Abstract: Aimed at the optimization problem of multi-OD demand-response bus routes, comprehensively considering the service time window, operating expenses and passenger selection, a bi-level planning model for route optimization is constructed. The upper model aimed at minimizing the total operating cost, and the lower layer used users to balance traffic distribution Model. A greedy algorithm was designed to solve the feasible solution of the model, and the hybrid particle swarm optimization was used to find the relatively optimal solution for the feasible solution. A small case was used to prove that the solution method could find a plan that was more in line with the actual operation of the bus. In terms of the quality of the solution algorithm, compared with the result obtained by the genetic algorithm, the total operating mileage was reduced by 3 km, and the total operating cost was reduced by 146.9 yuan, the calculation time was reduced by 17 s, which could effectively optimize the bus route plan.