Research On Multi-Objective Particle Swarm Optimization Based On Competition Mechanism

Many engineering problems and scientific research in the real world can be ed as multi-objective optimization problems.There are many methods can be used to solve the multi-objective optimization problems,e.g.multi-objective genetic algorithms,multi-objective particle swarm optimization algorithms and multi-objective differential evolution algorithms.Multi-objective particle swarm optimization has attracted a lot of interests in dealing with the multi-objective optimization problem,due to the fast convergence speed and simple implement of particle swarm optimization.Most multi-objective particle swarm optimization algorithms can generally achieve a good performance,while for the complex problems with many local optimum,they are easy to falling into the local optimum because of the influence of the best solution.Besides,for the multi-objective optimization problems with a large number of decision variables,these algorithms cannot obtain the efficient information when they face a high dimensionality search space.In order to improve the performance of multi-objective optimization algorithm,we introduce the competition mechanism to balance the convergence and diversity of algorithm.Therefore,in this thesis,we proposed a competitive mechanism based multi-objective particle swarm optimizer with fast convergence for handling those problems with many local optima.Meanwhile,we proposed a particle swarm optimization based efficient search for large-scale multi-objective optimization.The main research of this thesis is shown as follow:(1)This thesis proposed a competitive mechanism based multi-objective particle swarm optimizer with fast convergence.In this thesis,we replaced the personal and global best particles with winner to update the particles for accelerating the convergence speed.In addition,the proposed algorithm does not need any external achieve to record the historical information in the search process,whereas,it just updated the particles by using the current swarm and made the algorithm clear.Compared with the three multi-objective evolutionary algorithms and three multi-objective particle swarm optimization algorithms,the experimental results demonstrate the proposed algorithm has a promising performance.(2)Based on the competition mechanism,this thesis proposed particle swarm optimization based efficient search for large-scale multi-objective optimization.The proposed algorithm designed a new updating method by using a pre-process strategy to accelerate the convergence speed of the swarm and make the swarm fast converge to the true Pareto front.Besides,the competition mechanism in the proposed algorithm is used to ensure the diversity of swarm and improves the quality of solution set.Compared with four popular large-scale multi-objective optimization algorithms,the proposed algorithm in this thesis shows good convergence and diversity.