An improved particle swarm optimization for global optimization

This paper introduces an improved particle swarm optimization algorithm for continuous global optimization problems. Different from the traditional PSO, the proposed PSO adjusts dynamically the update modes of particles in terms of the population density where a new method based on the center of the particles is defined to measure the population density. Initial candidate solutions are generated by uniform design, and a local PSO search mechanism is designed to keep the diversity of swarm. In addition, an elite acceleration strategy is also introduced into the proposed PSO to improve the algorithm's intensification ability further. Using those operators, the proposed algorithm can get good balance between exploitation and exploration. The experimental results on unimodal and multimodal standard test problems demonstrate the good performance of the proposed PSO.