Parametrization and adaptation of gasoline engine air system model via linear programming Support Vector Regression

Air charge estimation is an essential task for gasoline engine control, as its performance determines that of the air-fuel-ratio control and torque control, thereby dictating the fuel economy and emissions of the vehicle. While the problem of air charge estimation has been addressed by the automotive and control communities for many years, assuring adaptivity and robustness of air charge estimation continues to be a challenge, especially as performance requirements become more stringent. In this paper, we propose a new air system model based on Support Vector Regression (SVR). The model leads to a new parameterization which facilitates effective adaptation with simple update laws. Simulation and experiment results demonstrate its real-time implementation performance, computational efficiency, and calibration simplicity.