Urban Navigation with LTE using a Large Antenna Array and Machine Learning

Channel fingerprinting entails associating a point in space with measured properties of a received wireless signal. If the propagation environment for that point in space remains reasonably static with time, then a receiver with no knowledge of its own position experiencing a similar channel in the future might reasonably infer proximity to the original surveyed point. In this article, measurements of downlink LTE Common Reference Symbols from one sector of an eNodeB are used to generate channel fingerprints for a passenger vehicle driving through a dense urban environment without line-of-sight to the transmitter. Channel estimates in the global azimuthal-delay domain are used to create a navigation solution with meter-level accuracy around a city block.