Decision tree-extended finite impulse response filtering for pedestrian tracking over tightly integrated inertial navigation system/ultra wide band data

Although the tightly integrated inertial navigation system/ultra wide band (INS/UWB) improves the localization accuracy, it suffers from UWB distance outage. In order to reduce the outage effect on the position accuracy, in this paper we propose using a novel decision tree (DT)-extended finite impulse response (EFIR) filter. When all the UWB distances are available, the EFIR filter tightly fuses the INS and UWB data. Otherwise, the DT builds a relationship between the INS and UWB position errors. Once at least one UWB distance is unavailable, the DT bridges a gap over unavailable measurements. It is shown experimentally that the DT-EFIR filter is an efficient tool to reduce the effect of the UWB distance outage on the INS/UWB system operation, which can improve the localization error by about 40% as compared with the UWB solution. (paper)