Adaptive Nonlinear Model Predictive Control of an Engine Air Path

It is well known that high performing control based on a nominal model may underperform when system parameters change e.g. due to wear. Adaptive controls are a well established way to tackle this problem, but their usage is not trivial, especially when they are used for complex nonlinear systems with constraints like the air path of a combustion engine. In particular, excitation can be insufficient for the update of a model with many parameters. Against this background, we extend and test an approach based on directional forgetting on a production engine, showing that this approach is viable for practical applications.