Adaptive Driver-Centric ADAS Using Machine Learning for Personalized Assistance

Advanced Driver Assistance Systems (ADAS) are becoming increasingly integral to vehicle safety; however, most current implementations rely on static algorithms with fixed thresholds that do not account for individual driver behavior. This can lead to false positives or unnecessary alerts, thereby reducing user trust and system effectiveness. This paper presents a Driver-Adaptive ADAS framework that leverages machine learning to dynamically tailor assistance logic based on real-time driver behavior and driving style. Instead of treating all drivers uniformly, the system learns patterns such as reaction time, maneuver preferences, and braking tendencies. It adjusts key parameters like Time-To-Collision (TTC) thresholds, Forward Collision Warning (FCW) timings, and Automatic Emergency Braking (AEB) triggers accordingly. A software-based proof of concept is developed using Python and MATLAB, where simulated vehicle and driver behavior data are processed to classify driver types and adapt the ADAS response logic. The approach enables personalized intervention strategies for different driver profiles—such as defensive, assertive, or distracted—enhancing system reliability and minimizing driver irritation. The proposed framework is modular and scalable, designed for eventual integration into production ADAS software pipelines. It bridges the gap between fixed-rule systems and intelligent, human-aware safety solutions by embedding adaptability into the core decision-making logic. This work demonstrates that incorporating machine learning-based driver adaptation into ADAS can significantly improve safety performance and driver acceptance, laying a strong foundation for the next generation of intelligent driver assistance systems.