Currently, e-commerce orders are characterized by diverse needs. In traditional warehouse operations, picking such orders requires a lot of manpower. More and more e-commerce companies are using Robotic Mobile Fulfillment Systems (RMFS) to improve order picking efficiency. In this system, the storage assignment problem is the basis for a number of decision problems, and unreasonable storage assignment can seriously affect the picking efficiency of orders with diverse demand, resulting in increased logistics costs for the enterprise. Therefore, the storage assignment optimization problem is an urgent problem to be solved.Based on the complex network analysis technology, this paper selects the warehouse data of a large B2C enterprise′s RMFS for the study of storage assignment. Facing orders with diverse commodity needs in RMFS, complex network analysis technology is used to explore the correlation of commodities, and the clustering grouping using the k-faction community structure in the complex network is proposed to form a storage assignment scheme.In the first part, the paper describes the storage assignment problem of RMFS, and then gives a way to build a commodity association network using complex network theory in conjunction with historical order data. In this commodity association network, commodities are used as nodes, and the edges between two nodes represent the number of times two commodities have been purchased in association. Finally, the concept and structural form of the k-faction community used in this paper are introduced.In the second part, this paper gives the flow of the method for storage assignment based on the k-clique community structure. Firstly, the parameters required for the method are explained. Secondly, a model is constructed for storage assignment based on k-clique community structure. The core of the k-clique community structure-based assignment method is to maximize the sum of the k-clique community structure levels for each clustering result when clustering and grouping commodities using the k-clique community structure to complete the assignment. When clustering commodities in the association network with this goal, it is possible to prioritize commodities with high level k-clique community structures on the same pod. The advantage of this clustering method is that it can take into account the sales characteristics while taking into account the commodities correlation.In the third part, this paper compares the methods using the warehouse data of a large B2C enterprise with RMFS in China, and also conducts sensitivity analysis for both key parameters. Firstly, in the small-scale case, the proposed method has advantages in both the value of the objective function and the solving time, and can efficiently generate the storage assignment scheme. Secondly, taking a real RMFS as an example, the method in this paper can significantly reduce the number of pods handling when compared with the baseline storage assignment method. Compared with other methods used in the current literature, the method in this paper reduces the number of pods handling by 1.58%, 3.69% and 9.58% respectively, which verifies the effectiveness of the method. Finally, the effects of pods capacity limitation and the adoption of distributed storage mode for different levels of commodities on the effect of commodities storage assignment are explored separately. It is known that when pods capacity is 30, the enhancement effect of storage assignment effect is the best, and continuing to increase the pods capacity does not significantly enhance the effect, but will cause the waste of storage resources. At the same time, increasing the number of pods can be placed for high level community structure commodities can more effectively improve the effect of storage assignment. However, it is not advisable to increase the number of pods too much, and the excessive increase will cause the waste of pods resources or even play a counterproductive role.To conclude, in the face of the prevalence of diverse order requirements today, the method of assigning commodities storage based on product correlation in RMFS has become more important. Through complex network analysis technology, the clustering is performed by using the network community structure, which is conducive to taking into account the correlation and sales characteristics of products, and can further improve the effect of storage assignment. Finally, in the process of commodities assignment, the setting of parameters is also worthy of attention. The reasonableness of parameter setting can help avoid the waste of storage resources, and the appropriate increase in the number of commodities storage space for the high level structure of commodities in the commodity association network can optimize the effect of assignment.