Bank of Kalman Filters in Closed-Loop for Robust Localization Using Unsynchronized Beacons

Indoor localization based on time difference of arrival has been recently a promising field of study. We consider the previously unsolved problem of locating a moving target receiver by using unsynchronized stationary beacons, showing an error in the range of centimetres in a real environment. The received line-of-sight signals have to be detected and their times of arrival have to be assigned to a beacon. In order to do this, the estimated positions and the velocities of the target are fused with the received symbols and timestamps in a bank of Kalman filters. The system is automatically calibrated by estimating the time offsets between the senders, their positions and the initial receiver position. The variables of the scenario are estimated in two steps, a calibration phase, where the variables of the scenario are estimated with local optimization algorithms and a tracking phase, where the continuous movement of the target is tracked using an unscented Kalman filter and a particle filter. Real life experiments show that the local optimization algorithms are capable of initializing the system, while the Kalman filters in closed-loop are able to track the target with an error of 0.065m at high velocity (up to 1.8 m/s).