수렴 성능이 개선된 차원 교환 입자 군집 최적화 기법 설계

Abstract

The particle swarm optimization technique, one of the widely used optimization techniques, allows individual particles to move to the optimum position, using other particles experiences. For this purpose, particles memorize their experiences from the optimum position to which they moved by themselves and share this with other particles. In addition, the technique adjusts the balance between global search ability and local search ability of the particles through the adjustment of parameters. The particle swarm optimization technique, as compared to a genetic algorithm widely used as a stochastic algorithm, has strengths that it has an excellent ability to find the global optimum, the operation for the application of the algorithm is very simple and the implementation is easy. However, the particle swarm optimization technique has weaknesses that it is at high risk of falling into the local optimum and has a slow speed of convergence towards the local optimum. This study presented a new dimension exchange particle swarm optimization technique, adding a swarm exchange operation and a one-dimensional exchange operation to the existing particle swarm optimization technique. A new particle swarm was generated through the swarm exchange operation to allow a fast search for the global optimum. A new particle swarm was generated through the primary exchange operation to resolve the problem that the particle falls into the local optimum and allow accurate convergence towards the global optimum. In order to demonstrate the performance of the dimension exchange particle swarm optimization technique proposed in this study, the results of simulations using benchmark functions were presented, which were compared with the results of the simulations by the existing particle swarm optimization technique. The results of an application of the dimension exchange particle swarm optimization technique proposed in this study to the decision of the PID controller parameters of BLDC motors were presented, which were compared with the results of an application of a genetic algorithm and the existing swarm optimization technique. In addition, the results of an application of the dimension exchange particle swarm optimization technique proposed in this study were presented, for the path search of mobile robots with swarm intelligence, which were compared with the results of an application of the existing particle swarm optimization technique. Through a series of simulations using benchmark functions mentioned above and the result of applications to BLDC motors and mobile robots, it was proven that the performance of dimension exchange particle swarm optimization technique proposed in this study was excellent.