Swarm-Genetics: A Hybrid PSO-GA Regeneration Model for Global Optimization Benchmark Problems

Abstract

Particle Swarm Optimization (PSO) is widely used in global optimization due to its simple structure and fast convergence, but it may suffer from premature convergence in multimodal search spaces. This study proposes Swarm-Genetics, an iteration-wise hybrid PSO-GA regeneration framework that combines PSO-based particle movement with Genetic Algorithm operators. In each iteration, particles are first updated using PSO velocity and position equations, then regenerated through crossover and mutation, followed by the selection of the best particles for the next iteration. The proposed method was evaluated on fourteen benchmark functions and compared with standard PSO and GA using mean fitness values. The results show that Swarm-Genetics achieved the lowest mean fitness values across the tested benchmark functions, with several cases producing mean errors close to zero, such as  and  It also obtained a lower mean value on the Schwefel function than both baseline methods, indicating better exploration in a complex multimodal landscape. These findings provide descriptive numerical evidence that genetic regeneration can improve PSO search performance by enhancing exploration while maintaining exploitation-oriented swarm guidance.