Improved Particle Swarm Optimization Based on Natural Flocking Behavior

摘要

Nature-based particle swarm optimization (NBPSO) is a technique which improves the performance of particle swarm optimization by using happenings in nature. It utilizes the concept of mature particles, which has the decisive ability to find out the solutions. In this paper, NBPSO is used for solving multidimensional and multimodal problems very easily, which are difficult to solve by other techniques. This new technique is proposed to move the swarm out of the stagnation region, by avoidance of taking reference of the global best particle, which causes the stagnation. The proposed technique also considers the direction of randomly selected two particles, which gives better acceleration to move away from the stagnation region. The algorithm is tested for 300 dimensions on 13 unimodal and multimodal functions from the test suit provided in AGPSO. Performance of NBPSO is compared with AGPSO1, AGPSO2, AGPSO3, IPSO, TACPSO and MPSO. To test the scalability, the proposed method is compared with CCPSO2 upto 1000 dimensions. Results and analysis show that NBPSO is highly competitive algorithms on higher-dimensional problems.