Road accidents are a leading cause of injury and death worldwide, with human factors such as fatigue and stress playing a critical role in impairing driver performance. Understanding and mitigating these risks is essential for enhancing road safety and saving lives. Recent developed sensors, technologies and machine learning have enabled the creation of systems that monitor driver states in real time, focusing on physiological and behavioral indicators of impairment. However, existing systems are often limited in their ability to integrate multimodal data effectively, resulting in reduced accuracy and delayed interventions. Here, we show that combining physiological sensors (e.g., heart rate, skin conductance) with behavioral monitoring (e.g., eye tracking, steering behavior) significantly improves the detection of dizziness and stress, achieving an accuracy improvement compared to traditional single-modality approaches. This advancement reveals that multimodal integration not only enhances detection reliability but also allows for more adaptive and timely feedback mechanisms. In contrast to earlier systems that relied on isolated data streams, our approach demonstrates the critical value of synthesizing diverse indicators for a comprehensive understanding of driver states by proactively identifying and mitigating risks associated with driver impairment, this research contributes to the broader goal of reducing road accidents and enhancing transportation safety. These researches underscore the transformative potential of intelligent monitoring systems in creating safer roads, advancing the field of intelligent transportation, and improving overall public safety.