A SVM approach to UWB-IR based positioning under NLOS conditions

This paper presents a machine learning approach, namely the Support Vector Machine (SVM), to solve a particular localization problem. The problem is to ascertain whether an object carrying a localization tag is inside or outside a particular area. As the area becomes smaller and as the object approaches the boundaries of the area, even minute errors can result in a completely wrong estimation. SVM was chosen for this problem due to its generalization capability in handling noisy data. Training and test data for the SVM were obtained from an experimental setup of the test scenario. The results obtained proved that SVM was a suitable tool for this application, due to its ability in handling the noisy data caused by the NLOS condition.