Kevin Guelton
Professor of Automatic control - Robotics and mechatronics
Research activities



'Sister and tree, Auvergne landscape' (2000)


Biographical sketches :

Kevin Guelton received the M.S. degree in 2000 and completing a PhD defended in 2003, both in Human and Industrial Automatic Control from the University of Valenciennes. In 2000, he also received the French Diploma of Engineering School (M.Eng.) from the “Ecole d’Ingénieurs en Génie Informatique et Productique”. In 2010, he succeeded the French habilitation for research direction (HDR) degree in automatic control.

From 2003 to 2005, Kevin Guelton served as a temporary assistant professor in Human and Industrial Automatic Control at the “Ecole Nationale Supérieure d'Ingénieurs en Informatique Automatique Mécanique Energétique Electronique” of Valenciennes and was a researcher affiliated with both the Biomechanics team and the Fuzzy Systems team of the “Laboratoire d’Automatique et de Mécanique Industrielles et Humaines”. In 2005, he has been appointed as Associate Professor, then Full Professor since 2019, in automatic control at the University of Reims Champagne-Ardenne and is affiliated with the automatic control team of the "Centre de Recherches en Sciences et Techniques de l'Informaion et de la Communication".


Research path and interests :

Keywords :

Takagi-Sugeno models, Quasi-LPV models, Hybrid systems, Fuzzy and hybrid control theory, LMI stability conditions, conservatism reduction, dynamics, mechanical modeling, biomechanics/biomedical applications.

From biomechanics to automatic control (1999-2004) :

Kevin Guelton began his research work during his engineering then master studies in 1999 at the LAMIH laboratory (Valenciennes, France). He was first involved in biomechanics and his works was focussed on human stance modeling using physical and robotic point of view. Within this subject, his first intend was to refine some previously proposed human stance models in order to better understand some human postural stability disorders. These works have, among other, lead to the proposition of a tridimensional inverted pendulum model of human stance allowing to estimate the spacial excursion of the center of gravity (which is expected to stay in a closed volume when a quiet stance is achieved). See [J1] for more details.

During his PhD at the university of Valenciennes (from 2000 to 2003), Kevin Guelton have followed these works dealing with human standing and was interested in estimating the joint torques deployed to ensure the stance stability. These torques are considered as the biomechanical system's inputs (the control signal beeing provided by the central nervous system). Therefore, analyzing joint torques dynamics may be of interest to highlight human motor control strategies as well as diagnosis some postural disorders. To achieved these torque estimations, the well known inverse dynamics has been first applied from force-plate and video measurements. However, due to slow motion in quiet stance and measurement precision, the latter approach lead to poor results. To overcome these problems in slow motion estimations, the use of the automatic control concept of unknown input observers has been proposed. Due to the nonlinear character of the considered human stance model (based on the dynamic equations of a double inverted pendulum), a Takagi-Sugeno model (also known as quasi-LPV model) in the descriptor form have been derived. Therefore, based on the latter modeling approach, a class of Takagi-Sugeno observers in the descriptor form has been studied and their design has been proposed through the Lyapunov theory and convex optimization to solve Linear Matrix Inequalities (LMI) constraints. For more details on these studies, see Kevin Guelton's PhD thesis (in french) or, e.g. [J4].


LMI based controller design and conservatism reduction for Takagi-Sugeno (Quasi-LPV) and Hybrid systems (main topic since 2005) :

Since 2005, Kevin Guelton as been appointed as associate professor in automatic control at the University of Reims Champagne-Ardenne and he leads his research at the CReSTIC laboratory. His works deals with advanced control of complex dynamical systems and are mainly focussed on modeling, analysis, observer and controller design for the class of affine nonlinear systems. Within these topics, Kevin Guelton is interested in studying convex polytopic systems such as Quasi-LPV models, also known as Takagi-Sugeno models, as well as a class of hybrid dynamical systems such as switched systems.

