Deep learning-driven sensing for sustainable Internet of Things

The rapid proliferation of the Internet of Things has transformed various domains of everyday life, from familiar household items like smart light bulbs to healthcare resources, including medical devices, wearable technologies, smart devices, and ultimately, smart cities. However, the increasing use of these devices has raised concerns about their energy consumption, which poses a significant challenge to their sustainability. This study presents a novel, deep-learning-based sensing framework for energy-efficient data collection. Using predictive models to forecast sensor readings, the framework adapts to environmental changes and optimises the time spent sensing and transmitting data. This leads to significant energy savings while preserving data quality. Extensive simulations, combined with experiments conducted on a case study based on three popular IoT hardware platforms, demonstrate the approach’s effectiveness, achieving up to 89% energy savings in the active states, compared to traditional monitoring methods, while maintaining high data accuracy and system performance. This research contributes to the sustainable growth of IoT ecosystems, promoting more energy-efficient and environmentally responsible systems for the future.