A Comparison between Nonlinear Estimation based Algorithms for Mobile Robot Localizations

Nonlinear estimation of uncertainties are generally solved by using Extended Kalman Filter (EKF), however performance of EKF algorithm is sluggish when system is highly nonlinear on scale for updating time interval. Tracking the ground truth, mobile robot model is highly nonlinear. In these conditions Unscented Kalman Filter (UKF) algorithm estimation performance is compared with EKF algorithm, while performing complexity of the same order. Difference between the algorithms are EKF performance based on Taylor approximations using jacobian matrices while UKF performance is based on the tuning of Unscented Transform (UT) parameters. In this paper, we are comparing mobile robots for both EKF and UKF navigation based algorithms for localization with respect to ground truth, Global Positioning System (GPS) and dead reckoning.