Robust Generalized Predictive Control with Uncertainty Quantification

Abstract

DOI: 10.2514/1.54510 This paper presents a robust predictive control method with uncertainty quantification for problems, such as online aircraft flutter suppression. The proposed approach begins by identifying the multistep output prediction matrices under various test conditions from online time-domain input/output data. Singular value decomposition is then used to characterize and quantify the parameter uncertainties of the prediction matrices. The predicted responseerrorduetoparameteruncertaintiesisintegrated intothepredictivecontroldesignprocessforrobustness. The proposed method provides an innovative and efficient way to incorporate the quantified uncertainty in predictive control design with a linear process. The proposed approach is demonstrated by application to the Benchmark Active Controls Technology wind-tunnel model.