Hybrid Reinforcement Learning Based Eco-Driving Approach for Connected and Automated Vehicle.

This video is a briefly visualized testing results of our Hybrid RL Eco-Driving method.
The ego-vehicle is a CAV with a front camera.
The traffic environment is a signalized intersection, with mixed human-driven vehicles which are controlled by (intelligent driver model) IDM.

The numerical experiments illustrate that our method can save 1.2% travel time and 12.25% energy consumption.

The study has been submitted to TRB2020. The preprint on arXiv will online soon.

More information at

Reinforcement Learning,Autonomous Driving,Connected and Automated Vehicle,Driving Strategy,Signalized Intersection,