Realization of the leanburn SI engine's potential for improved fuel economy strongly depends on precise control of the air-fuel ratio (AFR), especially during transients, for acceptable driveability and low exhaust emissions.The development of an adaptive-feedforward model-based AFR controller is described. A discrete, nonlinear, control-oriented engine model was developed and used in the AFR control algorithm. The engine model includes intake-manifold airflow dynamics, fuel wall-wetting dynamics, process delays inherent in the four-stroke engine cycle, and exhaust-gas oxygen (UEGO) sensor dynamics. The sampling period is synchronous with crank-angle (“event-based”) for more precise control.The controller relies on the engine speed and throttle position for load information. An intake-manifold pressure (MAP) sensor is used for identification of the airflow dynamics, but not for control. The MAP sensor would also be useful for the cold start and for engine diagnostics. No mass-airflow sensor is used nor indicated. Accurate in-cylinder air-mass estimation is made possible by controlling the airflow with an electronic throttle.An adaptive model for the airflow dynamics has been developed. The key dynamic parameter in the airflow model is determined by a linear least-squares identification technique using information from the throttle position and MAP sensors. The adaptive model-based control structure minimizes the need for engine controller calibration and ensures that the control accuracy is maintained throughout the lifetime of the engine.A 4 cylinder, 2.2 liter, port fuel-injected engine with drive-by-wire throttle was used to demonstrate this model-based AFR control structure. Air-fuel ratio followed the commanded value to about 0.8% RMS, with a peak error of 2% during throttle and RPM transients at various operating points and AFR set points. The transient AFR control during lean engine operation was accurate enough to prevent the occurence of misfires, even though the engine was running near the lean limit.