Real-time Eco-Driving Algorithm for Connected and Automated Vehicles using Quadratic Programming

The Eco-driving problem aims to reduce the energy consumption of a Connected and Automated Vehicle (CAV) over an itinerary through velocity planning, by exploiting connectivity-enabled look-ahead information. In this work, an Eco-driving algorithm was developed for CAVs using Quadratic Programming (QP). The proposed Model Predictive Control (MPC) scheme computes the energy-efficient velocity trajectory in real-time by leveraging information from traffic lights and predictively modeling the surrounding traffic. The resulting framework compactly formulates and applies the optimization problem constraints simultaneously in the QP setting (rather than sequentially). This algorithm is validated for a plug-in hybrid electric vehicle using extensive closed-loop traffic simulations based on a real-world corridor model in San Antonio, TX. Across these simulations, the Eco-driving framework reduces the energy consumption of the ego vehicle by over 12%. Further, experimental evaluation on a hub dynamometer for downselected cases resulted in over 16% energy savings compared to the baseline.