Graph Neural Networks for Learning Robot Team Coordination

TL;DR

This paper explores using Graph Neural Networks to enable distributed coordination in robot teams by modeling communication links as graphs, allowing robots to learn message passing and state updates for collective behaviors.

Contribution

It introduces a novel application of GNNs for robot team coordination, focusing on local estimation of network algebraic connectivity.

Findings

Robots successfully learn to estimate network connectivity.

GNN-based coordination improves team communication efficiency.

The approach demonstrates potential for scalable multi-robot systems.

Abstract

This paper shows how Graph Neural Networks can be used for learning distributed coordination mechanisms in connected teams of robots. We capture the relational aspect of robot coordination by modeling the robot team as a graph, where each robot is a node, and edges represent communication links. During training, robots learn how to pass messages and update internal states, so that a target behavior is reached. As a proxy for more complex problems, this short paper considers the problem where each robot must locally estimate the algebraic connectivity of the team's network topology.