Localization in Dynamic Indoor MIMO-OFDM Wireless Systems using Domain Adaptation

TL;DR

This paper introduces a domain adaptation approach using neural networks to improve user equipment localization accuracy in dynamic indoor wireless environments where environmental changes affect fingerprint distributions.

Contribution

It presents a novel domain adaptation framework that aligns wireless fingerprint distributions before and after environmental changes for improved localization.

Findings

Effective localization in dynamic environments demonstrated

Neural network-based distribution alignment improves accuracy

Robustness to environmental changes confirmed through experiments

Abstract

We propose a method for predicting the location of user equipment (UE) using wireless fingerprints in dynamic indoor non-line-of-sight (NLoS) environments. In particular, our method copes with the challenges posed by the drift, birth, and death of scattering clusters resulting from dynamic changes in the wireless environment. Prominent examples of such dynamic wireless environments include factory floors or offices, where the geometry of the environment undergoes changes over time. These changes affect the distribution of wireless fingerprints, demonstrating some similarity between the distributions before and after the change. Consequently, the performance of a location estimator initially designed for a specific environment may degrade significantly when applied after changes have occurred in that environment. To address this limitation, we propose a domain adaptation framework that utilizes neural networks to align the distributions of wireless fingerprints collected both before and after environmental changes. By aligning these distributions, we design an estimator capable of predicting UE locations from their wireless fingerprints in the new environment. Experiments validate the effectiveness of the proposed methods in localizing UEs in dynamic wireless environments.