Asynchronous distributed information leader selection in robotic swarms

This paper presents asynchronous distributed algorithms for information leader selection in multi-robot systems based on local communication between each robot and its direct neighbours in the system's communication graph. In particular, the information leaders refer to a small subset of robots that are near the boundary of the swarm and suffice to characterize the swarm boundary information. The leader selection problem is formulated as finding a core set that can be used to compute the Minimum-Volume Enclosing Ellipsoid (MVEE) representing the swarm boundary. Our algorithms extract this core set in a fully distributed manner and select core set members as information leaders, thus extending abstract centralized MVEE core set algorithms for robotic swarm applications. We consider different communication conditions (e.g. dynamic network topology) and system configurations (e.g. anonymous robots or uniquely identified robots) and present a variety of approaches for core set selection with associated proofs for convergence. Results for simulated swarms of 50 robots and experiments with a swarm of 10 TurtleBots are provided to evaluate the effectiveness of the proposed algorithms.