Minimizing memory effects of a system

Abstract

Given a stable linear time-invariant system with tunable parameters, we present a method to tune these parameters in such a way that undesirable responses of the system to past excitations, known as system ringing, are avoided or reduced. This problem is addressed by minimizing the Hankel norm of the system, which quantifies the influence of past inputs on future outputs. We indicate by way of examples that minimizing the Hankel norm has a wide scope for possible applications. We show that the Hankel norm minimization program may be cast as an eigenvalue optimization problem, which we solve by a nonsmooth bundle algorithm with a local convergence certificate. Numerical experiments are used to demonstrate the efficiency of our approach.