Successive Threshold-Based Multipath Mitigation Aided by Neural Network for UWB Ranging

Despite the recent advances in UWB systems especially after releasing the IEEE 802.1S.4z standard, achieving the “cm” level ranging accuracy and precision using UWB in a general multipath environment is very challenging. In this paper, we introduce a successive threshold-based multipath mitigation algorithm (STM) that improves the ranging performance in multipath conditions. We further improve the ranging performance of STM in multipath conditions by aiding the STM with STMnet as a neural network that estimates the ranging error of the STM. Using a UWB packet simulator, we show that the (mean, Q95) of ranging error for the STM aided by STMnet achieves up to (-0.07cm, 0.63cm), (-0.49cm, 21.3cm), and (0.06cm, 39cm) for LOS environment, LOS residential and LOS office multipath environments, respectively. The improved performance using STMnet is at the cost of 330KB of required memory and 232K of floating point operations.