To speed up gradient estimation in a slope-seeking controller two different modifications are proposed in this study. In a first approach, the gradient estimation is based on a locally identified black-box model. A further improvement is obtained by applying an extended Kalman filter to estimate the local gradient of an input-output map. Moreover, a simple method is outlined to adapt the search radius in the classical extremum-and slopeseeking approach to reduce the perturbations near the optimal state. To show the versatility of the slope-seeking controller for flow control applications two different wind tunnel experiments are considered, namely with a two-dimensional bluff body and a generic threedimensional car model (Ahmed body).

No abstract

No abstract