An improved parallel genetic algorithm based on particle swarm optimization and its application to packing layout problems

Packing layout problems belong to NP-Complete problems theoretically. They are concerned more and more in recent years and arise in a variety of application fields such as the layout design of spacecraft modules, plant equipments, platforms of marine drilling well, shipping, vehicle and robots. The algorithms based on swarm intelligence are relatively effective to solve this kind of problems. But usually there still exist two main defects of them, i.e. premature convergence and slow convergence rate. To overcome them, an improved parallel genetic algorithm based on particle swarm optimization (PSO-PGA) is proposed on the basis of traditional parallel genetic algorithms (PGA). In this algorithm, parallel evolution of multiple subpopulations based on improved adaptive crossover and mutation is adopted. And more importantly, in accordance with characteristics of different classes of subpopulations, different modes of PSO update operators are introduced. It aims at making full use of the fast convergence property of particle swarm optimization (PSO). The proposed arithmetic-progression rank-based selection with pressure can prevent the algorithm from premature in the early stage and benefit accelerating convergence in the late stage as well. An example of packing layout problems shows the proposed PSO-PGA is feasible and effective.