Tensor Data Conformity Evaluation for Interference-Resistant Localization

Abstract

We consider the problem of robust, interference-resistant localization in GPS-denied environments. Each asset to be self-localized is equipped with an antenna array and leverages time-domain coded beacon signals from anchor nodes that are placed at known locations. Collected data snapshots over time at the antenna array are organized in a tensor data structure. The conformity of the received tensor data is evaluated through iterative projections on robust, high-confidence data feature characterizations that are returned by L1-norm tensor subspaces. Non-conforming tensor slabs are more likely to be contaminated by irregular, highly deviating measurements due to interference, thus they are removed from the received dataset. Subsequently, we estimate the direction-of-arrival of the beacon signals by using L2-norm and L1-norm tensor decomposition techniques on the conformity-adjusted dataset. Finally, the relative position of the asset to the anchor nodes is estimated via triangulation. We consider two anchor nodes, one interferer, and one asset to be self-localized using radio frequency signals at the 2.4 GHz ISM band in an indoor laboratory environment. We evaluate the performance of the proposed localization system in terms of angle-of-arrival estimation accuracy experimental measurements from a software-defined radio testbed.