基于多层前馈神经网络仓储拣货优化问题的研究

Research on the Optimization of Warehouse Picking Based on Multi-Layer Feed Forward Neural Network

摘要: 仓储拣货问题中，实时规划拣货路线一直是难点。由于货架之间互相形成障碍，距离难以用欧氏距离给出，且已知距离后路线规划缺乏行之有效的方法。本文分为两部分，首先采用曼哈顿距离和Kronecker delta给出货架间的距离，将拣货路线转换为TSP (旅行商)问题。第二部分采用性能高、训练快等优点的神经网络智能算法，基于多层前馈神经网络的自适应反馈，形成结果反馈控制和闭环，实现求解TSP问题过程的循环优化。结果发现上述方法简化了拣货路径计算，得到了拣货单的最优拣货顺序及最小距离。

Abstract: In the problem of warehouse picking, it is always difficult to plan the picking route in real time. Due to the barriers formed between shelves, distance is difficult to be given by Euclidean Metric, and route planning after known distance lacks effective methods. This paper is divided into two parts. First, Manhattan distance and Kronecker Delta are used to give the distance between shelves, and the picking route is converted into TSP problem. The second part adopts the neural network intelligent algorithm with advantages of high performance and fast training, and forms the result feedback control and closed-loop based on the adaptive feedback of the multi-layer feed forward neural network, so as to realize the cyclic optimization of the process of solving the TSP problem.