Section: Research Program
Applications to control theory and signal processing
Robust stability analysis and stabilization problems for functional systems
Our expertise in the computer algebra aspects to stability and stabilization problems for multidimensional systems and for differential constant/distributed/varying delay systems , , ,  will further be developed.
Computation of Lyapunov functions for homogeneous dynamical systems
We shall investigate the possibility to develop a computer algebra package for the design of Lyapunov functions for homogeneous dynamical systems based on an effective study of the differential algebra of generalized forms, i.e. of Puiseux polynomials in signed powers , .
Robust stability analysis of differential time-delay systems
The symbolic-numeric study of the robust stability of a differential constant time-delay system with respect to the delay , via the variation of the zero locus of the associated quasipolynomial  in the stability (resp. unstability) region (resp. ) of , initiated in , will be further developed. This problem is another motivation for the development of a fast numerical algorithm for the computation of Netwon-Puiseux series and its implementation in a C library. They will be used to study how the different branches of a quasipolynomial at a critical pair (namely such that ) vary in and in with respect to . See  and the references therein.
Moreover, in collaboration with Mouze (Centrale Lille, France), we want to develop an effective study of the ring , where is the ring of entire functions. The ring plays an important role for differential time-delay systems , . Effective computation of Smith normal forms for matrices with entries in and its implementation in a symbolic-numeric package will have many applications for synthesis problems of differential time-delay systems.
Stabilization problems for functional systems
We want to use the results developed on the module-theoretic aspects of the ring of multivariate rational functions without poles in the closed unit polydisc of to effectively compute stabilizing controllers of multidimensional systems, as well as the Youla-Kuera parametrization of all the stabilizing controllers . This last parametrization can be used to transform standard -optimal control problems, which are nonlinear by nature, into affine, and thus convex optimal problems. See ,  and the references therein. Applications addressed in the former ANR MSDOS (2014–2018) will be developed in collaboration with Bachelier (U. Poitiers). The algorithms obtained in this direction will be unified in a unique Maple package.
Finally, the noncommutative geometric approach to robust problems for infinite-dimensional linear systems (e.g. differential time-delay or PD systems) , initiated in , will be further studied based on the mathematical concepts and methods introduced by (the Fields medalist) Connes . We particularly want to investigate generalizations of Nyquist's theorem to infinite-dimensional systems based on index theory (pairing of -theory and -homology), model reduction based on Connes' interpretation of infinitesimal operators, robustness metrics particularly the -gap metric, etc. The quantized differential calculus , based on Hankel operators, as well as the connections and curvatures on stabilizable systems will be further studied . We aim to exploit these noncommutative differential geometric structures on the systems to get new inside in both the topology and geometry aspects of the -control theory for infinite-dimensional systems .
Parameter estimation for linear & nonlinear functional systems
Linear functional systems
Our expertise on algebraic parameter estimation problem, coming from the former Non-A project-team, will be further developed. Following , this problem consists in estimating a set of parameters of a signal which satisfies a certain dynamics when the signal is observed, where denotes a structured perturbation and a noise. For instance, can be a multi-sinusoidal waveform signal and phases, frequencies, or amplitudes . Based on a combination of algebraic analysis techniques (rings of differential operators), differential elimination theory (computation of annihilators), and operational calculus (Laplace transform, convolution),  shows how can sometimes be explicitly determined by means of closed-form expressions using iterated integrals of . These integrals usually help to filter the effect of the noise on the estimation of the parameters .
A first aim in this direction is to develop to a greater extent our recent work  that shows how the above approach can cover wider classes of signals such as holonomic signals (e.g. signals decomposed into orthogonal polynomial bases, special functions, possibly wavelets).
Moreover,  explains how larger classes of structured perturbations can be considered when the approach developed in , , based on computation of annihilators, is replaced by a new approach based on the more general algebraic concept of syzygies . This general approach to the algebraic parameter estimation problem will be developed. Following the ideas of , an effective version of this general approach will also be done based on differential elimination techniques, i.e. Gröbner basis techniques for rings of differential operators. It will be implemented in a dedicated Maple package which will extend the current prototype NonA package .
Furthermore, as an alternative to passing forth and backwards from the time domain to the operational (Laplace/frequency) domain by means of Laplace transform and its inverse as done in the standard algebraic parameter estimation method , , we aim to develop a direct time domain approach based on calculus on rings of integro-differential operators as described by the following picture ( denotes the Laplace transform):
The direct computation will be handled by means of the effective methods of rings of integro-differential operators described in the above sections.
Nonlinear functional systems
For nonlinear control systems, the approach to the parameter estimation problem, recently proposed in  and based on the computation of integro-differential input-output equations, will be further developed based on the integration of fractions . Such a representation better suits a numerical estimation of the parameters as shown in .
In , we have recently initiated an extension of the results developed in  to handle integro-differential equations such as Volterra-Kostitzin's equation. This general approach, based on an extension of the input-output ideal method for ordinary differential equations to the integro-differential ones, will be further developed based on the effective elimination theory for systems of integro-differential equations. An important advantage of this approach is that not only it solves the identifiability theoretical question but it also prepares a further parameter estimation step .