MmWave user positioning accuracy challenged by blockage and calibration

mmWave UE Positioning Under Blockage: Limits for ISAC-Assisted Handover

Networking and Internet Architecture

Summary

Millimeter-wave (mmWave) networks can lose connection when something blocks the signal, like a person walking by. This paper looks at whether the network can figure out a user's exact location from radio signals they send, which helps keep the connection alive by switching to another antenna early. The authors show that errors in antenna setup limit accuracy more than signal bandwidth when there's a clear line of sight. When the line of sight is blocked, using known reflections from the environment helps a lot, but simple methods to estimate position fail. Their improved method reduces the location error significantly but still doesn't outperform current industry standards.

What this means in practice

  • For mobile network operators: Improve mmWave network handover by accurately tracking user position using multipath reflections to reduce connection drops during signal blockage.
  • For wireless infrastructure engineers: Design more precise antenna calibration and multipath-aware estimators to enhance positioning performance in urban mmWave deployments with blockages.

Authors

Seyed Bagher Hashemi Natanzi, Bo Tang

Abstract

Proactive handover can save a millimeter-wave (mmWave) link before a passing body blocks it, but only if the network knows where the user is. We ask whether the radio units that serve a user can localize it accurately enough from its uplink sounding reference signals. We derive position error bounds that include synchronization and array calibration errors, with and without reflected paths known from a digital twin, and test two estimators in a ray-traced street canyon on held-out scenarios. Three findings stand out. First, with line of sight to both radio units, array calibration, not bandwidth, limits the bound. Second, blockage degrades the line-of-sight-only bound sharply, whereas known reflections keep the map-aided bound low. Third, a simple dominant-path estimator fails where it matters most: its errors grow sharply around blockage, and proactive handover becomes worse than the 3GPP baseline. A map-aided multipath tracker cuts the median error from 12 to 3.7 cm and removes this significant penalty at link margins of 15 dB and above, but does not yet beat the baseline.
View PDFOpen arXiv