Unknown input identification in nonlinear systems based on interval observers

Abstract This paper addresses the problem of the unknown input estimation (identification) in technical systems described by nonlinear dynamic models, taking into account measurement noise. Such an estimate is needed to compensate for their effect. The solution of the problem is based on the reduced‐order model of minimal dimension of the original system. Such a model is used to design the interval observer, which, together with a differentiator, generates the low and upper bounds of the interval estimate. It is shown that the virtual sensor can be used in specific cases to solve the problem. The theoretical results are illustrated by a practical example of estimating the unknown inputs in the electric drive of a manipulation robot. The simulation based on the MATLAB package confirmed the correctness of the theoretical results. The obtained relations can be used to design fault‐tolerant systems.