Particle filter for mobile station Positioning in a cellular network

Global Positioning System is the most commonly employed localization technique to localize mobile devices in outdoor environments. However, this cannot be operating when the line-of-sight visibility to the satellites is lost, as in indoor, dense environments or bad weather conditions. This motivates the growing network or hybrid based positioning techniques that use signal strength and network topology. Our methodology uses TEMS Investigation software to retrieve network information including signal strength and cell-identities of various network transmitters, and a nonlinear estimation like technique to estimate the mobile position. Typically, under linearity and Gaussian additive noise constraint, the conventional Kalman filters yields optimal estimation solution, provided the noise statistics is known. However, when such constraint is violated, e.g., either the measurement or state model is non-linear, the convergence of the filter cannot be granted. In this paper, we present a suboptimal estimation method using the particle filter where the cellular network data are combined to yield a close to optimal solution. The algorithm is tested on synthetic and real word dataset, where the results are compared with conventional Kalman filtering and unscented transform, where the superiority of the particle filtering like approach is demonstrated.