Abstract: The path planning problem for car-like mobile robots is studied, and an environment modeling method is presented which builds the global compound roadmap by combining the local regular roadmap with the probabilistic roadmap. With path planning based on this model, the quality of local roads near the obstacles is improved, time consumption caused by the frequent collision-avoiding check is lowered, and the loss of feasible space is avoided. The simulation experiment shows that this method is very efficient in the car-like mobile robot path planning.