Implicit channel charting with application to UAV-aided localization

Traditional localization algorithms based on features such as time difference of arrival are impaired by non-line of sight propagation, which negatively affects the consistency that they expect among distance estimates. Instead, fingerprinting localization is robust to these propagation conditions but requires the costly collection of large data sets. To alleviate these limitations, the present paper capitalizes on the recently-proposed notion of channel charting to learn the geometry of the space that contains the channel state information (CSI) measurements collected by the nodes to be localized. The proposed algorithm utilizes a deep neural network that learns distances between pairs of nodes using their measured CSI. Unlike standard channel charting approaches, this algorithm directly works with the physical geometry and therefore only implicitly learns the geometry of the radio domain. Simulation results demonstrate that the proposed algorithm outperforms its competitors and allows accurate localization in emergency scenarios using an unmanned aerial vehicle.