Adaptive localization techniques in WiFi environments

Indoor localization of a mobile user can be performed by using the off-the-shelf 802.11 (WiFi) infrastructure. However most of the existing position estimators are based on a stationary environment assumption that turns out to be rarely true in practice. We analyze two different approaches for the simultaneous estimation of the position and of the signal statistical model. The first uses a discrete state approach and is based on the Expectation-Maximization (EM) algorithm; the second employs a continuous state space and Kalman or Particle Filtering methodology. Numerical simulations and implementation show the effectiveness of the latter for real-time applications in nonstationary environments.