Graph Neural Networks for Learning Robot Team Coordination
This work addresses coordination challenges in connected robot teams, but it is incremental as it applies an existing method to a new domain-specific problem.
The paper tackles the problem of distributed coordination in robot teams by using Graph Neural Networks to model robots as nodes and communication links as edges, enabling robots to learn message passing and state updates to achieve target behaviors, with a specific application to locally estimating the algebraic connectivity of the network topology.
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.