The work aims to improve the sorting efficiency and dynamic response capability of orders for e-commerce logistics distribution centers. Based on analyzing such factors as the characteristics of retail e-commerce orders with multiple varieties, small batches and high aging, movement and load constraints of sorting robots were taken into consideration; a rolling window scheduling strategy and a high-dimensional sparse dynamic clustering algorithm were proposed, and a simulation experiment model of a large-scale e-commerce distribution center was established for data comparison analysis. The simulation experiment was carried out on 500 orders of an e-commerce enterprise in rush hours. Compared with the fixed batch sorting strategy, the optimized sorting strategy reduced the average moving distance of the robot by 66.9% and the sorting time by 23.9%. The conclusion is that the high-dimensional sparse dynamic clustering strategy effectively improves the sorting efficiency, reduces the sorting cost, and the algorithm is more open and flexible. It is of great significance for the e-commerce enterprise to reduce costs and increase efficiency of logistics business.