Autocalibration of a wireless positioning network with a FastSLAM algorithm

The calibration (measurement of the position) of a network of wireless nodes used for indoor localization purposes is a tedious process and prone to errors if done manually. This paper presents a method for the autocalibration of that network, using the measurement of the received signal strength (RSS) of RF signals coming from the nodes, and captured while a person is taking a random walk in the environment. The calibration method is adapted from a Simultaneous Localization and Mapping (SLAM) technique from Robotics, and is based on a Bayesian particle filter modeling the unknown position of the user and the location of the beacons. Information coming from RSS measurements is incorporated to the filter using a rather generic measurement model (the path loss law), producing a sequence of beacon nodes position estimates with decreasing uncertainty over time. The accuracy and convergence of the method can be further enhanced by using pedestrian dead-reckoning (PDR) techniques from the handheld smartphone used to capture the RF data. The method is demonstrated with a deployment of 60 unknown position active RFID tags (and 4 known position tags) in an indoor environment, and a trajectory lasting 1054 s. The results are a median beacon positioning error of 4.9 m using only the RSS information, and 3.4 m if PDR information is incorporated to the particle filter. This error can be further decreased by adding the results of more calibration routes.