ZETA: a library for Zonotope-based EsTimation and fAult diagnosis of discrete-time systems
Brenner S. Rego, Joseph K. Scott, Davide M. Raimondo, Marco H. Terra, Guilherme V. Raffo
TL;DR
ZETA provides a comprehensive MATLAB library for zonotopes, constrained zonotopes, line zonotopes, and interval arithmetic to enable guaranteed, set-based estimation and fault diagnosis in discrete-time systems. It unifies linear, descriptor, and nonlinear dynamics with multiple propagation and estimation strategies (MV, FO, DC, PR) and introduces polyhedral relaxations via a Polyrelax framework to automate nonlinear enclosures. Key contributions include efficient CZ complexity reduction, robust state estimation for linear and descriptor models, active fault diagnosis using MIQP/MILP, and scalable nonlinear propagation and estimation methods. The work offers an accessible, extensible toolset for reachability, estimation, and fault diagnosis, with demonstrations and comparisons to existing tools to highlight efficiency and reduced conservatism.
Abstract
This paper introduces ZETA, a new MATLAB library for Zonotope-based EsTimation and fAult diagnosis of discrete-time systems. It features user-friendly implementations of set representations based on zonotopes, namely zonotopes, constrained zonotopes, and line zonotopes, in addition to a basic implementation of interval arithmetic. This library has capabilities starting from the basic set operations with these sets, including propagations through nonlinear functions using various approximation methods. The features of ZETA allow for reachability analysis and state estimation of discrete-time linear, nonlinear, and descriptor systems, in addition to active fault diagnosis of linear systems. Efficient order reduction methods are also implemented for the respective set representations. Some examples are presented in order to illustrate the functionalities of the new library.
