Design of energy-saving driving strategy based on proximal policy optimization considering urban transport information

Eco-driving has always been an ongoing topic. In urban driving conditions, traffic regulations, other vehicle behaviors, and special driving scenarios will have a major impact on the energy consumption of autonomous vehicles. As a representative algorithm of artificial intelligence, reinforcement learning has the ability to perform well under complex tasks. This paper uses deep reinforcement learning algorithms to design the economical driving strategies of autonomous vehicles in three driving scenarios: driving at signalized intersection under free traffic flow, car-following on ramps, and driving at signalized intersection considering queue effects. In the above three driving scenarios, the driving strategy proposed in this paper achieves economical driving performance while satisfying the driving scenario requirements.