Planner-Guided Robot Swarms

Abstract

Robot swarms have many virtues for large-scale task execution: this includes redundancy, a high degree of parallel task implementation, and the potential to jointly complete jobs that a single agent could not do. But because of their distributed nature, robot swarms face challenges in large-scale coordination, task serialization or ordering, and synchronization. We investigate the use of a central automated planner to guide a robot swarm to perform complicated, multistep operations normally beyond the capabilities of purely decentralized swarms. The planner orchestrates the actions of task groups of agents, while preserving swarm virtues, and can operate over a variety of swarm communication and coordination modalities. We demonstrate the effectiveness of the technique in simulation with three swarm robotics scenarios.