Rigorously computed orbits of dynamical systems without the wrapping effect

From MaRDI portal





Consider the discrete dynamic system \[ \Omega_n=f_n(\Omega_{n-1}), \quad\Omega_0\in R^d,\tag{1} \] \noindent where \(f_n\) are the maps, \(f_n:R^d\rightarrow R^d\) for stages \(n=1,2,\dots\). The goal is to construct supersets (enclosures) for the orbit of (1) such that the overestimation is kept small. (1) is not immediately suitable for the iteration on a computer. Therefore, the overestimation of the range \(f_n(\Omega_{n-1})\) at every stage \(n\) by wrapping it into a feasible superset is necessary to get a result suitable for the iteration on a computer (marching method). A collection of such sets and maps \(F_n\) operating on this collection are selected such that \(f_n(\Theta)\subseteq F_n(\Theta)\) for \(\Omega_0\subseteq\Theta_0\). Such maps are called extensions of \(f_n\) over the particular collection. The iterates are enclosures for (1). The paper presents a marching method (called cascade reduction) which beats the wrapping effect effectively and efficiently by constructing high order zonotope extensions of the orbit. A zonotope is the Minkowski sum of straight line segments. The number of line segments is always a multiple \(m\) of \(d\) here. The integer \(m\) is a freely selected measure for geometric complexity and is considered as a performance parameter for the method. The cascade reduction for finite precision, parallel computing and sparse maps are analyzed. Several examples are supplied.




Cited in
(63)








This page was built for publication: Rigorously computed orbits of dynamical systems without the wrapping effect

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1266309)