Improved On-Line Pose Estimation for Mobile Robots by Fusion of Odometry Information and Environment Map

A new method for on-line pose estimation in mobile robotics applications is presented in this paper. Using the information provided by the wheel encoders usually is accompanied with additive systematic and nonsystematic errors. Much effort has been done to overcome this problem. In the proposed approach, it is assumed that the robot is making an occupancy grids map of the environment during exploration. Some parameters are defined for compensation of the non-systematic error, existing in the estimated value by odometry information. These parameters will be gradually estimated and tuned by a maximum likelihood estimation method. In this estimation the similarity between the current map of the environment and the local map, generated by the sensors, is considered as a likelihood ratio. Some experimental results by Khepera are presented. The results show that the performance of our method increases even compared to the case where the wheels parameters are improved by the famous UMB benchmark.