One time‐step particle smoothing for radio range‐based indoor position tracking

In the context of sequential estimation of radio range-based indoor position tracking, Bayesian smoothing framework is promising as involving past, present and future observations. The performance and practicability of a smoothing method greatly depend on how many and how future observations are incorporated. Aiming at real-time locating systems, the authors propose one time-step smoothing form on sequential Monte Carlo methods, including four popular Bayesian smoothers and a novel one time-step smoothed filtering (SF) algorithm. The smoothing algorithms are evaluated through two-dimensional position tracking on a real-world indoor test-bed. The authors present results that the proposed SF improves tracking performance requiring very limited computation and memory, which is applicable for real-time indoor position tracking. Moreover, the one time-step smoothing form is validated to mitigate ranging errors and smooth positioning trajectories.