A large scale multi-delivery centre multi-type vehicle scheduling problem is considered. Firstly, based on the method of gravity center partition, the multi-delivery centre multi-type vehicle scheduling problem is transformed into the single-delivery centre multi-type vehicle scheduling problem, the partition method is simple, and reduces the amount of computations. Then based on the minimum delivery cost, a mathematical model of multi-delivery centre multi-type vehicle scheduling problem is established; the model considers not only the basic transportation cost, but also the driver's salary expenses, including the basic wage and overtime pay. In the model of multi-delivery centre multi-type vehicle scheduling, one car is able to provide service for more customers, however a customer must be served only by one car. According these characteristics, a new chromosomal hybrid encoding scheme is put forward, and the techniques of segmented crossover and mutation is used, the method effectively retains the father's generation superiority of chromosome fragments, therefore remarkably improves the convergence speed of the genetic algorithm. The results of numerical simulation verify the feasibility and effectiveness of the algorithm.