Abstract: For the accumulated errors of the odometry in the process of localization, a general odometry error model for mobile robots is established to compensate the odometry errors via feedback in real time. According to the errors of original data from the laser radar, an observation error model for laser radar is established. The observation model for mobile robots is designed according to the environmental features and the relative position of mobile robots. Further, extended Kalman filter is used by combining the odometry and laser radar error model to realize the mobile robots' localization based environmental feature tracking. The experimental results illustrate that the location accuracy of the robots is improved effectively with increasing a little locating time by introducing odometry and laser radar error model.