Orthogonal Design Assisted Particle Swarm Optimization

This paper proposes an improved orthogonal design assisted particle swarm optimization (ODPSO), effectively addressing the issue of traditional PSO easily getting trapped in local optima. Constructed using the orthogonal experimental design method, the orthogonal table of ODPSO integrates information from the current position of particles, individual best positions, and global best positions, enhancing global search capabilities while reducing reliance on random initial particles, thereby avoiding premature convergence, and strengthening optimization capabilities. Across 13 benchmark functions in the continuous domain, ODPSO demonstrates faster convergence speed and higher optimization accuracy compared to traditional particle swarm optimization (PSO), orthogonal predicted genetic algorithm (OPGA), and other algorithms, particularly excelling in multimodal functions. The innovation of this study lies in proposing a novel orthogonal learning strategy and demonstrating the effectiveness of ODPSO through systematic evaluation, providing robust support for the improvement of PSO.