Abstract: This paper considers the ship routing optimization problem in a hub-and-spoke network. A routing optimization model for multi-type containerships with time deadlines is established, and the target is to minimize the total cost, which consists of the total travelling cost, total service cost and total waiting cost. The model is set up through an improved genetic algorithm. The study data are from the Pearl River Delta region of China, which include 1 hub port and 29 feeder ports and have a population of 30 million. Result shows that when the iteration time reaches 190, the total cost comes to 521 thousand yuan near the optimal value. There are 6 routes, including 3 containerships of 100 TEU, 2 containerships of 150 TEU and 1 containership of 200 TEU. At the same time, in the single-type containerships case, there are 7 routes, and when the iteration time reaches 120, the total cost comes to 573 thousand yuan, which is close to the optimal value. Comparing the two cases, it shows that the model for multi-type containerships with time deadlines is reasonable, and the algorithm is practicable. In the last, three factors, which may affect the total cost to carry out sensitivity analysis are chosen. It shows that time deadline, containership capacity and cargo handling capacity of each port have significant influence on the total cost. It is also shown that the total cost for multi-type containerships is always less than that for the single-type containerships.