Bias reduction in the Optimal Controller Identification approach through optimal filtering

Non-iterative data-driven controller design methods are very convenient for their ease of implementation and for only using a collected data set for the design. One of these methods is Optimal Controller Identification (OCI), which uses the Prediction Error Method to identify a model of the system parameterized by the controller's parameters and the desired response of the controlled system, the Reference Model. The Virtual Reference Feedback Tuning (VRFT), another data-driven method, has a widely used filter that benefits its performance in the case of designing underparameterized controllers, but the same has not been yet presented for OCI. In this paper, a specific filter for OCI is developed for the underparameterized controller case and tested in a series of situations, resulting in a gain of performance.