Data fusion for ground target tracking in GSM networks

摘要

Positioning in mobile cellular networks is an exciting research area. The Global System for Mobile communications (GSM) network, as a widely used mobile communication standard around the world, has shown the potential to provide position information. Ground target tracking is a significant application of finding the position of a mobile station (MS). However, a GSM positioning system based on current specifications faces many difficulties to yield an accurate position estimate. Since the signals are designed by communication needs rather than positioning, the resolution of the measurements in GSM networks for positioning is coarse. The ambiguities of the position estimate arise when there are not a sufficient number of measurements available. Moreover, due to the restriction of terrain, road and traffic, the ground target often maneuvers. Therefore, data fusion approaches, which integrate redundant information from different sources, are applied in this work to obtain improved position estimation accuracy. This work focuses on the state estimation problem of the MSu27s position given the measurements from the GSM networks and a priori road information. A data fusion solution, which integrates time of arrival (TOA) and received signal strength (RSS) measurements using an extended Kalman filter (EKF), is proposed to provide an improved position estimate. The theoretical best achievable performance, posterior Cramer-Rao lower bound (PCRLB), is derived for the data fusion approach. The PCRLB is used to demonstrate the benefits of the fusion approach and applied as a benchmark to compare different estimators. The road constraint is incorporated into the estimation process as a pseudomeasurement. Simulations of the linear and nonlinear road segments prove the advantages of the road-constrained approach. Moreover, the motion mode uncertainty problem is considered and solved by a multiple model (MM) approach. In particular, an adaptive road-constrained interacting MM (ARC-IMM) estimator, which incorporates the road information into a variable structure MM mechanism, is proposed and demonstrated to be effective and robust to provide a significantly improved position estimate.