Improved Kalman filtering algorithms for mobile tracking in NLOS scenarios

Abstract

This paper presents an improved positioning approach for cellular-network based mobile tracking in severe non-line-of-sight (NLOS) propagation environments. The proposed approach consists of two stages: the smoothing stage to suppress the NLOS errors in the distance measurements; and the position tracking stage. An improved distance smoothing method is proposed to significantly reduce the NLOS errors. It applies online distance mean and variance estimates to identify LOS and NLOS propagations. The online LOS and NLOS identification results, the distance mean and variance estimates are employed to update the Kalman filter (KF) for smoothing distance measurements. A data fusion technique is developed to combine distance measurements, mobile velocity and heading angle estimates provided by motion sensors through the extended KF. Simulation results demonstrate that the proposed two-stage approach significantly improves position accuracy compared to the existing NLOS mitigation algorithms, at the cost of increased computational complexity.