Induced norm analysis of linear systems for nonnegative input signals (Q6884597)

The paper studies induced norms of continuous-time linear systems when the input signals are restricted to be nonnegative. In this setting the authors introduce the notion of the \(L_p^+\) induced norm, which is defined similarly to the classical \(L_p\) induced norm but with the supremum taken over nonnegative input signals. This concept arises naturally in several contexts, including the analysis of positive systems, the evaluation of differences between positive models, and the stability analysis of nonlinear feedback systems in which the nonlinear components return nonnegative signals.\N\NThe work focuses mainly on the analysis of the \(L_2^+\) induced norm. Since its exact computation is generally difficult, the paper develops methods for estimating both upper and lower bounds. For the upper bound analysis, the authors propose a framework based on copositive programming. The nonnegativity of the input signals is captured through copositive multipliers, leading to a sequence of optimization problems that produce progressively tighter upper bounds. In order to obtain numerically tractable formulations, inner approximations of the copositive cone are employed, resulting in semidefinite programming relaxations.\N\NThe paper also investigates the relationship between the standard \(L_p\) induced norm and the \(L_p^+\) induced norm. In particular, it derives the uniform infimum of the ratio \(\|G\|_p^+ / \|G\|_p\) over the class of stable linear time-invariant systems, including infinite-dimensional ones. Explicit expressions for this infimum are obtained for all \(p \in [1,\infty]\).\N\NFinally, the authors address the computation of lower bounds of the \(L_2^+\) induced norm for a given system. By exploiting frequency-domain properties and reducing the problem to a semi-infinite programming formulation, they develop procedures for estimating such bounds. The methods are discussed for both single-input and multi-input systems. Numerical examples are provided to illustrate the computation of the proposed bounds and their application in representative system analysis problems.