Optimal Finite and Receding Horizon Control for Identification in Automotive Systems

Abstract

The paper illustrates the use of optimal finite horizon and receding horizon control techniques to generate input excitation to enhance parameter identification in automotive systems. Firstly, it is shown that a Design of Experiments (DoE) problem of determining transient trajectories for off-line engine parameter identification can be posed as an optimal control problem, where either the determinant or the trace of the inverse of Fisher information matrix is minimized. Then an approach to adaptation of parameters in transient feed-forward compensation algorithms is discussed. Both Fisher information matrix computation and parameter update law utilize output sensitivity with respect to parameters being identified. Finally, a receding horizon optimal control framework for on-line parameter identification is considered where through an appropriate formulation of the cost function being minimized, it is shown that the system can be controlled to maintain tracking performance, satisfy constraints, and generate persistent excitation for parameter identification. For illustration, we use throughout an example based on the identification of transient fuel model parameters. An additional example of fast engine steady-state mapping is discussed to suggest another application of the receding horizon approach to on-line parameter identification.