改进多种群遗传算法的AutoStore系统多AGV调度优化

新兴紧致密集型仓储系统AutoStore存在出、入库单独作业及联合作业并存的情况，使用传统单一作业模式下所得AGV调度方案易导致资源浪费或效率低等问题。在分析多作业模式工作流程基础上，建立多AGV任务分配模型，优化目标为系统总作业时间最短。对传统多种群遗传算法进行改进。首先，为获得具有多样性的初始解，给出适用于实数编码的初始解判断式；其次，为提高搜索效率，给出交叉、变异概率计算式，使得遗传操作能随着进化过程和适应度值变化而不同。算例分析验证所给算法的可行性与有效性，能为系统提供更优的AGV调度方案。

Multi AGV Scheduling Optimization of AutoStore System Based on Improved Multi Population Genetic Algorithm

The emerging compact-intensive storage system AutoStore has the coexistence of separate operations and joint operations for outbound and inbound operations. If the AGV scheduling scheme obtained under the traditional single operation mode is used, it is easy to cause resource waste or low efficiency. Therefore, based on the analysis of multi-operation mode process, the AGV task allocation model with the shortest total operation time of various operation modes is established, and the objective function is the shortest total operation time of the system. The traditional multi-population genetic algorithm is improved. Firstly, in order to make the distribution of the initial solution uniform, the distribution of the generated initial solution is judged. Secondly, the rule that the cross-mutation probability changes with the fitness value is given to enhance the search efficiency of the algorithm. The analysis of the example verifies the feasibility and effectiveness of the improved algorithm, which can provide a better system for the system AGV scheduling scheme.