A Particle Swarm Optimization Based on Dynamic Parameter Modification

A new particle swarm optimization based on dynamic parameter modification is proposed in this paper (Dynamic Parameter Modification Particle Swarm Optimizer, DPSO). In DPSO algorithm, w is doing oscillating decay breaking through the constraint of topical linear decreasing, and the Euclidean distance vertical bar p(i) - x(i)(t)vertical bar, and vertical bar p(g) - x(i)(t)vertical bar is calculated, which respectively stand for the Euclidean distances form the position X-i, of particle i to the best position P-i that the particle has passed and the best position that all the particles have passed under the time t. Parameters c(1) and c(2) of topical PSO are modified dynamically based on the comparison of vertical bar p(i) - x(i)(t)vertical bar, and vertical bar p(g) - x(i)(t)vertical bar in order to coordinate between global search and local search. Then find out the optimal value of Goldstein-Price function using topical PSO and the improved DPSO respectively, and the results demonstrate that compared to topical PSO, DPSO algorithm avoids falling into the local minimum and improves the search efficiency.