Application of modern techniques to SI-engine torque control

Abstract

This contribution investigates the application of three modern design techniques to the torque control problem of a spark-ignited direct injection engine. The system to be controlled is a highly nonlinear system characterized mainly by the intake manifold dynamics. The first scheme applies feedack linearization in order to achieve a linear and uniform system behaviour over the whole range of operation. The second approach, nonlinear model predictive control, optimizes the control law over a finite time horizon taking the input and state constraints into account during optimization. The third approach is a gain-scheduled LQ-optimal control scheme based on the state-space description of the system. A comparative study of these schemes with the help of computer simulations is presented. The schemes are compared for achieved performance, computational cost, and implemetability on an electronic control unit.