[Paper Review] On the Computation of $\pi$-Flat Outputs for Linear Time-Delay Systems

Vincent Morio, Franck Cazaurang+1 more|arXiv (Cornell University)|Oct 19, 2009

Numerical methods for differential equations20 references3 citations

TL;DR

This paper presents a constructive algorithm for computing π-flat outputs in linear time-varying delay systems using polynomial matrix algebra. By extending the system ring to 𝕂(δ)[d/dt] and applying Smith-Jacobson decomposition, it characterizes π-flatness via hyper-regularity of system matrices, enabling systematic computation of flat outputs through unimodular transformations and matrix diagonalization.

ABSTRACT

This paper deals with linear time-varying, delay systems. Extensions of the concept of differential flatness \\cite{Fliess_95} to this context have been first proposed in \\cite{Mounier_95,Fliess_96} (see also \\cite{Rudolph_03,Chyzak_05}), by the introduction of $\\pi$-flat output. Roughly speaking, it means that every system variable may be expressed as a function of a particular output $y$, a finite number of its time derivatives, time delays, and predictions, the latter resulting from the prediction operator $\\pi^{-1}$. We propose a simple and constructive algorithm for the computation of $\\pi$-flat outputs based on concepts of polynomial algebra, in particular Smith-Jacobson decomposition of polynomial matrices. Some examples are provided to illustrate the proposed methodology.

Motivation & Objective

• To provide a constructive characterization of π-flatness for linear time-varying delay systems.
• To develop an algorithm for computing π-flat outputs based on algebraic matrix decomposition.
• To extend existing flatness theory to systems with time delays and meromorphic time-varying coefficients.
• To enable practical computation of flat outputs using standard computer algebra tools.

Proposed method

• The system is modeled as Ax = Bu with coefficients in the ring 𝕂[δ, d/dt], where δ is the delay operator and d/dt the time derivative.
• The ring is extended to 𝕂(δ)[d/dt], a principal ideal domain, to enable application of Smith-Jacobson decomposition.
• The Smith-Jacobson decomposition diagonalizes the system matrix into a form with invariant factors, enabling identification of free modules.
• Unimodular transformations are used to reduce the system to a canonical form, preserving module structure.
• The algorithm localizes results at powers of a polynomial π ∈ 𝕂[δ] to ensure well-posedness of predictions and delays.
• A flat output is constructed as a basis of the free module generated by the system variables, derived from the diagonalized matrix.

Experimental results

Research questions

• RQ1 How can π-flatness be characterized algebraically for linear time-delay systems with time-varying coefficients?
• RQ2 What conditions on the system matrices ensure the existence of a π-flat output?
• RQ3 How can π-flat outputs be computed constructively using standard algebraic tools?
• RQ4 What role does the Smith-Jacobson decomposition play in the computation of π-flat outputs?
• RQ5 How does the extension to the field 𝕂(δ) enable effective computation of flat outputs?

Key findings

• π-flatness is characterized by the hyper-regularity of the system matrices in the extended ring 𝕂(δ)[d/dt].
• The proposed algorithm computes π-flat outputs through a sequence of unimodular transformations derived from Smith-Jacobson decomposition.
• The method applies to systems with meromorphic time-varying coefficients, extending prior results limited to polynomial or constant coefficients.
• The computation of flat outputs relies only on operations over the larger ring 𝕂(δ)[d/dt], avoiding complex Gröbner basis techniques.
• The approach enables explicit parameterization of all system variables in terms of a flat output and its time derivatives and delays.
• The methodology is generalizable to systems with multiple delays, as demonstrated on a vibrating string model.