betaFeedback

WorksMathematics

A Lifted Hessian-Riemannian Gradient Flow for LQR Synthesis under Passivity Constraints

Abstract

The static state-feedback gain that renders a linear system passive at an external port and minimizes a linear quadratic regulator (LQR) cost cannot be computed by convex optimization, since the passivity certificate and the LQR cost are coupled bilinearly through the common gain. A previous indirect approach restricts the gain to a convex inner approximation of the passivity region, which is conservative and whose construction scales exponentially with the number of gain entries. This paper proposes a direct approach. The set of strictly passivating gains is the image of a convex lifted set of inverse storage matrices and transformed gains. We describe this set by a strict linear matrix inequality and an affine equality for every feedthrough matrix with positive semidefinite symmetric part, including feedthrough matrices with a singular symmetric part, and minimize the LQR cost over it by a gradient flow whose metric is the Hessian of a logarithmic barrier. For multi-input systems, this Hessian is singular along gain changes that preserve the storage function, and we complete the metric along these directions by the metric of the gain space. We prove that the flow exists for all times, that every gain along the flow is strictly passivating with an explicit storage function, and that the cost decreases monotonically while the gains remain in a compact set of stabilizing gains. Numerical examples show that the flow attains a lower cost than the indirect approach, that the completion is required for multi-input systems, and that the flow can converge to a local minimizer on nonconvex passivity regions.

Paper details

vlnc:W_ggcd3bq9
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
arXiv nonexclusive license
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Not recorded for this paper

View statement
Looking for related papers…