Abstract: As a new kind of evolutionary algorithm, Differential Evolution（DE） algorithm with the characteristics of remembering individual optimal solution and information sharing can be regarded as a real coded and excellent security greedy genetic algorithm. To solve the Vehicle Routing Problem（VRP）, which belongs to NP problems, the paper puts forward an improved differential evolution algorithm. A greedy algorithm is used to generate the initial population, legalized method is used to repair mutation, improved order crossover is used, then, after the mutation operator, a new selection mechanism is added in. The new algorithm is implemented in Matlab, the experimental results show that the improved differential evolution algorithm can efficiently solve the VRP.

Key words: Differential Evolution（DE）, Vehicle Routing Problem（VRP）, greedy algorithm, NP problem, evolutionary algorithm

摘要： 差分进化算法是一种具有记忆个体最优解和种群内部信息共享的特点的新型进化算法，本质上可看做是一种基于实数编码的、具有保优思想的贪婪遗传算法。针对具有NP难的车辆路径优化问题，提出了一种改进的差分进化算法。利用贪心算法产生初始种群，定义合法化修复变异个体的方法，采用改进的顺序交叉，并在变异操作之后，加入新的选择机制。使用Matlab进行了算法的实现，实验结果表明了改进DE算法能够高效地解决VRP问题。

关键词: 差分进化算法, 车辆路径问题, 贪心算法, NP问题, 进化算法