This paper proposes a new SLAM (Simultaneous Localization and Mapping) algorithm based on hybrid map method. We express the environment surrounding mobile robot with a grid and a feature map. Using the reliability of estimation for individual map, we calculate the importance factor for Rao-Blackwellized Particle Filter (RBPF) resampling. In this way, we improve the accuracy of the algorithm and reduce computational complexity. Experimental results verify the feasibility and effectiveness of our algorithm.