A corrective direction particle swarm optimization for large-scale multi-objective optimization

In multi-objective optimization, Large-Scale Multi-Objective Optimization Problems (LSMOPs) involve a large number of decision variables and present significant challenges in finding satisfactory results with a limited number of function evaluations. To address this issue, a Corrective Direction Particle Swarm Optimization (CDPSO) is proposed for large-scale multi-objective optimization in this paper. The core idea of CDPSO is to maintain the correctness of the search direction of the particles in the decision space to improve convergence speed. Specifically, a competitive learning strategy that is employed by normal particles in the decision space to select elite particles with the smallest directional deviation improves the population’s ability to discover the Pareto optimal solutions. Moreover, a novel exploration mechanism is used by the elite particles, which spreads the exploration across each dimension of the decision variables and adaptively adjusts the velocity components in each dimension to enhance the algorithm’s performance further. Experimental results show that our proposed CDPSO exhibits strong competitiveness compared to the other nine state-of-the-art algorithms when tested on both benchmark problems and real-world problems.