IFAC PapersOnLine 52-5 (2019) 279–284 ScienceDirect ScienceDirect Available online at www.sciencedirect.com 2405-8963 © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. Peer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2019.09.045 © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. 1. INTRODUCTION In a 2015 survey, the Department of Energy, DOE (2015), estimated that the average annual fuel use of Class 8 trucks in the US is almost 13 thousand Gasoline-Gallon Equiva- lents (GGEs) and that the average annual mileage for the same vehicle category stands at 68 thousand miles per ve- hicle. In this scenario, even a relatively small fuel economy improvement will result in a significant reduction of fuel, emissions, and vehicle operating costs. As Connected and Automated Vehicles (CAVs) are becoming more popular, there is potential for leveraging the available traffic and road information to reduce vehicles’ fuel consumption. For instance, Uhlemann (2015) shows that Vehicle-to-Vehicle (V2V) and Vehicle-to-Infrastructure (V2I) communica- tions allows for a systematic implementation of intelligent safety features as well as eco-driving. For this reason, a significant body of literature has been focusing on utilizing GPS and traffic information for com- bined vehicle velocity and powertrain optimization. For instance, in Hellstr¨om et al. (2009) the authors extend the work presented by Fr¨oberg and Nielsen (2007) and implement a Look-Ahead powertrain controller for the optimization of the vehicle speed, gear selection and en- gine fueling command of a heavy duty truck under the assumption that the road and the road grade is know.  The authors would like to acknowledge the ARPA-e NEXTCAR program (Award number: DE-AR0000801) and thank them for their financial support. The authors would also like to thank Volvo Trucks North America for their support. Experimental results showed a reduction of both time and fuel compared to the production cruise control. A similar result has been presented by van Keulen et al. (2010) for a hybridized truck under the assumption that the gear selection was an exogenous input. The solution of the optimization problem was implemented as a veloc- ity advisory to the truck driver and showed a consistent fuel consumption reduction for different repetitions of the cycle. The work was then extended by Ngo et al. (2013) to focus on the gear selection in an automated-manual transmission with drivability considerations. To address the computational complexity of the vehicle velocity and gear selection algorithms presented above, Larsson et al. (2015) proposes an analytical approximation of the cost- to-go function, while Xu et al. (2018) uses the optimal solution obtained off-line to train a neural network. When vehicle and gear selection optimization algorithms, such as the ones discussed above, are to be coupled with the Engine Control Unit (ECU), the issue of the inconsis- tent sampling time must be addressed. Specifically, while the relevant dynamics for the power train optimization are in the order of seconds, the ECU operates at a much faster sampling rate (in the order of milliseconds). In order to make the powertrain optimization problem tractable, fast engine dynamics are neglected, and it is often assumed that the engine torque production is instantaneous, and the torque is only saturated based on engine speed. In addition, because the ECU takes the reference engine torque and converts it into set points for the actuators’ positions, neglecting the engine fast dynamics can result Keywords: Diesel engines, model based control, predictive control, connected and automated vehicles Abstract: As connected and automated vehicles are becoming more popular, there is potential for leveraging the available traffic and road information to reduce vehicles’ fuel consumption. Many optimization strategies have been proposed at the powertrain level, providing a slow torque reference setpoint to the engine controller. This paper explores the opportunities provided by the different controller time scales to provide a fast torque reference accounting for (i) torque smoothing and avoiding actuation saturation, (ii) fuel consumption reduction while meeting constraints, and (iii) drivability. The control oriented model of the engine and its validation is presented first. Then, the control hierarchy is outlined. Then the input shaping and reference generator solution strategies are presented. Finally, results for each solution method are discussed, showing about 0.17% fuel consumption increase for the input shaping algorithm and about 2.5% of fuel consumption reduction with the predictive reference generator, and substantial improvements in drivability for each. * Department of Mechanical Engineering The Pennsylvania State University, State College, PA 16802 (e-mail: sfb5244@psu.edu). Stephen Boyle * Stephanie Stockar * Comparison of Input Shaping and Predictive Reference Generator Techniques for IC Engine Setpoints Commands 