Probabilistic swarm guidance for collaborative autonomous agents

This paper extends a probabilistic guidance approach for the coordination of swarms of autonomous agents, which contain sub-swarms with different mission objectives. The main idea of probabilistic guidance is to drive the swarm to a prescribed density distribution in the configuration space. Both non-collaborative and collaborative swarm guidance methods are considered. In the non-collaborative case, the probabilistic approach is decentralized and does not require communication between agents. Agents make statistically independent probabilistic decisions based solely on their own state, which ultimately guides the swarm to the desired density distribution. The probabilistic guidance idea is then extended to the collaborative case where there are sub-swarms of agents with different desired distributions. In this case, agents collaborate to decide on a common objective. This introduces collaboration at a higher level of decision making, and makes useful connections with consensus theory for networked agents. Our main result establishes that, under quite general conditions, the collaborative swarm behavior converges to the weighted average of the sub-swarm desired distributions. The formal development of this result now allows us to consider complex swarm behaviors in a mathematically rigorous framework.