［Purposes］According to the characteristics of milk-run in automotive parts inbound logistics, which can help automotive company achieve automotive parts transportation in smaller batch, more frequency, a simultaneous delivery and pickup vehicle routing problem with three-dimension loading constraints was discussed to meet the demand to delivery circulation boxes and pickup automotive parts simultaneously. ［Methods］A mathematical model was established to minimize the total cost of transportation and inventory, and then a two-stage ant colony algorithm was designed to solve the model. In the first stage, the memory bank of the approximate optimal solution was obtained by the vehicle path planning sub-algorithm, and in the second stage, the feasible solution was obtained by the loading feasibility test sub-algorithm. ［Findings］The simulation results show that the pickup scheme based on the fuzzy vehicle loading constraint with simple arithmetic addition cannot be successfully loaded. By contrast, the routes of the pickup scheme with three-dimensional loading constraint increased from 2 to 3, and the vehicle loading rate reduced by 27.45%, but the loading feasibility of the scheme is ensured. ［Conclusions］The milk-run routes with fuzzy loading space constraint are not practical because of the lack of three-dimension loading constraints test. The simultaneous delivery and pickup vehicle routing scheme with three-dimension loading constraints not only meets the practical needs of recycling circulation box, but also ensures the feasibility of the three-dimensional loading.