Multi-agent systems learn to avoid collisions using communication
Learning Communication-Conditioned Generative Policies for Decentralized Multi-Agent Collision Avoidance
Robotics
Summary
Avoiding collisions when many robots or agents move around is tricky, especially when they can't see everything. This paper shows how agents can learn to send and understand short messages that help them coordinate their moves safely without a central controller. They trained these behaviors using examples that already knew the whole situation but then learned to work with just local information and communication. This learning allows the agents to keep working even if some messages get lost and to handle new, busier situations they haven't seen before.
What this means in practice
- •For robotics engineers: Develop decentralized robot teams that avoid collisions using learned communication without central control.
- •For drone fleet operators: Manage large drone groups in complex airspaces where independent units coordinate safely via learned message exchanges.
Authors
Prajwal Koirala, Mark Campbell
Abstract
In this work, we propose a decentralized communication-conditioned generative framework for multi-agent collision avoidance. Agents generate short-horizon action sequences using a flow-matching policy trained from privileged offline demonstrations with access to global state. The demonstrations do not include explicit communication signals; instead, agents learn to exchange and aggregate latent messages that encode interaction-relevant intent under partial observability. This formulation supports flexible inference at test time, where unconditioned generation corresponds to independent behavior and communication-conditioned generation enables coordinated interaction without centralized planning. The resulting policies operate in a fully decentralized manner at execution time, relying only on local observations and learned messages. Combined with a receding-horizon inference scheme, the proposed approach enables efficient single-step inference of short-horizon action sequences and degrades gracefully under communication dropouts. Extensive simulation results demonstrate near-expert collision avoidance performance and strong generalization to denser, unseen multi-agent scenarios, along with zero-shot transfer to real-robot experiments.