Data-based controller tuning: Improving the convergence rate

Data-based control design methods most often consist of iterative adjustment of the controller´s parameters towards the parameter values which minimize an H₂ performance criterion. Typically, batches of input-output data collected from the system are used to feed directly a gradient descent optimization - no process model is used. The convergence to the global minimum of the performance criterion depends on the initial controller parameters, as well as on the size and direction of the steps taken at each iteration. This paper discusses these issues and provides a method for choosing the search direction and the step size at each optimization step so that convergence to the global minimum is obtained with high convergence rate.