Reference Point-Based Particle Sub-Swarm Optimization

In this paper, a novel optimization method named reference point-based particle sub-swarm optimization (RPB-PSWO) is presented. RPB-PSWO utilizes the particle position update method of PSO and with the non-dominance and diversity selection methods of NSGA-II. The multi-objective optimizer utilizes a reference point-based system to allocate particles into an equidistant sub-swarm, in which particles are attracted to a pareto optimal solution in that sub-swarm. To encourage diversity and avoid local minima, density and turbulence factors are included. RPB-PSWO is capable of optimizing problems with many dependent variables, as the position update method of PSO inherently preserves dependent relationships, but suffers from an increased computation cost compared to NSGA-II. The proposed algorithm, although less computationally efficient, is capable of creating diverse pareto front solutions for standardized and custom optimization problems.