Improving UWB Positioning Accuracy by Fusion With Pedestrian Dead-Reckoning

Ultra-wide Band (UWB) positioning systems are a system of choice for indoor positioning. Though UWB is more resilient against multipath propagation, the positioning accuracy can still be hindered in its presence. The positioning accuracy and availability can be improved by combination with other sensors, typically inertial ones. In the case of pedestrian positioning the use of a dead-reckoning system, which uses the inertial data from a smartphone, may be advantageous. In this paper we propose the use of Extended Kalman filtering for the data fusion of UWB and STEP system, which detects the user making steps and estimate the step-length from the vertical acceleration of the user. For the fusion we assume two models: fusion of UWB and step-length with heading and fusion of UWB and step-length. The performance is evaluated on real data and it is shown that the UWB positioning error of 0.59 m is improved to 0.22 m and 0.36 m when fusion with both the step-length and heading or with only the step-length is used, respectively.