Hybrid Reinforcement Learning-Based Eco-Driving Strategy for Connected and Automated Vehicles at Signalized Intersections

Taking advantage of both vehicle-to-everything (V2X) communication and automated driving technology, connected vehicles are quickly becoming one the transformative solutions to many transportation problems. However, in a mixed traffic environment at signalized intersections, it is still challenging task improve overall throughput energy efficiency considering complexity uncertainty system. In this study, we proposed hybrid reinforcement learning (HRL) framework which combines rule-based strategy deep (deep RL) support eco-driving intersections traffic. Vision-perceptive methods integrated with vehicle-to-infrastructure (V2I) communications achieve higher mobility The HRL has three components: manager that operates collaboration between policies RL policy; multi-stream neural network extracts hidden features vision V2I information; RL-based policy generate longitudinal lateral actions. order evaluate our approach, developed Unity-based simulator designed mixed-traffic intersection scenario. Moreover, several baselines were implemented compare new design, numerical experiments conducted test performance model. show method can reduce consumption by 12.70% save 11.75% travel time when compared state-of-the-art model-based Eco-Driving approach.