Idle Speed Control plays a crucial role to reduce fuel consumption that turns in both a direct economic benefit for customers and CO\d reduction particularly important to tackle the progressive global environmental warming. Typically, control strategies available in the automotive literature solve the idle speed control problem acting both on the throttle position and the spark advance, while the Air-Fuel Ratio (AFR), that strongly affects the indicated engine torque, is kept at the stoichiometric value for the sake of emission reduction. Gasoline Direct Injection (GDI) engines, working lean and equipped with proper mechanisms to reduce NOx emissions, overcome this limitation allowing the AFR to be used for the idle speed regulation.In this paper, an effective model of the GDI engine dynamics is derived, tuned and then used to synthesize a gain scheduling control strategy which comprises a feedback action acting on the throttle position, and a feedforward compensator which varies dynamically the demand of the AFR control task. The former control action is mainly exploited to accomplish smooth transitions from/to idle speed regime, whereas the latter copes with torque disturbances at idle speed mainly due to the intermittent use of accessory loads. In so doing, a faster actuation path, provided through the AFR control, is added to the air control path to increase performance both in terms of disturbance rejection and fuel economy. Comparison between performance provided by our control approach and a classical LQ strategy, which controls both the throttle angle and the spark advance when the AFR is kept at the stoichiometric value, confirms the effectiveness of the proposed control architecture with respect to different cost indexes.Model validation as well as the effectiveness of the control design are carried out by means of ECU-1D Engine Co-Simulation tools. The combination in a one integrated designing environment of control systems and virtual engine, simulated through high predictive commercial 1D-code, becomes a high predictive tool for automotive control engineers and fast prototyping.