On-line optimization of RBF network feedforward compensation for load disturbance in idle speed control of automotive engine

The paper considers an application of a novel on-line nonlinear parametric optimization technique based on radial basis function (RBF) network approximation to the problem of idle speed control for an automotive engine. The control problem is formulated for a phenomenological model of an idling automotive engine. The system is highly nonlinear and includes delays in the control loop. In this paper, we assume that the load disturbance is known to the controller and demonstrate high-performance nonlinear adaptive feedforward compensation of this disturbance. The on-line parametric optimization technique applied in this paper to the design of the nonlinear adaptive feedforward controller uses an RBF network for approximating nonlinear vector fields defining the controller. The simulation results presented in this paper show that the designed feedforward controller provides good idle speed control performance and fast adaptation of the RBF network weights. The technique could be extended to other problems in powertrain control