In this paper, a sensorless control of an electromagnetic valve actuator for automotive applications is presented. Based on a nonlinear model, a switching esti mator combined with a two-stage observer structure is proposed. The presented structure is basically a high-gain observer (HGO) to reduce the effects of the uncertainties on the electromagnetic valve actuator model and unmeasurable external disturbances coming from the burning phase of the engine and acting against the actuator movement. The proposed sliding-mode observer has the capability of guaranteeing zero average estimation error in finite time, even in the presence of model uncertainties and bounded disturbances. Considerations on constructive aspects concerning observability are made. Two constructive propositions are proven in this paper to realize the observer. A general comparative analysis is proposed, considering other contributions which use Kalman filters, in which aspects of optimality, robustness, and chattering problems are discussed. In particular, an analysis of the computational effort is also reported. Laboratory experiments of the controlled system, which demonstrate promising behavior, are presented and discussed for exhaust valves of an engine.