Keywords:
windsurfer, fully probabilistic control design, Jobcontrol,
Anotation:
A control design converts knowledge about the controlled system, constraints and control aims into the controller. In fully probabilistic control design (FPD), the desired system behaviour (ideal) is ex- pressed by a probability density function of respective quantities and the controller is selected to minimize the Kullback-Leibler divergence of probability density functions describing actual and ideal closed-loop behaviour. However, ideal behaviour of the closed loop must be realistic. Therefore, the ideal is adapted according to the observed behaviour and gradually modified towards the desired control aim. This iterative process, called windsurfer approach, results in reaching of the control design respecting properties of the controlled system and ability to control it using the given means and constraints. The windsurfer approach to control aim elicitation was integrated into the Jobcontrol package, a UI to Mixtools for normal mixtire-based identification and control.