The application of particle swarm optimization and maneuver automatons during non-Markovian motion planning for air vehicles performing ground target search

Abstract

This paper presents a centralized receding discrete time horizon controller for use in cooperative high level unmanned air vehicle (UAV) motion planning during a search for land based targets over large ground regions with sensor occlusions. Due to the non-Markovian, dynamic, multimodal, and discontinuous objective function employed, the controller presented uses a stochastic optimizer in the form of either a particle swarm optimizer (PSO) or a comprehensive learning PSO (CLPSO) to calculate UAV routes. A hybrid maneuver automaton is applied to model vehicle dynamics and a Fisher information theory based model of discrete ground regions is used as an objective function. Finally, the non-Markovian process PSO (NMPPSO) algorithm is introduced to achieve better performance by customizing the PSO and its variants to decision processes with non-Markovian reward through the use of a suffix trie.