Simultaneous localization and mapping (SLAM) based on pseudolinear measurement model with a bias reduction approach

摘要

This paper describes an improved solution to mobile robot localization and map building problem based on pseudolinear measurement model through bias reduction approach. Accurate estimation of vehicle and landmark states is one of the key issues for successful mobile robot navigation if the configuration of the environment and initial robot location are unknown. Direct linearization of nonlinear models has always come up with information loss of original nonlinear models. A state estimator which uses linearized models can be a solution to a problem where high accuracy is not expected. A state estimator which can be designed to use the nonlinearity as it is coming from the original model has always been invaluable in which high accuracy is expected. Thus to accomplish the above highlighted point, pseudolinear measurement model is directly applied to the Kalman filter equations without losing the original nonlinearity. We address the pseudolinear bias problem through simple geometry translation of system states. This system is simulated using Matlab for vehicle-landmarks system and results show that the new approach performs much accurately compared to that of well known Extended Kalman filter (EKF).