Based on the scene of ``parts-to-picker'' warehouse system, the order batching problem of autonomous vehicle storage and retrieval system is studied by considering the quantity of every item in orders and SKU bins. Given a set of orders to be picked, the order batching problem is formulated into an integer programming model, the goal of the model is to minimize the number of SKU bins to be transported. A two-stage heuristic algorithm is designed based on the clustering idea. Simulation experiments are carried out on different size of examples to verify the effectiveness of the model and algorithm. By comparing with the order batching results obtained by the first-come-first-served strategy, we find that the picking efficiency can be improved by 25\%--45\% based on the batching results obtained by the model and algorithm of this paper. Furthermore, the sensitivity analysis is done on parameters of the picking station's capacity and the weighted coefficient of similarity. Finally, the relationship between the order batching model considering the quantity of items in orders and the order batching model without considering the quantity of items in orders is compared and analyzed by using two examples. And the superiority of the order batching model considering the quantity of items in orders is verified