adaptive locomotion controller on various quadruped robots and configurations without redesigning control software.
Authors
Hyunsik Oh, Seunghun Jeon, Donghoon Youm, Jungwoo Hur, Hyeongjun Kim, Jeong Hyun Lee, Moonkyu Jung, Jemin Hwangbo
Abstract
Reinforcement learning has enabled robust locomotion for legged robots, but learned policies are typically tied to a specific robot morphology and require additional engineering when the number of legs, actuators, or mechanical properties change. We present PolyMorph, a morphology-conditioned framework that learns a single locomotion policy for heterogeneous legged robots. PolyMorph converts each robot's URDF into joint-link modules and combines a recurrent encoder, graph-based Transformer, and per-joint decoder to generate actions across variable leg counts and joint configurations. To improve training stability, PolyMorph uses morphology-normalized rewards, adaptive robot sampling, and URDF-derived PD-gain initialization. We evaluate PolyMorph in simulation across bipedal, tripedal, and quadrupedal robots, including unseen morphology variants and platforms, and deploy the policy on real quadruped platforms operated in multiple configurations. The results show that a single PolyMorph policy generalizes across diverse robot morphologies without per-robot policy retraining.