Enhanced Unscented Kalman Filtering for Robust Mobile Tracking in NLOS Environments

Abstract

This paper proposes a newly enhanced robust unscented Kalman filter for mobile positioning in environments with various probabilistic non-line-of-sight (NLOS) events. To develop our filter for significantly reducing the adverse effect of NLOS outliers and enhancing tracking performance, we first propose trimmed measurements obtained from raw observations. Next, we propose nonlinear regression model-based M-estimation for adaptively adjusting the measurement noise covariance matrix. We then incorporate the trimmed measurements and M-estimation scheme into the conventional unscented Kalman filter to yield the proposed enhanced UKF. The effectiveness of the proposed filter is demonstrated using an example.