Energy-aware optimization of electric vehicles’ dual-motor coupled powertrain based on heterogeneous synchronous reinforcement learning

Dual-motor coupled powertrain (DMCP) is a promising drive type that can improve the energy utilization efficiency and driving range of electric vehicles (EVs). In this paper, a heterogeneous synchronous reinforcement learning (HSRL) approach is proposed to optimize the energy utilization efficiency of the DMCP in EVs. First, the models of the DMCP and vehicle dynamics are established. Then, the HSRL framework is constructed to select the drive mode and determine the torque allocation coefficient of the DMCP simultaneously. Within the HSRL framework, an actor-critic network is adopted, and a cross-domain learning strategy is proposed. The cross-domain learning strategy incorporates discrete domain learning (DDL) and continuous domain learning (CDL) to learn both the discrete and continuous decision-making tasks concurrently. Additionally, gradient strategies for DDL and CDL are designed. Based on the proposed HSRL, the energy-aware drive mode and torque allocation coefficient of the DMCP are selected and determined. To demonstrate the superiority of the proposed method, two state-of-the-art (SOTA) methods are chosen as benchmarks. The simulation and experimental results show that the proposed method outperforms these benchmarks in optimizing the energy utilization efficiency of the EVs’ DMCP.