Population Diversity Based Inertia Weight Adaptation in Particle Swarm Optimization

In this paper, we propose two new inertia weight adaptation strategies in Particle Swarm Optimization (PSO). The two inertia weight adaptation strategies are based on population diversity. In the search process of an optimization algorithm, there must be a balance between exploration and exploitation. Exploration means to explore different areas of the search space in order to have high probability to find good promising solutions. Exploitation means to concentrate the search around a promising region in order to refine a candidate solution. The exploration and exploitation are two conflicted objectives of an optimization algorithm. A good optimization algorithm should optimally balance the two conflicted objectives. With the first strategy, the algorithm focus on the exploration at the beginning of the search, and focus on exploitation at the end of search. Particles' inertia weights are randomly initialized within the range [0.4, 0.9], the minimum weight gets increased at the beginning of search to enhance the exploration ability, and the maximum weight gets decreased at the end of search to enhance the exploitation ability. With the second strategy, particles' inertia weights are set at the same value, and this value is adaptively changed according to the distance between the gbest and the centre of swarm. The porposed PSOs are compared with the standard PSO. Experimental results show that a PSO with adaptive inertia weight could obtain performance as good as the standard PSO, and even better on some multimodal problems.