Multi-robot Collaboration

Multiple robots dividing up work and coordinating to finish a task that one robot can't do alone or fast enough.

Multi-robot collaboration studies how multiple robots, identical or different, divide tasks, coordinate paths, and avoid colliding with each other while working toward a shared goal. Core problems include task allocation, multi-agent path finding, and communication or information sharing; control can be centralized, with one scheduling system directing everything, or distributed, where each robot decides based on local information — swarm robotics is the extreme case of the latter. The most mature deployment is in warehouse logistics: Amazon says it has deployed around one million robots and uses a foundation model called DeepFleet to coordinate its fleet, expecting it to cut robot travel time by 10%. Since large language models emerged, some work has also had multiple robot arms negotiate how to divide a task through language dialogue, such as 2023's RoCo.

ExampleIn RoCo, two robot arms collaborate on a tabletop task: each arm's LLM agent discusses in natural language who handles which step, then generates waypoints for a motion planner to execute, renegotiating whenever feedback from the environment signals a collision.

Also called

Multi-Robot Coordination, Multi-Robot Systems (MRS)

Related

Swarm Intelligence · Multi-Agent Path Finding · Multi-Agent Reinforcement Learning · Fleet Management System (e.g. Open-RMF) · Human-Robot Collaboration · Task Planning

Sources

Large Language Models for Multi-Robot Systems: A Survey
RoCo: Dialectic Multi-Robot Collaboration with Large Language Models
Amazon: 1 million robots and the DeepFleet foundation model

As of

2025