Scalable decentralized control coordinates large robot teams accurately
Controlling Collectives of AI Agents in Reasoning Space with Spatial Transformers
RoboticsArtificial Intelligence
Summary
Robots working together in large groups often struggle with coordination, especially when the team grows large. The authors present COMPASS, a new system that helps many robots communicate locally using a method that processes messages in a spatial way to control their movements effectively. Their approach lets even very large groups of robots follow natural language commands and form organized patterns without central control. The authors show that using learned compact communication signals helps keep the robots coordinated better than methods that send raw data or rely on a single centralized brain.
What this means in practice
- •For robotics engineers: Coordinate large robot fleets with decentralized message aggregation for better team navigation and formation control under natural language instructions.
- •For drone fleet operators: Operate large groups of aerial robots to perform complex coordinated flight patterns through natural language commands processed on-board each drone.$Commercial implications: Enables commercial-scale drone fleet control for delivery and surveillance via efficient decentralized coordination, reducing reliance on centralized systems.
Authors
Frederic Vatnsdal, Roshan Gopal, Romina Garcia Camargo, Vijay Kumar, Alejandro Ribeiro
Abstract
Large Language Models (LLMs) introduce an exciting new paradigm for planning and navigation in robotics, but fail on even simple multi-robot tasks as team sizes grow. We propose COMPASS, a scalable, decentralized multi-robot architecture for controlling large collectives of agentic robots with reasoning space feedback control. Feedback is generated locally on each robot by a spatial transformer which aggregates multi-hop messages across the fleet into a learned feedback token. Our experiments find that collectives of language models demonstrate performance gains from structured diversity of the input command, which can cancel biases; an advantage that is held across scale. Compared against a centralized frontier LLM policy and a language-only communication ablation, we find that the coupled design of COMPASS decisively produces cohesive flocking formations that accurately fly the commanded intent. We show that reasoning feedback works best when composed with a compact learned token. Our ablations show that hand engineered feedback with raw state appearing in the language channel obliterates cohesion. COMPASS generalizes zero-shot to unseen instructions of ambiguous meaning while commanding flocks up to 16 times its training scale, flying up to 1024 robots under natural language commands.