A novel approach is presented to the problem of robust control strategy for an Exhaust Gas Recirculation (EGR) system. Modeling issues and controller design for an electric valve are presented. Robustness is one of the main issues due to nonlinearities (hysteresis, friction), computational time delays, and the effects of complex exhaust gas dynamics. A high level of noise, due to engine vibrations, makes the input-output model identification very difficult. We use a special windowing technique to successfully complete the identification. It was found that a good approximation of the system dynamics consists of a first order transfer function with a time delay whose parameters depend on the operating conditions (engine load, r.p.m.). Thus, the nonlinear effects of friction and hysteresis are replaced, for the purpose of control design, by variable gain, time delay, and time constant of the linear model. Replacing friction and hysteresis, especially when they vary in time (mostly due to temperature changes), may seem a bit controversial. However, an alternative is to model precisely these nonlinear time varying and sometimes random processes which is a very costly and unreliable approach. We propose instead a less accurate modeling compensated by more robust and more intelligent control. We found that it is sufficient to identify only nine models corresponding to nine different operating conditions and assume linear interpolation for transfer function coefficients to characterize the system dynamics in a general case. Thus, the controller design can be reduced to nine controllers corresponding to nine identified models and interpolation for coefficients of the controller transfer function. Identification and simulation results for such designs are presented as well as the results of laboratory experiments.