Scalable Real-Time Predictive Eco-Driving based on a Rollout Algorithm

The deployment of real-time predictive eco-driving control algorithms remains an active research area, primarily due to their computational complexity. This study addresses this challenge by examining the scalability of a Dynamic Programming based eco-driving algorithm. We propose a parallelized framework that leverages the multi-core and multi-threaded architecture of modern CPUs. The framework is then deployed on a low-cost compute platform to offer a practical solution to the computational and financial barriers associated with integrating eco-driving technologies into modern Connected and Automated Vehicles (CAVs). This work demonstrates the potential for more accessible and widespread adoption of these technologies.