Pretrained robot skills adapt to team coordination without losing solo ability
Residual Denoising Enables Sample-Efficient Multi-Agent Coordination on Demand
Robotics
Summary
Robots often learn to do tasks alone, but working together with other robots is harder. The authors show a way to adjust a robot’s existing skill to help it team up with others, using only a little group training data. Their method keeps the robot good at working alone while improving teamwork by adding a special coordination step that changes how the robot moves only when needed. This works without the robots directly talking to each other and was tested both in simulations and on real robots.
What this means in practice
- •For robotics engineers: Improve teamwork of multiple robots on tasks without retraining all skills from scratch, using minimal collaborative data.
- •For drone fleet operators: Enable drones to collaborate effectively in a team while retaining their individual flight capabilities without explicit communication.
Authors
Dayi Dong, Maulik Bhatt, Aayushi Shrivastava, Lasse Peters, Negar Mehr
Abstract
Pretrained robot policies offer strong manipulation skills but are typically limited to single-agent settings, where a robot acts in isolation. In this work, we study how to adapt pretrained single-agent diffusion policies to multi-agent settings using minimal collaborative data, co-optimizing for two key objectives: high coordination performance and single-agent skill retention. To this end, we introduce ALTER, an adaptation method for coordination on demand: the adapted policy coordinates with other robots when deployed in a team while remaining capable of acting independently when operating alone. Execution is decentralized: each robot acts only on its own visual observations, without explicit inter-agent communication. Our method trains a coordination head that predicts a residual denoiser to transform single-agent behavior into coordinated multi-agent behavior when necessary while also preserving single-agent capabilities. To preserve single-agent capabilities, we augment a small number of collaborative demonstrations with self-distilled data generated by the base policy during training of the residual denoiser. In simulation,ALTER achieves higher coordination success over our baselines while retaining much higher source-skill retention. In our hardware experiments, we find similar trends where ALTER better co-optimizes for coordination success and single-agent skill retention than the baselines