Improving Deep Learning-Based UWB LOS/NLOS Identification with Transfer Learning: An Empirical Approach

Abstract

This paper presents an improved ultra-wideband (UWB) line of sight (LOS)/non-line of sight (NLOS) identification scheme based on a hybrid method of deep learning and transfer learning. Previous studies have limitations, in that the classification accuracy significantly decreases in an unknown place. To solve this problem, we propose a transfer learning-based NLOS identification method for classifying the NLOS conditions of the UWB signal in an unmeasured environment. Both the multilayer perceptron and convolutional neural network (CNN) are introduced as classifiers for NLOS conditions. We evaluate the proposed scheme by conducting experiments in both measured and unmeasured environments. Channel data were measured using a Decawave EVK1000 in two similar indoor office environments. In the unmeasured environment, the existing CNN method showed an accuracy of approximately 44%, but when the proposed scheme was applied to the CNN, it showed an accuracy of up to 98%. The training time of the proposed scheme was measured to be approximately 48 times faster than that of the existing CNN. When comparing the proposed scheme with learning a new CNN in an unmeasured environment, the proposed scheme demonstrated an approximately 10% higher accuracy and approximately five times faster training time.

Node Calibration in UWB-Based RTLSs Using Multiple Simultaneous Ranging

“…A crucial consideration for anchor placement is to avoid NLOS conditions between mobile and nearby anchors, since they would induce positive biases on the estimated TOFs that may vary dynamically and significantly as the mobile moves around. Several NLOS mitigation techniques are present in the literature [ 37 , 38 , 39 ]. However, a TOF estimated from NLOS conditions cannot have LOS-equivalent results.…”

Antenna Delay Calibration of UWB Nodes

“…This source of error is contrasted to antenna delay, which produces a rather quasi-static bias. Currently, although there are available NLOS-mitigation techniques [21]- [23] that can be useful to some extent, a TOF estimated from NLOS conditions cannot be remedied to have LOS-equivalent quality. Hence, we assume a rule of anchor placement such that a tag in the coverage area should typically find LOSs to at least three nearby anchors.…”

“…This is to ensure that a TOF from NLOS conditions is only infrequently used in the position estimation. Namely, when a tag sees LOSs to a sufficient number of nearby anchors, the position-estimation process with a LOS/NLOS identification technique [21]- [23] can simply ignore the other nearby NLOS anchors to avoid significant performance impairment. Then, based on our current office building, it seems to be difficult to have LOS over a range greater than 12 m. Therefore, based on the assumed rule of anchor placement, we assume that TOFs of range less than 12 m are typically used by the RTLS in our considered application.…”

“…Antenna delay calibration relies on a pre-measurement of the antenna delay, which can then be used in operating applications to correct the TWR bias caused by the antenna delay itself. Such bias correction is considered to be useful only when the LOS condition is the case because the NLOS condition adds a variable bias that needs specific techniques [21]- [23] to deal with. In addition, we do not have a ground truth of the antenna delay of any UWB node.…”

NLOS identification and mitigation in UWB positioning with bagging-based ensembled classifiers

“…As UWB signals possess large bandwidths, the identification/classification can be based on the signal properties [13]. Some of these approaches are hypothesis testing [4,14], filter-based [16], feature-based (UWB signal) classifiers (machine learning (ML) approaches) [17][18][19][20] and deep learning-based [21,22]. Cooperative localization in NLOS environments is investigated in [23,24].…”