MorFiC achieves zero-shot locomotion transfer across quadrupeds of different sizes and masses with up to 5x speed gains over standard baselines.
arXiv · March 17, 2026 · 2603.14554
The Takeaway
It fixes 'value miscalibration' in multi-morphology training by using morphology-aware modulation in the critic network. This allows a single policy to work on entirely different robots (e.g., Unitree Go1 to B1) without any fine-tuning.
From the abstract
Generalizing learned locomotion policies across quadrupedal robots with different morphologies remain a challenge. Policies trained on a single robot often break when deployed on embodiments with different mass distributions, kinematics, joint limits, or actuation constraints, forcing per robot retraining. We present MorFiC, a reinforcement learning approach for zero-shot cross-morphology locomotion using a single shared policy. MorFiC resolves a key failure mode in multi-morphology actor-critic