A Hybrid of Nelder-Mead Simplex and Attractive-Repulsive Particle Swarm Optimization

• Main Authors: Chien-Peng Yu, 游鈐芃
• Other Authors: Juan-Chen Huang
• Format: Others
• Language: zh-TW
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/25048816048508062536

碩士 === 國立臺灣海洋大學 === 商船學系所 === 98 === In this study, a hybrid method of Attractive and Repulsive Particle Swarm Optimization (ARPSO) and Nelder and Mead simplex method for global optimization algorithm was developed. The proposed algorithm incorporates strategies for search intensification and diversification. The particles of ARPSO can adjust the aggregation or dispersion directions of particle movement based on particle distribution in space. Comparison with original PSO method, ARPSO has more powerful global search capability. The Nelder and Mead direct search method is a simplex method which don't need gradient calculation, and with fast convergence properties, but applying for local optimization problems. Combining the above search methods, the hybrid ARPSO+NM search method for global optimization was implemented. The accuracy and effectiveness of ARPSO+NM method were verified by some benchmark problems. And also, the ARPSO+NM method was applied to both steady the dynamic shortest path planning. It is shown that ARPSO+NM method retain the advantages of both ARPSO and NM methods.