An adaptive chaotic hill-climbing particle swarm optimization was presented in order to overcome the unability to find all extreme points, local optimum and slow convergence speed at later time caused by Particle Swarm Optimization (PSO) in multimodal function optimization. An improved PSO was proposed , and the population of diversity was measured by entropy. A dynamic chaos mechanism was used to increase the diversity when there is a lack of population diversity, and a hill-climbing method was introduced to improve the convergence speed of PSO in later period. Four kinds of typical multimodal functions were chosen to test the performance of the improved algorithm in solving complex multimodal function optimization problems. The results show that the improved algorithm has better performance than the existing algorithms.