EN FR
EN FR
##### MASAIE - 2012

Overall Objectives
Scientific Foundations
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
Dissemination
Bibliography

Overall Objectives
Scientific Foundations
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
Dissemination
Bibliography

## Section: Scientific Foundations

### Observers

The concept of observer originates in control theory. This is particularly pertinent for epidemiological systems. To an input-output system, is associated the problem of reconstruction of the state. Indeed for a given system, not all the states are known or measured, this is particularly true for biological systems. This fact is due to a lot of reasons : this is not feasible without destroying the system, this is too expensive, there are no available sensors, measures are too noisy ...The problem of knowledge of the state at present time is then posed. An observer is another system, whose inputs are the inputs and the outputs of the original system and whose output gives an estimation of the state of the original system at present time. Usually the estimation is required to be exponential. In other words an observer, using the signal information of the original system, reconstructs dynamically the state. More precisely, consider an input-output nonlinear system described by

 $\left\{\begin{array}{c}\stackrel{˙}{x}=f\left(x,u\right)\hfill \\ y=h\left(x\right),\hfill \end{array}\right\$ (1)

where $x\left(t\right)\in {ℝ}^{n}$ is the state of the system at time $t$, $u\left(t\right)\in U\subset {ℝ}^{m}$ is the input and $y\left(t\right)\in {ℝ}^{q}$ is the measurable output of the system.

An observer for the the system (1 ) is a dynamical system

 $\stackrel{˙}{\stackrel{^}{x}}\left(t\right)=g\left(\stackrel{^}{x}\left(t\right),y\left(t\right),u\left(t\right)\right),$ (2)

where the map $g$ has to be constructed such that: the solutions $x\left(t\right)$ and $\stackrel{^}{x}\left(t\right)$ of (1 ) and (2 ) satisfy for any initial conditions $x\left(0\right)$ and $\stackrel{^}{x}\left(0\right)$

$\parallel x\left(t\right)-\stackrel{^}{x}\left(t\right)\parallel \le c\phantom{\rule{0.166667em}{0ex}}\parallel x\left(0\right)-\stackrel{^}{x}\left(0\right)\parallel \phantom{\rule{0.166667em}{0ex}}{e}^{-a\phantom{\rule{0.166667em}{0ex}}t}\phantom{\rule{0.277778em}{0ex}},\phantom{\rule{4pt}{0ex}}\forall t>0.$

or at least $\parallel x\left(t\right)-\stackrel{^}{x}\left(t\right)\parallel$ converges to zero as time goes to infinity.

The problem of observers is completely solved for linear time-invariant systems (LTI). This is a difficult problem for nonlinear systems and is currently an active subject of research. The problem of observation and observers (software sensors) is central in nonlinear control theory. Considerable progress has been made in the last decade, especially by the “French school", which has given important contributions (J.P. Gauthier, H. Hammouri, E. Busvelle, M. Fliess, L. Praly, J.L. Gouze, O. Bernard, G. Sallet ) and is still very active in this area. Now the problem is to identify relevant class of systems for which reasonable and computable observers can be designed. The concept of observer has been ignored by the modeler community in epidemiology, immunology and virology. To our knowledge there is only one case of use of an observer in virology ( Velasco-Hernandez J. , Garcia J. and Kirschner D. [38] ) in modeling the chemotherapy of HIV, but this observer, based on classical linear theory, is a local observer and does not allow to deal with the nonlinearities.