Residual Denoising Enables Sample-Efficient Multi-Agent Coordination on Demand

2026-09-26

AI Summary

This study addresses the inefficiency and skill degradation encountered when adapting pretrained single-agent policies to multi-agent collaboration. To overcome these challenges, we propose ALTER, a framework that introduces a predictive residual denoising coordination head integrated with self-distillation-based data augmentation and diffusion models, enabling on-demand coordination alongside decentralized visual execution. This work presents the first approach to jointly optimize collaborative performance and independent capabilities. Both simulation and hardware experiments demonstrate that ALTER significantly improves collaboration success rates under minimal data regimes while substantially outperforming baselines in preserving source skills.

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

Problem

• multi-agent coordination
• pretrained robot policies
• diffusion policies
• sample efficiency
• skill retention

Innovation

• Residual Denoising
• Diffusion Policy
• Multi-Agent Coordination
• Self-Distillation
• Decentralized Execution