Cubature Kalman Filter Based Localization and Mapping

Simultaneous Localization and Mapping (SLAM) is the process of simultaneously building a map and localizing in it, and can be used for autonomous navigation. SLAM deals with estimation of vehicle states and landmarks. Most SLAM algorithms are based on extended Kalman filters (EKFs). However, the use of EKF for SLAM is not the best choice, as it works only for `mild' nonlinear environments owing to the assumption of first order Taylor series approximations of process and observation models. A few researchers has also proposed the use of other estimation techniques like particle filters, unscented Kalman filters (UKFs), etc for SLAM. In this paper, we propose the use of a cubature Kalman filter (CKF) for the estimation of the SLAM augmented state vector. The proposed SLAM is derivative less SLAM. A comparison of CKF SLAM and UKF SLAM is given through numerical simulations.