PERA: Plug-in Experts for Robotic Actions in Continual Whole-Body Control
AuthorsShangchao Hu, Yadan Sun, Xibo Yu, Wei Huang
PERA teaches humanoid robots new skills by plugging small experts into an existing controller without disrupting its original motions.
Key results
Approximate share of the base actor’s parameters.
PERA completes all ten Unitree G1 simulation tasks.
Fully executed with the added expert disabled.
With the expert enabled, versus 59.38% with it disabled.
Mean error in cm across ten tasks, down from 173.36 cm for HoloMotion alone.
What the paper found
PERA adds new whole-body skills without fine-tuning or replacing a robot’s pretrained controller. It freezes a mixture-of-experts policy and trains a small task-specific feedforward expert and routing score; the expert joins Top-2 selection when enabled, and disabling it restores the base policy’s original computation path. Each added module is approximately 1.03% of the base actor’s parameters. Training uses PPO when task rewards are available, or behavior cloning followed by online DAgger when a teacher provides actions. In Unitree G1 simulation, PERA completes all 10 new tasks while retaining execution of all 126 original motions when the expert is disabled. On eight fast-dance references excluded from expert training, enabling the expert raises the fraction of fully completed episodes from 59.38% to 71.88%. The method also transfers to HoloMotion: adding an expert reduces mean endpoint error across the ten tasks from 173.36 cm to 28.51 cm. Together, the results show how modular experts can extend a frozen controller while preserving access to its existing skills.
Original abstract
Motion imitation enables humanoid robots to acquire whole-body skills such as walking and dancing. However, fine-tuning an entire policy for a new skill can affect existing behaviors, while training a separate policy for each skill increases training and maintenance costs. We present PERA, a method for learning new skills by reusing existing motion capabilities. PERA freezes a pretrained mixture-of-experts policy and trains an additional task expert and its routing score using task rewards or teacher-action supervision. In Unitree G1 simulation, each added task module contains approximately 1.03% of the parameters of the base actor. Built on our pretrained policy, PERA successfully completes all ten new tasks. Disabling the expert restores the original computation path of the base policy, which still fully executes all 126 original motions. Enabling the expert also improves the execution of some dance motions not used to train that expert, indicating that its benefits extend beyond its training motions. Experiments with HoloMotion further demonstrate the applicability of PERA to different base policies.
Read the original paperMore in Robotics
Browse all 68 papers →ELA: End-to-End Latent Action Policy Learning from Video Demonstrations
Cheng-Chun Hsu, Peter Stone
ELA turns plentiful robot videos into useful control experience by jointly learning latent actions, policies, and action decoders.
GRASP: Gradient-Guided Sampling for Contact-Rich Physics-Based Motion Retargeting
Yang Liu, Tong Wang, Yiming Bao, Heng Yang
GRASP uses past simulation rollouts to guide robot motion retargeting, producing contact-rich movements that work in simulation and on real hardware.
Alloy: A Multi-Material Physics Engine for Robot Learning
Pranay Gupta, Deepanshu Rohilla
Alloy brings robots, cloth, cables, liquids, and other materials into one GPU-capable simulator for robot learning.