A Multirate, Multiscale Economic Model Predictive Control Approach for Velocity Trajectory Optimization of a Heavy Duty Truck

Abstract

This paper introduces a hierarchical economic model predictive control (MPC) approach for maximizing the fuel economy of a heavy-duty truck, which simultaneously accounts for aggregate terrain changes that occur over very long length scales, fine terrain changes that occur over shorter length scales, and lead vehicle behavior that can vary over much shorter time/length scales. To accommodate such disparate time and length scales, the proposed approach uses a multilayer MPC approach wherein the upper-level MPC uses a long distance step, a long time-step, and coarse discretization to account for the slower changes in road grade, while the lower-level MPC uses a shorter time-step to account for fine variations in road grade and rapidly changing lead vehicle behavior. The benefit of this multirate, multiscale approach is that the lower-level MPC leverages the upper-level's sufficiently long look-ahead while allowing for safe vehicle following and adjustment to fine road grade variations. The proposed strategy has been evaluated over four real-world road profiles in both open-highway and traffic environments, using a medium-fidelity simulink model furnished by Volvo Group North America. Compared with a conventional cruise control system plus vehicle following controller as a baseline, results show 4–5% fuel savings in an open highway setting and 6–8% fuel savings in the presence of traffic, without compromising trip time.