Abstract: Based on the stochastic environment, an actual bus scheduling optimization method is proposed. Considering bus lines operating environment of random factors, a bi-level programming model was proposed to formulate the bus scheduling problem (BSP), with the upper deck upgrading the passenger's satisfaction degree as the main function, and the lower deck assuring the enterprise its profit. A intelligence algorithm to find the optimal scheduling strategy was proposed. The experiments demonstrate that the model has proved its validity and feasibility actually. Combined bus model, departure time interval, operating costs and other factors, it can quickly give optimal results based on the statistical data of the actual site.