Research And Implementation Of Commodity Shelf Strategy Based On Network Representation

In recent years,with the rapid development of E-commerce,huge numbers of orders with a small numbers of goods in each order are often encountered in nowadays warehouse management.In order to improve the user experience and efficiency of processing orders,auto robots are often used to help human do trifle things.And the traditional human-oriented good location assignment strategy is not suitable for efficiently processing these massive orders.Therefore,in this paper,we studied various good location assignment strategies with "order processing efficiency" as the evaluate indicators,including common-used machine learning methods.This paper mainly studies good location assignment strategies by the following four steps:1)We define good location assignment formally,and prove the factor that affects the efficiency of robots-involved warehouse is the Rack Moved Number(RMN)by auto robots when processing a batch of orders rather than good retrieval speed,space usage rate,etc.in traditional human-only warehouse.2)A natural sequence of history orders storage location assignment strategy is proposed.Firstly,all goods are arranged as a long sequence of history orders;then the sequence is divided according to the capacity of racks;finally,the goods are sequentially put into racks.The experimental results show that this algorithm has low computing complexity and small RMN on test data.3)A clustering based good location assignment strategy is proposed.Firstly,the relation graph between the goods is constructed according to the historical orders;then the vector representation of all goods is obtained through the network embedding method;finally,the K-means clustering method is used to find the strong related goods and the strong related goods are placed in the same rack.The results show that the clustering strategy based on network representation is better than the random location assignment strategy.4)A genetic algorithm based good location assignment strategy is proposed.Firstly,RMN is used as the fitness function of genetic algorithm.Then in order to find a better solution,the natural sequence of history orders location assignment is used as the seed to initialize the chromosome population.In addition,a specific crossover and mutation operation is defined to avoid checking the validity of the chromosome by exchanging the values of gene pairs.The experimental results show that the genetic algorithm performs better than other methods.