快速约束多目标进化算法及其收敛性

Fast multi-objective constrained evolutionary algorithm and its convergence

针对进化算法收敛速度缓慢、容易陷早熟的问题,提出了约束多目标优化问题的一种新的快速进化算法.设计了能够从可行解空间和不可行解空间同时搜索的交叉算子,将约束条件和目标结合在一起,引入一种新的偏序关系用于比较个体之间的优劣,提出一种新的Niche值计算方法作为维持种群均匀性的主要动力,并采用已搜索解集避免了算法的重复搜索.在此基础上,设计了具有全局搜索能力的进化算法,并证明了算法的收敛性.仿真结果表明,与同类进化算法相比,该算法能够快速收敛到Pareto前沿,并能很好地维持种群的多样性.

Aimed at the problems of slow pace of convergence and easy subsidence precocious problem, a new fast evolution algorithm is proposed for constrained multi-objective optimization problems. A crossover operator which search simultaneously from feasible and infeasible solution space is designed. With combining constraint condition and objective, a new partial-order relation for comparing individual is introduced. Thus, a new niche computation method for maintaining diversity of population is suggested and repeat search is avoided using searched solution space.Based on all these, a novel effective evolution algorithm for global optimization is proposed and its convergence is proved. The simulation results show that this algorithm can rapidly converge at global pareto solutions, and can maintain diversity of population, comparing with the current MOEAs.