PARTICLE SWARM OPTIMIZATION BASED ON PARTICLE MEAN DIMENSIONS WITH ELIMINATING VELOCITY COMPONENTS

Abstract

Premature convergence is the chief difficulty in solving hard optimization problems for most particle swarm optimization variants. To address this issue, a particle swarm optimization based on particle mean dimension with an eliminating velocity component algorithm is proposed in this thesis. In the proposed algorithm, every swarm updates its position by eliminating velocity components based upon cognitive, social, and particle mean dimension information, and the key aspect used here is that these parameters are no longer assumed to be accelerating components but rather position components to enhance the global search capability and avoid premature convergence. This strategy could make particles fly in a better direction by discovering useful information from the entire particle mean dimension. The idea of the method is to improve the solution quality, fast convergence, and robustness of particle swarm optimization. The proposed algorithm is coded in MATLAB R2021a and evaluated on fourteen benchmark test functions for unconstrained optimization which have different characteristics. The experimental results show that the proposed algorithm has superior performance compared to the previously reported results of the three other particle swarm optimization algorithms. This suggests that the particle swarm optimization based on particle mean dimension with the elimination of velocity components algorithm has the ability to jump out of local optimums and achieve the global optimum efficiently.