IMOPSO: An Improved Multi-objective Particle Swarm Optimization Algorithm

摘要

An improved multi-objective particle swarm optimization (IMOPSO) is presented because of the different demand for decision and state variables in engineering optimizations. IMOPSO adopts a new method of dynamic change about acceleration coefficients based on sine transform to improve the ability of global search in early period and the local search ability in the last runs of the algorithm. To expand the search area of particles, a drift motion is acted on the personal best positions. Moreover, a dynamic mutation strategy in which the mutation rates are generated by modified Levy flight is used to make the particles escape from the local optimal value. Finally, the efficiency of this algorithm is verified with test functions and the experimental results manifest that the IMOPSO is superior to MOPSO algorithm in wide perspectives like obtaining a better convergence to the true Pareto fronts with good diversity and uniformity.