A multi-strategy particle swarm optimization framework based on deep reinforcement learning

Abstract

The particle swarm optimization (PSO) algorithm is a well-known optimization algorithm that has shown good performance in solving engineering problems. However, the performance and convergence speed of the PSO algorithm is easily affected by the parameter settings. In this paper, we propose an adaptive parameter optimization framework (APOF) for the PSO algorithm by using the Deep Deterministic Policy Gradient (DDPG) of deep reinforcement learning. In order to achieve better optimization effect, the strategy group is extracted from the APOF, so that the APOF can be combined with more strategies to improve the searchability of the optimized algorithm. This paper also improves the PSO algorithm and proposes the hybrid cluster PSO algorithm (HCPSO) as the built-in algorithm of the APOF. In the experiment, twenty-one functions are selected to implemented, and the optimization effect of the APOF algorithm is tested. The experimental results show that the APOF has a good optimization effect and scalability, and the built-in HCPSO algorithm also achieves good performance.