A Novel Multi-objective Particle Swarm Optimization Based On Crowding Angle Maintenance

The traditional multi-objective particle swarm optimization algorithm, despite being easy to implement and converging quickly, often faces challenges in achieving sufficient convergence and diversity in solutions when applied to complex optimization problems. To address this problem, this study proposes a novel multi-objective particle swarm optimization algorithm based on crowding angle maintenance. Initially, the algorithm employs a nonlinear decreasing strategy for updating the particle's inertia weight and cognitive learning coefficient, while the social learning coefficient is updated using an increasing strategy. This dual strategy aims to improve the particles' capability to explore the solution space globally during the initial phase and to refine their local search capability during the later phase of the optimization process. Furthermore, the algorithm introduces a diversity maintenance strategy for the archive based on the crowding angle, to achieve a uniform distribution of the approximate Pareto front and thus maintain the diversity of solutions during the optimization process. Additionally, evolutionary search is applied to the archived Pareto optimal solutions to generate superior solutions, thereby effectively addressing potential deficiencies in the search process. The algorithm is assessed by comparing its performance to five advanced multi-objective optimization algorithms across 21 benchmark problems. The experimental findings indicate that the introduced algorithm exhibits superior performance in terms of both convergence and diversity of solutions, demonstrating a clear advantage compared to the other five algorithms.