Simultaneous Localization and Mapping: A Pseudolinear Kalman Filter (PLKF) Approach

This paper describes an improved solution to the simultaneous localization and mapping (SLAM) problem based on pseudolinear models. 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. 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 model based Kalman filter (PLKF) state estimator is introduced. Evolution of vehicle motion is modeled using vehicle frame translation derived from successive dead reckoned poses as a control input. A pseudolinear process model is proposed to improve the accuracy and the faster convergence of state estimation. The general sensor model is presented in a pseudolinear form to preserve the nonlinearity in the observation model. The PLKF-based SLAM algorithm is simulated using Matlab for vehicle-landmarks system and results show that the proposed approach performs much accurately compared to the well known EKF-SLAM algorithm.