Radio Spectrum Sensing Framework for Large-Scale, High-Density IoT Sensor Networks

The rapid growth of the Internet of Things (IoT) requires innovative solutions to address the challenges posed by constrained radio spectrum resources in Wireless Sensor Networks (WSNs). This paper presents an Artificial Intelligence (AI) based spectrum sensing framework designed to detect and classify various IoT communication technologies and generate advanced communication channel analytics, such as the technology in use, central frequency and the bandwidth of the radio channel, time-on-air parameter, channel usage patterns, and radio spectrum occupancy degree. In contrast to most existing other methods, the proposed framework follows an end-to-end approach that integrates dataset design, AI model training, and inference evaluation using IoT sensors controlled by a traffic orchestrator that emulates a large-scale, high-density WSN architecture. The system follows a complete design approach, automatically creating and enhancing datasets consisting of spectrogram images from both controlled tests and real-world scenarios (6,190 unique radio spectrograms). Using the YOLOv8 architecture, the AI model demonstrates high performance in detecting multiple narrow-band radio signals, including LoRa and Sigfox technologies, achieving precision and recall scores of 0.983 and 0.947, respectively. In live deployment tests, the framework processes spectrograms with an average latency of 30.12 ms. Through an integrated detection tracking mechanism, it efficiently generates advanced radio channel analytics, enabling better use of spectral resources and interference mitigation in large-scale IoT networks, making it a valuable tool for optimizing IoT deployments. In conclusion, the proposed framework offers significant advances in spectrum sensing methodologies, paving the way for improved scalability and reliability in the evolving IoT landscape.