PROBABILISTIC ROBUST OPTIMIZATION OF TWO LQ CONTROL PROBLEMS

摘要

The probabilistic robust optimization (PRO) of LQ control problems are studied in the framework of soft-bound descriptions. The improved robustness and the unity design in the parameter space are achieved by optimizing a probability-weighted sumof quadratic performances of the system with respect to possible parametric uncertainties. Two PRO problems are formulated considering real control requirements. The approach with the construction of a probability-extended system is proposed to determinethe optimal control profile and the gradient-based solution is presented to find the optimal feedback control. Numerical examples illustrate the advantages of the presented methods for systems with large parametric uncertainties.