Unknown input estimation for a class of nonlinear systems and its application to automotive engine controls

System unmodeled dynamics and uncertainties are common issues in the design of model based controllers and observers. One way to deal with this is to design an unknown input observer to estimate those unknown variables. However it is not feasible, if measurement noises corrupt the estimator significantly. This paper proposes a new approach in the design of an unknown input estimator with proportional and integral terms. Unlike existing high gain or sliding mode based unknown input observers where the high gain is applied at the proportional error term, the proposed one applies the high gain at the integral term, which will render of less sensitivity to measurement noises without sacrificing estimation accuracy. The reduction of measurement noises effect is due to the property of the integrator that can significantly diminish measurement noises. The presented techniques can also be applied to a class of uncertain systems to estimate both the unknown states and disturbances with less sensitivity to measurement noises and less restrictive conditions than those of the previous approaches. Two case studies will be presented for the application of the proposed estimator to automotive engines: The first one is a feedback linearization controller synthesized with the unknown input observer for airpath controls of turbocharged diesel engines and the second one is to reconstruct the signal of the thermal sensor which has a slow response.