Model Predictive Control of Automotive Powertrains - First Experimental Results

This paper illustrates the capabilities of model predictive control for the control of automotive powertrains. We consider the minimization of the fuel consumption of a gasoline engine through dynamic optimization. The minimization uses a mean value model of the powertrain and vehicle. This model has two state variables: the pressure in the engine manifold and the engine speed. The control input is the throttle valve angle. The model is identified on a universal dynamometer. Optimal state and control trajectories are calculated using Bock's direct multiple shooting method implemented in the software MUSCOD-II. The developed approach is illustrated both in simulation and experimentally for a test case where a vehicle accelerates from 1100 rpm to 3700 rpm in 30 s. The optimized trajectories yield minimal fuel consumption. The experiments show that the optimal engine speed trajectory yields a reduction of the fuel consumption of 12% when compared to a linear trajectory. Thus, it is shown that, even with a simple model, a significant amount of fuel can be saved without loss of the fun-to-drive.