| Summary: | 碩士 === 國立成功大學 === 航空太空工程學系 === 102 === The linear quadratic (LQ) optimization is a known approach for control system synthesis. In addition, a LQ control design can also be conducted based sorely on the open-loop plant test data, when a plant dynamic model is not explicitly known. On the other hand, the presence of a nonzero penalty on the control input causes an error to appear in the closed-loop output. In order to achieve a perfect command following operation, a LQ control design must be performed without penalizing the control input. However, the removal of the penalty on the control input also brings the information matrix of the data-based LQ (DBLQ) control design close to singular. In this work, the numerical difficulty of such a DBLQ computation is resolved by developing an ultra-precision (UP) arithmetic package and by conducting DBLQ computations using the UP package.
|
|---|
