As crucial components of logistics and distribution in e-commerce enterprises, the coordination and optimization of trunk and branch routes have significant importance in cost reduction and efficiency improvement. Based on the operational characteristics of trunk and branch logistics, the two-echelon vehicle routing problem involving soft time windows, multiple depots, multiple commodities and split delivery is extracted, and a mixed integer programming model is established with the objective of minimizing vehicle routing cost, fixed cost and penalty cost (violation of delivery time requirements). Employing a greedy idea focusing on dispatch frequency and transportation distance savings, an efficient heuristic algorithm is designed and validated through experiments involving 27 small-scale instances and 8 larger-scale instances. The experimental results show that: 1) The proposed model can solve instances involving 5 distribution centers, 20 satellites, and 4 commodities within 2 hours by using optimization software. The proposed heuristic algorithm efficiently produces feasible solutions for large-scale problems in a short time, demonstrating superior solving capability and efficiency compared to CPLEX. 2) The complexity of the problem increases sharply with the increase of the scale of the instance, and the number of satellites has the greatest impact on the complexity of the problem, followed by the number of city distribution centers, and the impact of the number of commodity types is the smallest. 3) For small-scale instances, linehaul distribution costs account for a relatively large proportion. With the increase of the scale of the instance, the number of vehicles in the delivery level increased significantly, which leads to a significant increase in the distribution cost of the delivery level. 4) After increasing the maximum working hours of vehicles, the number of vehicles in the linehaul level decreases significantly, and the total cost is significantly reduced. With the relaxation of the capacity limit in the delivery level, the number of vehicles used in the delivery level was significantly reduced. The research results can not only expand the existing theory of two-echelon vehicle routing problem, but also help e-commerce enterprises to achieve efficient and collaborative operation of multi-level distribution networks. In addition, by optimizing the distribution schedule and reducing the number of dispatches, it will help alleviate the pressure on urban traffic and improve the urban traffic environment.