A Distributed Boundary Detection Algorithm for Multi-Robot Systems

We describe a distributed boundary detection algorithm suitable for use on multi-robot systems with dynamic network topologies. We assume that each robot has access to its local network geometry, which is the combination of a robot's network connectivity and the positions of its neighbors measured relative to itself. Our algorithm uses this information to classify robots as boundary or interior in one communications round, which is fast enough for rapidly changing networks. We use the local boundary classifications to create a robust boundary subgraph, and to determine if the boundary is an interior void or the exterior boundary. A proof of the key property of the boundary detection algorithm is provided, and all the algorithms are extensively tested on a swarm of 25-35 robots in rapidly changing network topologies.