The Improvement Research On Particle Swarm Optimization

Particle swarm optimization (PSO) is an evolutionary computation technique developed by Dr. Eberhart and Dr. Kennedy in 1995, inspired by social behavior of bird flocking or fish schooling. Recently, PSO algorithm has been gradually attracted more attention over another intelligent algorithm. PSO is simple in concept, few in parameters, and easy in implementation. It was proved to be an efficient method to solve optimization problems, and has successfully been applied in the area of function optimization, neural network training and fuzzy control systems, etc. However, both theory and application of PSO are still far from mature.A modified niche PSO algorithm is constructed which allows unimodal function optimization methods to efficiently locate all optima of multimodal problems that the Niche PSO is unreached. In the new algorithm, the sequential niche technique is introduced. Firstly, a stretching technique is adopted in main swarm. Secondly, the dismissed mechanism is used in sub-swarms namely when a local best of value is found in sub-swarms, the sub-swarms would be dismissed and regressed to the main swarm. At last, the radius of created sub-swarms is confined in order to avoid the excessive of radius. The new stretching-NichePSO algorithm resolves the disadvantage of standard Niche PSO that the local best of value depending on the number of sub-swarms and easily have the problem of iteration and pretermission. Testing of the algorithm by using three benchmark functions indicate that the modified niche PSO has higher performance than standard Niche PSO in a better value found and a steady convergence.Based on the bionic study on dolphin, philosophy of dolphin Partner Optimization (DPO) is formulated and a so-called"nucleus"is introduced to predict the best position according to the positions and fitness of the team members. DPO is a heuristic method which performs with partner selection, roles identify and communication to determinate the roles of each dolphin in his virtual team, then the leader of the team need do more explorer to the "nucleus" and the common member will just following the pioneer. Test on several benchmark functions shows that DPO can achieve fairly better value in the beginning steps with rapid speed, and can often breakthrough the local minimum that GA always lost in. Meanwhile, the DPO has niche capability to find many minimum for multimodal functions. Benchmark function shows its accuracy.