Abstract

Artificial intelligence (AI) provides a powerful framework for improving energy efficiency in hybrid electric vehicles (HEVs). This paper presents an AI-driven approach that integrates driver behavior modeling into the design of energy management strategies (EMS). Real-world driving data were collected from a Toyota Prius using Controller Area Network (CAN) logging during repeated urban driving cycles. Unsupervised learning methods—including k-means, hierarchical, and fuzzy clustering—were applied to classify driver input patterns such as speed and acceleration. A validated parallel hybrid powertrain model was developed, and a genetic algorithm (GA)-based EMS was designed to optimize internal combustion engine (ICE) control under forecasted driver behavior. Simulation results show that AI-enabled, driver-aware EMS can achieve fuel consumption reductions of up to 12% compared with conventional rule-based strategies. The results highlight the potential of AI to bridge human driving patterns and intelligent vehicle control, moving toward adaptive, real-time EMS for sustainable mobility.

I. Introduction

Hybrid electric vehicles are a cornerstone of sustainable mobility. However, conventional energy management strategies are limited by static rule sets that do not account for human driving variability. Artificial intelligence offers a path toward adaptive, data-driven EMS design. This work explores how unsupervised learning and evolutionary algorithms can extract knowledge from driver behavior and transform it into optimized control strategies.

II. Methodology

– Data Collection: CAN-Bus logging from a Toyota Prius during urban driving cycles.
– AI-based Behavior Modeling: Unsupervised learning (k-means, hierarchical, fuzzy clustering) to categorize driver styles.
– Simulation Model: Parallel hybrid powertrain model validated against real data.
– Optimization: GA-based EMS that adapts ICE operation to clustered driver profiles.

III. Results and Discussion

Driver behavior clusters revealed clear efficiency differences between aggressive and conservative styles. When integrated into the GA-based EMS, fuel consumption was reduced by 8–12% compared to baseline strategies. These results validate AI as a driver-aware optimization tool, showing the potential for on-board adaptive EMS.

IV. Conclusion

This study demonstrates that AI methods—unsupervised learning and evolutionary optimization—enable intelligent EMS design by incorporating real-world driver behavior. Future work will focus on deploying real-time, adaptive AI controllers for HEVs, paving the way toward self-learning, driver-aware vehicle systems.

References

[1] L. Guzzella and A. Sciarretta, Vehicle Propulsion Systems, Springer, 2007.
[2] H. He, B. Wang, and J. Sun, “Application of AI techniques in hybrid electric vehicle energy management,” IEEE Trans. Veh. Technol., vol. 65, no. 6, pp. 4593–4605, 2016.
[3] C. Sun et al., “Deep reinforcement learning for hybrid vehicle energy management,” IEEE Trans. Control Syst. Technol., vol. 27, no. 6, pp. 2363–2370, 2019.