In order to improve the rationality of customized bus network scheme and resolve the problem of commuting during morning peak hour, a network optimization model of customized bus was built which took both interests of operating companies and passengers into consideration and met needs of multiple picking-up and dropping-off stops and multi-type vehicles. According to application information of residents' commuting demands during morning peak hour, parking lot conditions and passenger time window requirements, two objectives were determined to be optimized, that included vehicle operation cost and waiting time cost of passengers caused by vehicles early or late arriving at boarding stations. Meanwhile, real-coded genetic algorithm was adopted to resolve the model mentioned above, which included dispatched vehicles, OD pairs carried by vehicles and paths vehicles driving. The study result of the example shows that under the condition of multi-type vehicles at parking lots and “many-to-many” operation mode,a better and more reasonable customized bus network scheme can be obtained by balancing operation cost of operators and waiting time cost of passengers.