Control Design Based on Passivity Indices with Learning

Networked control systems may include some unknown plants. But it is usually possible to access input-output data sets of such plants. As a result, methods that use input-output pairs to design the controllers such as passivity based controllers (PBC) are gaining a lot of attention. On the other hand, robustness and compositionality of PBCs make them strong candidates for system designers. In this thesis, auto-tuning PBCs with a specific measure of passiveness, named passivity indices, are discussed.

After reviewing passive and dissipative systems, passivity indices are introduced, and then controller design based on passivity indices are discussed. We show how to find passivity indices of a system which has limited frequency inputs. It is specifically useful since most of practical systems usually have only limited frequency input. Next, different methods of determining passivity indices are reviewed. Specially we review how system designer can use specific input-output sets to learn passivity indices of a system. Finally, control design and tuning using passivity indices is covered in the last chapter. Specifically we focus on auto-tuning procedure for PI controllers based on passivity indices. In summary, this thesis includes main results on control design based on passivity indices.