Particle Swarm Optimization algorithm is based on swarm intelligence to search optimization algorithm, which is simple in principle, easy to understand, not dependent on the actual problem, strang incommonality, and easy to implement. The algorithm has been proposed to become a hotspot, and it can be applied to many fields.In the paper, the Particle Swarm Optimization’s principles, major steps and characteristics are introduced, two search modes of the algorithm are analyzed, the advantages and disadvantages of each mode are pointed out. Improvements of the algorithm are summarized from different angles.The inertia weight is an important parameter in the velocity update formula of the Particle Swarm Optimization algorithm, which controls the degree of influence of the contemporary speed to the next generation and plays a role of balancing global search and local search. In this paper, in the iteration process, the inertia weight is decreased linearly at the previous stage which facilitated a global search and decreased nonlinearly at the later stage which facilitated a local search. At the same time, in order to avoid groups into local optimum, the local version of PSO search mode was adopted. The population was divided into several subgroups, each subgroup in accordance with the improvements of the inertia weights searched independently, search to the optimal value of each sub-groups, the individual optimal values composed the initial population to search the final optimal value. The improved strategies are validated through several test functions.The general vehicle routing problem is resolved with the improved algorithm, using a typical case of the vehicle routing problem to test the improved algorithm. The results of Genetic Algorithms and other improved Particle Swarm Optimizations are compared in the paper, which shows that the proposed algorithm is effective. |