Batteryless Frontiers in Edge AI: Harvesting Intelligence via Energy-Efficient Sensor Selection and Policy Optimization

University Of Texas At Austin, Austin TX

Investigators

Abstract

Developing batteryless technology for Internet of Things (IoT) devices offers a promising future, especially for wearable technologies like fitness trackers, smartwatches, and medical devices. The project addresses fundamental issues such as the reliability and functionality of these devices without traditional power sources. By integrating artificial intelligence (AI) and deep learning techniques, this project aims to harness the potential of batteryless sensors for personalized data analytics. This integration requires substantial intellectual innovation, particularly in aligning machine learning (ML) with systems design. Consequently, this project not only aims to transform edge computing, but also has far-reaching implications for healthcare, IoT, and personal augmented reality/virtual reality (AR/VR) applications. Beyond technological advancements, the project has a significant educational impact as it proposes an interdisciplinary, research-based curriculum that combines machine learning (ML) and batteryless system design. This educational approach will foster a new generation of innovators equipped with the skills to advance both ML and batteryless technologies. In terms of new methods, this project seeks to develop new deep learning algorithms tailored for unstructured data generated by batteryless sensors, which differ fundamentally from traditional ML settings. A joint optimization approach is proposed to balance the energy usage strategies of batteryless sensors with the requirements of ML models for data analytics. The project will also develop a novel approach for selecting sensors specifically for batteryless systems. Real prototypes using kinetic energy harvesting will be built to validate simulations through hardware profiling. These prototypes will facilitate the collection of a real-world dataset, providing a comprehensive methodology for integrating batteryless sensors with ML. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.