Studying the order delivery strategy considering urban traffic restrictions to reduce the average delivery time of orders, improve the consistency of delivery time and customer satisfaction is of great significance. A steering-restricted distribution network was constructed by setting up steerable nodes. And with the goal of minimizing the average delivery time, this paper established a multi-vehicle real-time pickup and delivery path optimization model considering the dynamics of the orders based on the turn restriction network. And two delayed distribution strategies, IGNORE and wait & return（W&R), were designed to solve the problem. In the study analysis, the number and distance of steerable nodes were adjusted to simulate two kinds of road networks: dense and sparse. Through numerical simulation and genetic algorithm solution, the stability of the model and the applicability of the strategy were verified, the average delivery time of IGNORE and W&R strategies under these two networks, different order quantities, the number of delivery personnel and lengths of rolling time domains were obtained, and the reasons for the fluctuation of the average delivery time were analyzed. The results show that the IGNORE strategy is applicable to the case of small network with few orders, and the average delivery time of orders decreases as the rolling time domain time shortens. The W&R strategy is suitable for larger distribution networks, and the more the number of steerable nodes in the network, the average delivery time of orders decreases. The research conclusion has certain reference significance for the real-time delivery platform to optimize the delivery strategy and improve customer satisfaction.