An algorithm for amplitude-constrained input design for system identification

Abstract

We propose an algorithm for design of optimal inputs for system identification when amplitude constraints on the input and output are imposed. In contrast to input design with signal power constraints, this problem is non-convex and non-smooth. We propose an iterative solution: in the first step, a convex optimization problem is solved for input design under power constraints. In subsequent steps, the constraints considered are the p-norms of the input and output signals, p increases for each iteration step. This is an adaptation of the classical Poà ¿lya algorithm for function approximation, which has previously been used for the related problem of signal crest-factor optimization. Although the difficulty of the problem prevents a proof of optimality, the performance of the algorithm is discussed with reference to a simple example.