Intelligent Automation of Driver Behavior Analysis Using Machine Learning: CAN Bus Signals

In modern intelligent transportation systems, detecting driver distraction and abnormal vehicle behavior is crucial for enhancing road safety. Traditional methods rely on predefined thresholds and manually designed rules, which often fail to adapt to varying driving conditions. The Controller Area Network (CAN) signal, which provides real-time vehicle telemetry data, offers a promising avenue for automated analysis. However, abnormal driving characteristics such as erratic braking, inconsistent acceleration, and unusual steering inputs require an adaptive thresholding approach to dynamically assess risk levels. Existing feature extraction techniques often struggle with real-time adaptability, leading to increased false alarms (false positives) or missed critical events (false negatives). To address these challenges, this research proposes an AI-driven automated feature extraction framework that leverages machine learning models, which is used SVM-GRBF kernel function to intelligently analyze CAN signal patterns and training input data. By integrating adaptive thresholding techniques, the system dynamically adjusts decision boundaries based on contextual factors such as vehicle speed, road conditions, and driver behavior. This study aims to bridge the gap between rule-based CAN signal processing and AI-driven automation, ensuring a more robust and intelligent detection of abnormal driving behaviors. The proposed approach seeks to optimize real-time classification accuracy while reducing false alerts. Ultimately, this advancement contributes to the development of intelligent, self-learning ADAS systems, reinforcing AI’s transformative impact on automated vehicle safety and intelligent transportation.