Artificial Intelligence-Based Non-intrusive Driver Drowsiness Detection: A Multimodal Review of Behavioral and Physiological Solutions

Road accidents and fatalities, in part due to human fatigue and sleep disorders, are caused by several issues and result from driver drowsiness. The drowsiness is so important that many scientists around the world have begun studying it non-invasively using machine learning (ML) and deep learning (DL) algorithms to uncover behavioral and physiological signals that indicate a driver may be drowsy. In recent years, many researchers have also developed hybrid and multimodal ways to integrate different cues to detect drowsy drivers more accurately. This document is a systematic review of studies on Artificial Intelligence (AI)- based drowsiness detection that used non-intrusive ML and DL methods published between 2020 and 2025. In addition, it provides an overview of the types of sensors used for drowsiness detection, the ML algorithms used, how the accuracy of each method was measured, and the typical limitations associated with these studies, including the issue of data imbalance and lack of diversity in data sets; real-time implementation issues; power requirements; and ethical issues regarding privacy and fairness. This document also provides an overview of current research trends and identifies areas that remain open to investigation and further study. Multimodal integration of drowsiness detection with use, along with the integration of explainable AI techniques, can lead to the development of safer, more transparent, and more reliable intelligent transportation systems.