Using driver behavior classification to implement an H2-powered motor control system

Hydrogen-powered vehicles, particularly those using internal combustion engines (H2-ICE), represent a strategic solution for reducing emissions in the transportation sector. However, optimizing their energy efficiency remains a challenge, mainly due to variations in individual driving behavior. Aggressive driving styles lead to suboptimal fuel consumption, highlighting the need for adaptive engine control systems. In this study, we introduce an innovative machine learning-based control module designed specifically for hydrogen-powered engines. We implement a new physics-based labeling strategy to classify driving behavior as "aggressive"or "cautious"using vehicle dynamics from a comprehensive real-world dataset. We systematically compared six supervised learning classifiers, including Support Vector Machines, Decision Trees, Random Forests, Gradient Boosting, K-Nearest Neighbors, and Logistic Regression, taking into account accuracy, latency, and interpretability, which are critical for embedded automotive systems. Random Forest and Gradient Boosting emerged as optimal candidates due to their superior performance metrics and balanced computational complexity. Our results provide concrete guidelines for the implementation of real-time, behavior-sensitive control systems that improve the efficiency and adaptability of hydrogen mobility, laying a solid foundation for future implementations on embedded hardware.