Performance-dependent Adaptive Particle Swarm Optimization

The swarm collective behaviors, such as birds flocking and fish schooling, are complex, dynamic and adaptive processes, in which the differences among individuals play an important role. As a new swarm intelligent technique, the standard particle swarm optimization only provides a simple uniform control, omitting the above mentioned phenomenon entirely. Thus, a new modified version: performance-dependent adaptive particle swarm optimization incorporated with personal differences, is designed. It uses each particle’s adaptation score – fitness value of the current position to represent the effect dominated by differences, and guides the inertia weight of each particle to adjust its’ value adaptively. Furthermore, three adaptive adjustment strategies are discussed. Simulation results show the new version is effective and efficient, although the improved performance is problem-dependent.