Particle Swarm Optimization with Hybrid Jumps for Multimodal Function Optimization ⋆

Particle Swarm Optimization (PSO) has shown good performance in many optimization problems. However, it easily falls into local optima and suffers from premature convergence on complex multimodal problems. To help trapped particles escape from local minima, a novel hybrid jumps strategy is proposed. The main idea of the new jump strategy is to monitor the changes of previous best particle and the global best particle. If the best particles are trapped into local optima, hybrid jumps will be conducted to help the particles escape. Experimental studies on six multimodal benchmark problems show that PSO with Hybrid Jumps (PSOHJ) outperforms PSO, PSO with Gaussian Jump (PSOGJ), PSO with Cauchy Jump (PSOCJ), Classical Evolutionary Programming (CEP) and PSO with Gaussian Mutation (PSO-GM). Additionally, the number of successful jumps in different stages of evolution is also investigated.