The paper proposes an hybrid Kalman filter integrating a robust and optimal algorithm for the use as an observer in model-varying predictive control (MVPC) of a nonlinear system. Moreover, a position MPC is derived in detail. Even though the proposed approach is quite general, a real case coming from automotive application is studied using computer simulation to demonstrate the effectiveness of the proposed technique. Simulations and results with real data are also discussed. © 2004 IEEE.