Parallel particle swarm optimization (PPSO) on the coverage problem in pursuit-evasion games

A Parallel Particle Swarm Optimization (PPSO) algorithm using MPI is implemented to solve the coverage problem of pursuit-evasion (PE) games where multiple pursuers need to cooperate to cover an agile evader's possible escape area within reasonable time. The area to be covered is complex and thus difficult to calculate analytically. With the use of PPSO, maximum coverage is achieved in less time, given the minimum number of pursuers. The computation time can be further reduced by optimizing the fitness function based on data locality. In addition, using variable length of communication data frame performs better than fixed length in reducing inter-process communication time when the number of processors increases (more than four in the test example). Simulation results show a comparison of the speedup, the computation time before and after optimizing the fitness function, and communication time between fixed and variable data frame. Pursuers' positions and orientations are also presented to show the effectiveness of the PPSO algorithm.