Adaptive Eco-driving of Fuel Cell Vehicles Based on Multi-light Trained Deep Reinforcement Learning

With the advent of modern communication technology, future driving environment information such as the coming traffic lights can help vehicles make more intelligent planning and decision-making. This study proposed an adaptive eco-driving method of Fuel Cell Vehicles (FCVs) based on multi-light trained Deep Reinforcement Learning (DRL), aiming to improve the fuel economy and comfort of FCVs. By setting a reasonable reward function, the DRL agent takes actions to maximize the cumulative reward according to the environmental information received. An application method was further studied to fulfill the eco-driving of vehicles in complex arterial road scenarios. The result shows that the proposed method can adapt to variable scenarios with traffic lights, and quickly solve the approximate optimal speed trajectory. Hydrogen consumption of multi-light trained DRL models is reduced by more than 10% on average compared with that of the single-light trained DRL model.