To solve the low estimating accuracy of SLAM (simultaneous location and mapping) caused by particle degradation in fast simultaneous location and mapping algorithm, a FastSLAM algorithm which combine fading adaptive unscented particle filtering and Gaussian distributed re-sampling is proposed. An adaptive proposal distribution was generated by combining fading filtering and unscented particle filtering, and the high-weight particles were dispersed by Gaussian distribution to get new particles. The motion model and observation model of robot were established, and the performance was tested in simulation environment. The simulation result shows that, the algorithm can effectively alleviate particle degradation, increase system stability and improve SLAM estimation accuracy.