Multi-objective Particle Swarm Optimization Guided by Global Diversity and Cross-generation Competition

The application of multi-objective evolutionary algorithms in multi-objective optimization problems often faces challenges of diversity and convergence. A multi-objective particle swarm optimization algorithm based on the global diversity and cross-generation competition is proposed, and the cross-generation elite competition strategy is designed to guide the particles to the Pareto frontier. Introduce an efficient elite guidance mechanism based on global diversity, and select efficient elite guidance particles to form a new velocity guidance direction by evaluating the current algorithm diversity, and enhance the global search capabilities and local development capabilities of particles. When the algorithm diversity is good, select the elite particles with better convergence to guide and accelerate the algorithm convergence; when the algorithm diversity is not good, select the elite particles with better diversity to guide and increase the algorithm diversity. The experimental results show that under the same test problem, the reverse generation distance and hypervolume of this algorithm are significantly improved compared with other existing algorithms, and the obtained feasible solutions are more evenly distributed to the Pareto front.