Position and Velocity Tracking in Cellular Networks Using the Kalman Filter

Abstract

In this chapter, two estimation approaches are introduced to track the position and velocity of a MS in a cellular network. They are based on lognormal shadowing and Aulin's scattering models combined with the MLE and the EKF estimation algorithms, respectively. According to Aulin's channel model, the instantaneous electric field is a nonlinear function of the MS location and velocity. Consequently, the EKF is employed for the estimation process. Since the EKF approach is sensitive to the initial condition, we propose to use the ML estimate that employs the lognormal channel model, as the initial EKF state. Numerical results for typical simulations show that they are highly accurate and consistent. These methods also excel in using inherent features of the cellular system, i.e., they support existing network infrastructure and channel signalling. The assumptions are knowledge of the channel and access to the instantaneous received field, which are obtained through channel sounding samples from the receiver circuitry. Future work will focus on generating efficient channel estimation algorithms, to remove the assumption on partial knowledge of the channel. Work on building a pilot application to test the performance of the EKF in realistic conditions is on-going together with the incorporation of channel model parameters estimation algorithms. Another direction in future work is to use more advanced filtering techniques such as the unscented Kalman filter (Julier & Uhlmann, 1997) and the particle filter (Arulampalam et al., 2002), which are not based upon the principal of linearising the nonlinear state and measurement models using Taylor series expansions as the EKF. Some preliminary results for MS location and velocity estimation algorithm based on particle filtering are presented in (Olama et al., 2007; Olama et al., 2008).