Inhomogeneous Model Predictive Control Horizon Discretization for an Urban Truck Energy Efficient Driving Application

Abstract

This paper presents a novel approach on Model Predictive Control (MPC) using an inhomogeneously discretized preview horizon for the application of urban energy efficient driving. One solution for model predictive energy efficient driving is a direct solution of the underlying speed profile optimization problem using Quadratic Programming (QP), which allows computationally efficient and robust results. Our inhomogeneous horizon discretization allows to have a finer discretization of the typically important near future and a wider discretization of the less decisive far range of an MPC, while keeping a long preview horizon and at the same time limit the number of supporting points, hence limit the problem dimension, computational complexity, and proportional execution time. In extensive simulations of a real-world urban driving scenario, we demonstrate a significantly improved control performance in terms of fuel consumption, trip time, or constraint violation for the same computational complexity.