Transforming Decoder-Only Transformers for Accurate WiFi-Telemetry Based Indoor Localization
Nayan Sanjay Bhatia, Katia Obraczka
TL;DR
WiFiGPT reframes WiFi telemetry-based indoor localization as a deterministic regression task using a decoder-only transformer. By fine-tuning LLaMA models with LoRA adapters on structured telemetry prompts, it integrates diverse sources (CSI, FTM, RSSI) without hand-crafted features, achieving sub-meter to centimeter-level accuracy across LOS/NLOS environments. The approach demonstrates robust cross-device generalization, data-imputation capabilities, and deterministic inference, suggesting practical deployment on commodity hardware with minimal calibration. This token-based, schema-less framework reduces engineering overhead while capturing complex wireless propagation patterns like multipath, enabling scalable, vendor-agnostic indoor localization. The work highlights the potential of large language models to extend beyond language tasks into reliable regression for wireless sensing.
Abstract
Wireless Fidelity (WiFi) based indoor positioning is a widely researched area for determining the position of devices within a wireless network. Accurate indoor location has numerous applications, such as asset tracking and indoor navigation. Despite advances in WiFi localization techniques -- in particular approaches that leverage WiFi telemetry -- their adoption in practice remains limited due to several factors including environmental changes that cause signal fading, multipath effects, interference, which, in turn, impact positioning accuracy. In addition, telemetry data differs depending on the WiFi device vendor, offering distinct features and formats; use case requirements can also vary widely. Currently, there is no unified model to handle all these variations effectively. In this paper, we present WiFiGPT, a Generative Pretrained Transformer (GPT) based system that is able to handle these variations while achieving high localization accuracy. Our experiments with WiFiGPT demonstrate that GPTs, in particular Large Language Models (LLMs), can effectively capture subtle spatial patterns in noisy wireless telemetry, making them reliable regressors. Compared to existing state-of-the-art methods, our method matches and often surpasses conventional approaches for multiple types of telemetry. Achieving sub-meter accuracy for RSSI and FTM and centimeter-level precision for CSI demonstrates the potential of LLM-based localisation to outperform specialized techniques, all without handcrafted signal processing or calibration.