TinyML Applications, Research Challenges, and Future Research Directions
Hicham Oufettoul, Redouane Chaibi, Saad Motahhir
Abstract
In recent times, both the academic and industrial sectors have developed a greater interest in artificial intelligence (AI) and machine learning (ML). Conventional ML approaches usually need significant computational resources to attain the intended level of precision, thereby restricting their feasibility to high-capacity hardware including network nodes. A paradigm shift has occurred with the emergence of Tiny Machine Learning (TinyML), enabling ML tasks to be run on Internet of Things (IoT) hardware with severe limitations. The TinyML paradigm advocates the integration of ML-based processes into small devices powered by microcontroller units (MCUs). This article begins with an introduction to TinyML and then explains the tools that support it. We provide up-to-date information about TinyML frameworks. Subsequently, the most recent TinyML apps that use cutting-edge technology are discussed. Finally, several research problems and potential future directions are noted.