Nonlinear controller tuning based on linearized time-variant model

Presents an algorithm for tuning controllers for nonlinear plants. The algorithm iteratively minimizes a criterion of the control performance. In each iteration two experiments are performed with similar reference signals. The differences of the signals from these experiments are used to identify a linear time-variant model which is then used to calculate an update of the controller parameters. The algorithm requires an initial controller that stabilizes the plant so that similar reference signals give similar output signals