Adaptive Input and Parameter Estimation with Application to Engine Torque Estimation

This paper presents two estimation methods for systems with unknown time-varying input dynamics. By defining auxiliary filtered variables, an invariant manifold is derived and used to drive the input estimator with only one tuning parameter. Exponential error convergence to a small compact set around the origin can be proved. Robustness against noise is studied and compared with two well-known schemes. Moreover, when the input dynamics to be estimated are parameterized in a quasilinear form with unknown parameters, the proposed idea is further investigated to estimate the associated unknown time-varying parameters. The algorithms are tested by considering the torque estimation of internal combustion engines (ICEs). Comparative simulation results based on a benchmark engine simulation model show satisfactory transient and robustness performance.