An energy management strategy of deep reinforcement learning based on multi-agent architecture under self-generating conditions

To improve the driving efficiency of hybrid power vehicle, an energy management strategy of deep reinforcement learning based on multi-agent architecture under self-generating vehicle driving conditions is proposed. Firstly, the kinematics segments are self-generated based on the Wasserstein generative adversarial network. The generator network G is used to generate kinematics segments. The discriminator network D is used to judge the credibility of the generated kinematics segments with the Wasserstein distance. The speed distribution characteristics of the training conditions and verification conditions established based on the self-generated segments are verified. Afterward, a multi-agent algorithm based on twin delayed deep deterministic policy gradient algorithm for hybrid systems is proposed by introducing centralized training with decentralized execution framework. The engine and a motor are used as two independent agents respectively. Different reward functions are designed based on training objectives to establish a mutually beneficial relationship of cooperation-restraint between the two agents. A driving mode constraint is designed in the environment to improve sample utilization. Finally, the simulation results demonstrate that our method can achieve better performance compared with other existing works.