Observability-Constraint Extended Kalman Filter for SLAM Navigation Systems; Design and Comparison

Recently, the use of different visual navigation algorithms as a navigation-aided system has been become of interest by many researchers.In most of these approaches, to improve the accuracy of the navigation problem, undesired predicted landmarks are used again in comparison with the output of the algorithm by appropriate filtering units for fusion and integration in the system.However, the number of points with a specific feature, the amount of noise, and the presence or absence of color noise have a direct impact on the performance of the conventional filters such as extended Kalman filter (EKF).In this paper, in addition to demonstrating how to model SLAM-based video navigation systems and systematically design and implement filtering unit, an observability-constrained EKF (OC-EKF) has been proposed for improving the performance of the system and minimizing the system errors.The main contribution of the proposed scheme is that it has provided the SLAM filtering mechanism as a constrained optimization issue that considers the observability constraints in the estimation problem.Besides, the performance of OC-EKF is compared with the standard EKF (STD-EKF) and the first estimate Jacobean EKF (FEJ-EKF), H inf -based Kalman Filter (KF), and Robust KF; where, the robust performance of the proposed scheme against uncertainties, white and color noises have been verified through comprehensive simulation results.As shown in the Monte Carlo simulation results, the proposed OC-EKF not only provides more accurate navigation, but also behaves more robustly against system uncertainties and noises.