Adaptive Pulse-and-Glide for synergistic optimization of driving behavior and energy management in hybrid powertrain

Eco-driving has been extensively studied and recognized as an effective method for enhancing energy efficiency. However, much of the existing research focuses solely on optimizing driving behavior. Although there is growing interest in integrating eco-driving strategies with energy management systems (EMSs), synergistic optimization of driving behavior and energy management remains relatively underexplored. This study proposes a novel deep reinforcement learning (DRL)-based eco-driving strategy for hybrid electric vehicles (HEVs), termed Adaptive Pulse-and-Glide (A-PnG). A-PnG operates as a centralized neural network-in-the-loop system, simultaneously planning the longitudinal driving profile of the vehicle and generating equivalence factor (EF) signals to guide the Equivalent Consumption Minimization Strategy (ECMS) in managing energy flow within the powertrain. By coupling the energy-saving potential of the Pulse-and-Glide (PnG) technique with real-time energy optimization, A-PnG achieves both a stable State of Charge (SOC) and efficient powertrain control. Experimental evaluations reveal that A-PnG demonstrates robust adaptability across diverse driving conditions, including standard driving cycles and long-duration tests, and showcases remarkable performance across a range of initial SOC levels. In fuel economy benchmarks, A-PnG achieves 89.55 % of the theoretical optimum set by Dynamic Programming (DP), while Charge Depleting Charge Sustaining (CDCS) exhibits 5.62 % higher energy cost compared to our approach. With excellent ride comfort and a latency of less than 0.25 ms per decision, A-PnG shows its superior applicable potential as an efficient eco-driving solution.