Optimizing multi-agent dynamics for underwater tactical applications.
Access rightsWorldwide access.
Access changed 6/26/13.
Yu, Albert R.
MetadataShow full item record
Large groups of autonomous agents, or swarms, can exhibit complex emergent behaviors that are difficult to predict and characterize from their low-level interactions. These emergent behaviors can have hidden implications for the performance of the swarm should the operational theater be perturbed. Thus, designing the optimal rules of operation for coordinating these multi-agent systems in order to accomplish a given task often requires simulations or expensive implementations. This thesis project examines swarm dynamics and the use of inversion to optimize the rules of operation of a large group of autonomous agents in order to accomplish missions of tactical relevance: specifically missions concerning underwater frequency-based standing patrols and point-defense between two competing swarms. Modified genetic algorithms and particle swarm optimization are utilized in the inversion process, producing various competing tactical responses and patrol behaviors. Swarm inversion is shown to yield effective and often creative solutions for guiding swarms of autonomous agents.