两阶段动态差分智能元胞机算法

Two-stage dynamic differential agent cellular automata algorithm

针对传统进化算法在求解高维度优化工程问题时存在全局搜索和局部寻优的平衡难题,提出一种基于差分进化和元胞种群拓扑结构的两阶段动态差分智能元胞机算法。通过对个体的邻居结构进行调整,实现进化种群由结构化种群过渡到非结构化种群的效果,较好地兼顾全局搜索和局部寻优之间的协同问题;同时对外部种群保留的对象进行调整及完全反馈,提高算法的收敛速度。算法将智能体机制引入元胞种群,采用两阶段的外部种群多样性维护方法,将扰动因子引入变异操作使其跳出局部最优困境。通过对WFG系列基准函数测试表明,新算法相对于其他4种典型算法能获得更好的Pareto前端和竞争性的收敛结果。

To solve the balance difficulty of global search and local optimization for solving high dimensional optimization problems by traditional evolution algorithm, a two-stage dynamic differential agent cellular algorithm based on differential evolution and cellular population topology structure was proposed. By adjusting the individual neighboring structure, the effect of evolutionary population from the structured population to the unstructured population was achieved, and the algorithm could better balance the global search and local optimization. At the same time, the adjustment and complete feedback of the left individuals of the external population improved the convergence speed of the algorithm. The algorithm introduced the agent mechanism into the cellular population, adopted the two-stage external population diversity maintenance method, and introduced a disturbance variable in the mutation operation to avoid the algorithm falling into local optimum. Tests on the WFG series of benchmark functions showed that the new algorithm could obtain a more uniform Pareto front and competitive convergence results when compared to the other four typical algorithms.