The performance of the path planner based on probabilistic roadmap method often degrades seriously because of the irrational distribution of roadmaps, when narrow passages exist in the robot’s configuration space. Therefore, a hybrid roadmap sampling strategy, hybrid roadmap planner based on narrow passage recognition is proposed. This strategy recognizes narrow passages by using the proposed randomized star builder to increase the roadmap density within narrow passages and derive rational distribution of roadmaps, which improves the efficiency of path planning. The experiments in 2D and 3D configuration space show the effectiveness of the proposed algorithm.