On the Robustness of Adaptive Nonlinear Model Predictive Cruise Control

In order to improve the vehicle’s fuel economy while in cruise, the Model Predictive Control (MPC) technology has been adopted utilizing the road grade preview information and allowance of the vehicle speed variation. In this paper, a focus is on robustness study of delivered fuel economy benefit of Adaptive Nonlinear Model Predictive Controller (ANLMPC) reported earlier in the literature to several noise factors, e.g. vehicle weight, fuel type etc. Further, the vehicle position is obtained via GPS with finite precision and source of road grade preview might be inaccurate. The effect of inaccurate information of the road grade preview on the fuel economy benefits is studied and a remedy to it is established. It is shown that the effect of scale and value bias error in the road grade preview can be eliminated by the on-line adaptation of the model parameters performed by the constrained Recursive Least Squares (RLS) method and the estimation of the additive acceleration by the Extended Kalman Filter (EKF). The effect of phase error in road grade preview is eliminated by the iterative improvement of the estimation of the vehicle’s true position over the grade map. The algorithm is based on minimization of vehicle model fitting residuals over a past horizon, which is formulated as a one-dimensional nonlinear optimization problem solved by Secant method at each ANLMPC sample time. Success in grade error correction is verified by both model fitting quality and the resulting control performance benefit. The ANLMPC with road grade preview error correction has been validated in simulation with model of light duty truck towing a 4500kg trailer showing up to 10% fuel economy improvement compared to production cruise controller with the same time of arrival.