Data-driven co-optimization method of eco-adaptive cruise control for plug-in hybrid electric vehicles considering risky driving behaviors

Plug-in hybrid electric vehicles (PHEVs) with adaptive cruise control (ACC) systems are crucial for energy-efficient travel. However, dealing with risky driving behaviors of the preceding vehicle (PV) while optimizing mobility and energy management poses significant challenges. This study proposes a data-driven co-optimization method of eco-ACC for PHEVs that integrates potentially risky driving behaviors (RDBs) detection, motion prediction, and cost-based co-optimization. An LSTM-based autoencoder model is developed to identify potential signals of PV's RDBs, and an LSTM-based predictor is established to forecast the PV's future motion state. Additionally, the concept of environmental treatment cost is introduced, and a driving cost estimation model is designed using an improved extreme learning machine to streamline optimization. The results demonstrate that the proposed method accurately detects RDBs of the PV and proactively adjusts vehicle distance. Its tracking performance surpasses rule-based models and closely matches optimization-focused models. It achieves 6.47 %–15.86 % improvements in energy saving and emission reduction compared with existing real-time methods, while approaching the performance of theoretical optimal methods. Hardware-in-the-loop test validates its real-time feasibility and deployment potential.