Fast and robust adaptation of lookup tables in internal combustion engines: feedback and feedforward controllers designed independently

Abstract

Adaptive feedforward control is a common approach for handling uncertainties and time-varying effects in automotive control applications. The adaptation of the feedforward controller, usually formed by a lookup table, is often combined with a linear feedback controller. The feedforward controller is used to overcome the nonlinearities that are due to variations of the operating point. The feedback controller is used to deal with fast disturbances. If the system behavior is changing, for example due to varying fuel qualities or aging effects, the feedforward controller has to be adapted. A common problem is the fact that the bandwidths of the feedback controller and the adaptation of the feedforward controller have to be separated, i.e. the adaptation of the feedforward controller has to be slower than the feedback controller. Otherwise, the coupling effects lead to an unexpected behavior of the control system, which may also lead to instability. This paper presents an analysis of this effect using a linear representation of the adaptation of the feedforward controller. Based on the results of this analysis, the paper presents a method to decouple the feedback controller and the adaptation of the feedforward controller, which allows a fast adaptation of the feedforward controller. Ideally, the adaptation is as fast as the feedback controller. The system is then able to adapt during operating point transients without the need to stay at some operating points for longer periods. The proposed method does not depend on the structure of the feedforward controller, nor does it depend on the method of the adaptation. Experiments on a natural gas-diesel dual-fuel engine are used to validate the proposed method.