Enhancing comprehensive learning particle swarm optimization with local optima topology

Recently, particle swarm optimization (PSO) has been employed in many studies for solving numerous real-world problems. However, PSO may suffer from premature convergence when dealing with multimodal problems. Thus, we propose a local optima topology (LOT) structure based on the comprehensive learning particle swarm optimizer (CLPSO) called CLPSO-LOT. The local optima are found in the iterative process and a new topology space is composed. A random element from the space can serve as the next exemplar that the particle uses for learning. This topology structure comprises the local optima that enlarge the particle's search space and increase the convergence speed with a certain probability. We conducted numerical experiments based on various functions from CEC2005 and CEC2014, where the results demonstrated good performance of this algorithm. Furthermore, we applied the algorithm to the optimization of four-bar linkages, where the results indicated that the CLPSO-LOT performed better than other algorithms, and that the performance of the CLPSO was improved.