Lessons from Deploying Learning-based CSI Localization on a Large-Scale ISAC Platform

TL;DR

This work addresses the challenge of scalable CSI-based indoor localization in a large-scale ISAC platform by proposing GLow, a graph neural network framework that explicitly handles heterogeneous CSI and leverages unlabeled data. The method combines spatiotemporal pretraining with a graph-encoded representation of CSI and a confidence-aware fine-tuning stage to produce accurate 3D coordinates and associated confidence, even across dozens of AP configurations and multiple device types. Evaluated on over 70,000 data points across five floors and 25,600 m^2, GLow achieves a median localization error of 2.17 m in a leave-one-phone-out setting and a floor accuracy of 99.49%, outperforming state-of-the-art baselines with an 18.7% MAE improvement. The results demonstrate the practicality of large-scale, server-side CSI localization for ISAC deployments and highlight the value of unlabeled data and heterogeneous CSI encoding for real-world performance.

Abstract

In recent years, Channel State Information (CSI), recognized for its fine-grained spatial characteristics, has attracted increasing attention in WiFi-based indoor localization. However, despite its potential, CSI-based approaches have yet to achieve the same level of deployment scale and commercialization as those based on Received Signal Strength Indicator (RSSI). A key limitation lies in the fact that most existing CSI-based systems are developed and evaluated in controlled, small-scale environments, limiting their generalizability. To bridge this gap, we explore the deployment of a large-scale CSI-based localization system involving over 400 Access Points (APs) in a real-world building under the Integrated Sensing and Communication (ISAC) paradigm. We highlight two critical yet often overlooked factors: the underutilization of unlabeled data and the inherent heterogeneity of CSI measurements. To address these challenges, we propose a novel CSI-based learning framework for WiFi localization, tailored for large-scale ISAC deployments on the server side. Specifically, we employ a novel graph-based structure to model heterogeneous CSI data and reduce redundancy. We further design a pretext pretraining task that incorporates spatial and temporal priors to effectively leverage large-scale unlabeled CSI data. Complementarily, we introduce a confidence-aware fine-tuning strategy to enhance the robustness of localization results. In a leave-one-smartphone-out experiment spanning five floors and 25, 600 m2, we achieve a median localization error of 2.17 meters and a floor accuracy of 99.49%. This performance corresponds to an 18.7% reduction in mean absolute error (MAE) compared to the best-performing baseline.

Figures (18)

• Figure 1: General structure of the wireless transmission process. The diagram illustrates the transmission process where spatial streams undergo CSD and spatial mapping at the transmitter. These streams then propagate through the physical channel environment and are received to calculate the CSI at the receiver end.
• Figure 2: The CSI dimensions of multiple users showcasing variations.
• Figure 3: Envelope plots of CSI amplitude captured at the same transmitter-receiver pair and location.
• Figure 4: The processing pipeline of localization service of GLow.
• Figure 5: Definition of Edge Connections.