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 differ 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 pre-trained 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.

System Flow

Paper Link

The widespread use of Wi-Fi infrastructure in indoor environments allows for non-intrusive health monitoring. While traditional vital sign monitoring relies on specialized equipment that can be expensive, uncomfortable, or not viable for long-term use, recent advances in wireless sensing allow for continuous, contactless monitoring using commodity devices. Our work looks at how Channel State Information (CSI) from low-cost, single antenna Wi-Fi devices can be used for accurate heart rate monitoring using compact deep learning.

Our experimental validation is done using two datasets, including the most CSI-based heart rate dataset with 118 participants in 17 positions, which shows that PulseFi achieves state-of-the-art accuracy (Mean absolute error 0.08 BPM) compared to existing methods, while maintaining nearly perfect consistency across varying distances between transmitter/receiver, and positions. By using single antenna amplitude information alone, PulseFi significantly reduces the hardware requirements and computational complexity, with the models being able to run locally real-time on the ESP32 chips (<600KB of RAM).

The academic year has started and more advancements are taking place in the lab! This quarter, we’re happy to talk about new research in network cybersecurity, indoor localization, edge intelligence, and new light-based technologies!

• October 25 • Nayan Bhatia • Wi-Fi Indoor Localization
• November 1 • Harikrishna Kuttivelil • Emergent Decentralized Federated Learning
• November 8 • Li Xue • Device Fingerprinting for Network Security
• November 15 • Lakshmi Krishnaswamy • Load Balancing for Geo-Distributed Data Center Networking
• November 22 • Tyler Morton & Firouz Valdafari • Li-Fi: Light-based Technologies