Relevance vector machine for UWB localization

Wireless localization systems have a great importance in a variety of fields such as positioning and tracking systems. Specifically, in hash conditions, e.g., indoor environments, it is difficult to localize an agent with high accuracy due to radio blockage or insufficient information of anchors. Therefore, identification and mitigation of non-line-of-sight (NLOS) radio propagation are highlighted as a solution for improving localization accuracy by overcoming these limitations. In this paper, we develop a robust and efficient localization algorithm using relevance vector machine (RVM). We first design a RVM classifier to identify NLOS signals using features extracted from the received waveform. We then design a RVM regressor to predict a probability density function of range estimates (pair-nodes distance) by exploiting its prediction ability. Numerical results show that the RVM localization algorithm provides high localization accuracy with low complexity.