Reinforcement Learning for Truck Eco-Driving: A Serious Game as Driving Assistance System

Making fuel-economy for vehicles is an important and current challenge in particular for professionals of transportation. In this article, we address the challenge of providing a driving serious game based on artificial intelligence in order to significantly reduce the fuel consumption for trucks. Our proposition is based on a machine learning process consisting of a Self-Organizing Network (SOM) for clustering and subsequent reinforcement learning to deliver precise recommendations for eco-driving. Driving experts provide us knowledge in order to model the actions-rewards process. Experiments conducted on simulated data demonstrate that the recommendations are coherent and enable drivers to adopt eco-driving behavior.