[Paper] SIDSense: Database-Free TV White Space Sensing for Disaster-Resilient Connectivity

Source: arXiv - 2602.13542v1

Overview

The paper introduces SIDSense, an edge‑AI system that lets devices use TV White Space (TVWS) spectrum without relying on the cloud‑based PAWS database. By moving spectrum sensing and decision‑making onto the device, the authors demonstrate a way to keep critical communications alive during climate‑driven disasters—especially for Small Island Developing States (SIDS) where terrestrial networks are often the first to fail.

Key Contributions

• Database‑free TVWS operation: A CNN‑based classifier runs locally to identify vacant TV channels, eliminating the PAWS database as a single point of failure.
• Compliance‑gated controller: Guarantees regulatory adherence by only authorizing channels that pass a strict audit log and policy check.
• Hybrid workflow: “Sensing‑first, authorize‑as‑soon‑as‑possible” pipeline that quickly falls back to a safe default when confidence is low.
• Integrated 5G video backhaul: Co‑locates TVWS sensing with a private 5G stack on a maritime vessel, enabling real‑time situational‑awareness video during emergencies.
• Empirical Caribbean TVWS dataset: Publicly released propagation and occupancy measurements (470‑698 MHz) for researchers and operators.
• Open‑source components: Selected parts of the SIDSense pipeline are contributed to the community to accelerate resilient deployments.

Methodology

1. Edge AI Model – A lightweight convolutional neural network (CNN) processes raw I/Q samples captured by a low‑cost SDR (Software‑Defined Radio). The model outputs a probability that a given channel is free.
2. Hybrid Decision Engine
   • Sensing Phase – The device continuously scans the TV band, feeding samples to the CNN.
   • Authorization Phase – If the confidence exceeds a configurable threshold, the compliance controller checks regulatory constraints (e.g., power limits, protected services) and logs the decision.
   • Graceful Degradation – When confidence is insufficient or the device detects a PAWS outage, it falls back to a pre‑approved “safe channel” list until a reliable decision can be made.
3. GPU‑aware Scheduling – The CNN inference runs on an embedded GPU, while the 5G Layer‑1 (L1) stack runs on the CPU. A scheduler prioritizes L1 deadlines to guarantee zero missed transmission slots.
4. Field Deployment – The prototype was mounted on a research vessel off Barbados. The team simulated PAWS outages by disabling the internet connection and measured sensing accuracy, latency, and 5G performance under realistic sea‑state conditions.

Results & Findings

The results show that SIDSense can reliably identify free TVWS channels in real time, keep a private 5G link alive, and do so without any reliance on external databases.

Practical Implications

• Disaster‑resilient communications – Emergency responders can deploy portable TVWS radios that stay operational even when internet or cellular backhaul is down.
• Cost‑effective rural broadband – Communities in SIDS can set up low‑power TVWS base stations without paying for continuous PAWS subscriptions or building expensive backhaul links.
• Maritime and offshore use‑cases – Ships, research vessels, and offshore platforms can maintain high‑quality video streams for situational awareness without needing satellite links.
• Regulatory compliance made easy – The built‑in compliance controller and audit logs simplify certification processes for operators who need to prove they respect incumbent services.
• Open data for spectrum planning – The released Caribbean TVWS dataset helps regulators and network planners model interference and design more efficient frequency reuse strategies.

Limitations & Future Work

• Model Generalization – The CNN was trained on Caribbean‑specific measurements; performance in other regions (e.g., Europe, Africa) may require retraining or transfer learning.
• Hardware Dependency – Accurate sensing hinges on a calibrated SDR and an embedded GPU; ultra‑low‑cost nodes without GPU acceleration may see higher latency or lower accuracy.
• Regulatory Acceptance – While the compliance controller enforces rules, many jurisdictions still mandate PAWS database checks; policy changes will be needed for full adoption.
• Scalability Testing – The current experiments involve a single vessel; future work should evaluate multi‑node coordination, interference management, and handover mechanisms in dense deployments.

The authors plan to extend SIDSense to support collaborative sensing among neighboring nodes, explore model compression for micro‑controller‑class hardware, and work with regional regulators to pilot database‑free TVWS operation in other climate‑vulnerable locales.

Authors

• George M. Gichuru
• Zoe Aiyanna M. Cayetano

Paper Information

• arXiv ID: 2602.13542v1
• Categories: cs.NI, cs.DC
• Published: February 14, 2026
• PDF: Download PDF