Robust Indoor Localization in Dynamic Environments: A Multi-source Unsupervised Domain Adaptation Framework
Jiyu Jiao, Xiaojun Wang, Chengpei Han
TL;DR
This work addresses robust indoor localization in dynamic environments where signal distributions drift over time. It introduces DF-Loc, an end-to-end framework that combines quality-controlled CSI preprocessing, CSI fingerprint reconstruction, a multi-scale attention-based feature extractor, and a dual-stage multi-source unsupervised domain adaptation (MUDA) with domain-specific adapters to align source-target distributions and regressor outputs. The key contributions include marginal and conditional distribution adaptation via Maximum Mean Discrepancy (MMD), a two-stage alignment for cross-domain regression, and a Multi-Scale AFF network with MS-Conv and MS-CAM to extract transferable features. Experiments in office and classroom settings show that DF-Loc achieves superior localization accuracy and robustness, achieving average errors around $0.79$–$0.94$ m under varying training regimes and test conditions, outperforming several baselines and demonstrating strong generalization capabilities.
Abstract
Fingerprint localization has gained significant attention due to its cost-effective deployment, low complexity, and high efficacy. However, traditional methods, while effective for static data, often struggle in dynamic environments where data distributions and feature spaces evolve-a common occurrence in real-world scenarios. To address the challenges of robustness and adaptability in fingerprint localization for dynamic indoor environments, this paper proposes DF-Loc, an end-to-end dynamic fingerprint localization system based on multi-source unsupervised domain adaptation (MUDA). DF-Loc leverages historical data from multiple time scales to facilitate knowledge transfer in specific feature spaces, thereby enhancing generalization capabilities in the target domain and reducing reliance on labeled data. Specifically, the system incorporates a Quality Control (QC) module for CSI data preprocessing and employs image processing techniques for CSI fingerprint feature reconstruction. Additionally, a multi-scale attention-based feature fusion backbone network is designed to extract multi-level transferable fingerprint features. Finally, a dual-stage alignment model aligns the distributions of multiple source-target domain pairs, improving regression characteristics in the target domain. Extensive experiments conducted in office and classroom environments demonstrate that DF-Loc outperforms comparative methods in terms of both localization accuracy and robustness. With 60% of reference points used for training, DF-Loc achieves average localization errors of 0.79m and 3.72m in "same-test" scenarios, and 0.94m and 4.39m in "different-test" scenarios, respectively. This work pioneers an end-to-end multi-source transfer learning approach for fingerprint localization, providing valuable insights for future research in dynamic environments.