Large-scale multiobjective optimization with adaptive competitive swarm optimizer and inverse modeling

Abstract

Competitive swarm optimizer (CSO) is an efficient algorithm for solving larger-scale multiobjective optimization problems (LSMOPs). However, the winners are not updated in original algorithm, and the random parameters model affects the convergence speed. To improve the performance of CSO for solving LSMOPs, an adaptive competitive swarm optimizer with inverse modeling is developed. In the method, an adaptive parameter model is designed to accelerate the convergence speed when the population not traps in local convergence, and the random parameter model is used to help the losers jump out of local convergence when the population traps in stagnation. Moreover, inverse modeling is used to update the winners, and the number of them is maintained by a new designed environment selection method. The exploitation ability of algorithm is improved and the drawback that convergence speed of the algorithm might be slowed by few winners in the early stage of evolution is avoided. In addition, an adjacent individual competition method is proposed to improve the distribution of solutions. Finally, the proposed algorithm is tested on nine benchmark optimization problems, and the results indicate that the average ranks in terms of IGD and HV of the proposed algorithm outperform those of seven compared algorithms.