Passive Unsupervised Localization and Tracking using a Multi-Static UWB Radar Network

The indoor localization and tracking of objects and humans with high accuracy is becoming increasingly important in a number of applications including healthcare, ambient assistant-living, surveillance, among others. Since Ultrawideband (UWB) systems have a large operating bandwidth, they can provide centimeter (cm) level localization accuracy in Line-of-Sight (LoS) scenarios. However, current commercial UWB systems require the target to carry an active device (tag) so that it can be precisely located and this may be impractical in applications such as security and surveillance. In this work, we process experimental data obtained from a multi-static UWB radar network for the passive indoor localization of a moving target. Current fingerprinting-based passive localization techniques require a substantial radio-map survey in the offline training phase and labor-intensive fingerprint updates when there are changes in the environment. In this work, on the other hand, we propose to use the fine-grained physical layer information, known as Channel Impulse response (CIR), that is exchanged between UWB modules and capitalize on the fact that a moving person induces variations in the CIR that stand out against the background signal. Our results show that a walking target can be passively located with a median distance error as low as 0.55m in an indoor environment.