Publication
Towards Embodiment Scaling Laws in Robot Locomotion
Bo Ai; Liu Dai; Nico Bohlinger; Dichen Li; Tongzhou Mu; Zhanxin Wu; K. Fay; Henrik I. Christensen; Jan Peters; Hao Su
In: Computing Research Repository eprint Journal (CoRR), Vol. abs/2505.05753, Pages 1-33, arXiv, 2025.
Abstract
Cross-embodiment generalization underpins the vision of building gen-
eralist embodied agents for any robot, yet its enabling factors remain poorly un-
derstood. We investigate embodiment scaling laws, the hypothesis that increas-
ing the number of training embodiments improves generalization to unseen ones,
using robot locomotion as a test bed. We procedurally generate ∼1,000 embodi-
ments with topological, geometric, and joint-level kinematic variations, and train
policies on random subsets. We observe positive scaling trends supporting the
hypothesis, and find that embodiment scaling enables substantially broader gener-
alization than data scaling on fixed embodiments. Our best policy, trained on the
full dataset, transfers zero-shot to novel embodiments in simulation and the real
world, including the Unitree Go2 and H1. These results represent a step toward
general embodied intelligence, with relevance to adaptive control for configurable
robots, morphology co-design, and beyond.