Two Staged ANN Based UWB Ranging Error Mitigation for Real Time Self Localization on Mobile Robots

Ultra-wideband (UWB) offers global positioning in various, mostly indoor scenarios, whereby the localization can be performed by the device itself. The usual accuracy of its proximity detection under good conditions is ±10 cm. However, this accuracy can be strongly influenced by environmental influences resulting in highly inaccurate and unsatisfactory position estimation. This paper introduces a method utilizing artificial neural networks (ANNs) to mitigate errors by detecting non-line-of-sight (NLOS) conditions and correcting measured distances. Two ANNs are created, which are capable to exploit information using both the additional metrics such as the channel impulse response (CIR) provided by the Decawave DW1000 UWB radio chip.