Data-Driven Prediction and Predictive Control Methods for Eco-Driving in Production Vehicles

Abstract

This paper presents a study of perception and robust model-predictive control (MPC) strategies in realistic traffic environments, which are simulated using data from real-world driving experiments. In this paper, we consider a heterogeneous traffic environment, which includes human-driven vehicles, and study the performance of currently available automation in production vehicles. We then present a data-driven preceding vehicle's velocity and position prediction algorithm, and a robust MPC strategy that optimizes fuel consumption and takes into account the prediction errors. Data used in this paper are taken from experiments using a 2018 Cadillac CT6 vehicle. Simulation results show up to 6.39% energy efficiency improvement.