A novel mobile robot localization approach based on a model switching feature extraction

This paper studies natural feature based localization for mobile robot navigation in semi-structured outdoor environments using a laser range sensor. We propose an algorithm for feature extraction by using switching models between line model and circle model. In order to avoid the estimation error caused by the linearization in the extended Kalman filtering (EKF), a particle filter is applied to realize the prediction and validation process by integrating data from both the laser range sensor and encoders in outdoor environments. The proposed feature extraction and localization algorithms are verified in a artificial simulation environment. The results show that the proposed algorithms perform very well in an semi-structured outdoor environment.