Since their introduction in 1985, Takagi-Sugeno (TS) models have been the subject of interest of many control researchers. Historically based on the fuzzy formalism, TS models are known as universal approximators of a wide class of nonlinear systems. More precisely, TS are polytopic systems where polytopes may represent linear dynamics (e.g. LTI models) which are then aggregated by nonlinear convex functions (called membership functions). Therefore, using convex transformations such like the sector nonlinearity approach or the tensor product decomposition, a nonlinear system may be exactly matched by a TS model on a compact set of its state space. Thus, a distinctive feature of using TS models in control studies is to extend some of linear control concepts to the case of some nonlinear systems.

Stability analysis (by extension controller and observer design) of TS models is often achieved using the well-known direct Lyapunov method. Thus, from the choice of an appropriate Lyapunov functional candidate, a set of matrix inequalities to be solved (if a solution exists) is obtained. According to the considered control problem, sufficient conditions may be written in terms of Linear Matrix Inequalities (LMI) in the best case or in terms of Bilinear Matrix Inequalities (BMI). Despite LMI, BMI conditions suffer from the drawback of needing sequential optimization process (e.g. Generalized Eigenvalue Problems...) which are initial parameters dependent. Therefore, using BMI optimization, it is often difficult to conclude on the existence of a solution to the convex optimization problem when the current algorithm failed to find a solution. Hence, one of the main issue of many TS model based control studies is to find a 'good' way to obtain relevant LMI conditions.

Moreover, the choice of the Lyapunov functional candidate, the lost of membership functions information in LMI based stability conditions, the matrix transformations and boundary techniques used to obtain LMI conditions, are all constituting sources of conservatism. Indeed, for a given parametrized nonlinear control problem, the whole set of admissible solution may be not entirely recover by the set of feasible solutions of the LMI problems obtained through a Takagi-Sugeno modeling approach. Therefore, one other main issue of many Takagi-Sugeno model based control studies relies in reducing the conservatism (relax) of LMI conditions.

In this context, some of Kevin Guelton's works in the field of Takagi-Sugeno based control problems are concerned with:

- relaxed LMI conditions for some specific control problems (Robust trajectory tracking [J6], TS descriptor robust control [J11], decentralized control...),
- new LMI based robust output feedback controller design using descriptor redundancy approaches [J8],
- Relaxation based on non-quadratic / fuzzy Lyapunov approaches [J9]...

Complementary to its work in the field of TS model based control, Kevin Guelton is also interested in hybrid dynamical systems such like switched systems. Indeed, similarly to TS models, switched systems can be considered as convex polytopic systems where the activation function is not smoothed but discrete. Therefore, in the same way as TS based control problems, applying the direct Lyapunov method to switched systems leads to LMI conditions. In this context, some of Kevin Guelton's recent works focussed on obtaining relaxed LMI conditions for some control problems of switched linear systems [J5] as well as switched nonlinear systems where each mode is represented by a local TS model.

For a review of Kevin Guelton's research works between 2005 and 2010, one can refer to his HDR dissertation (in french).

Application of theoretical results / rehabilitation robotics :

Technical assistance to functional rehabilitation has attracted great interest recently. The design of a new device must start from specific needs expressed by clinicians in rehabilitation. For a specific muscular complex, several rehabilitation techniques may be considered. For muscular deficiency, two rehabilitation techniques can be used: the first is named open muscular chain (OMC), which is characterized by strengthening an isolated muscle group; the second is named closed muscular chain (CMC), which is characterized by recruiting both the agonist and antagonist muscle groups that contribute to the movement. These two techniques are complementary and are used in various stages of rehabilitation protocols. For technical reasons, it is difficult to design a device that is able to reproduce these two techniques simultaneously. OMC devices are commonly used in clinical contexts for lower-limb rehabilitation with isokinetic devices such as Cybex, Biodex or Multi-Iso (designed and developed at the CReSTIC, University of Reims Champagne-Ardenne [J2]).


'Multi-Iso, isokinetic device developed at the CReSTIC' (2006)

More recently, CMC rehabilitation has received particular attention because of its ability to stabilize the targeted joint during exercises similar to those used in daily life (such as walking, sitting or standing up). To cover a larger group of rehabilitation protocols, an isokinetic CMC lower limbs rehabilitation apparatus, called SYS-REEDUC, has been designed [J10] in collaboration with the CRITT MDTS (Charleville-Mézères).



'Prototype of the SYS-REEDUC CMC device realized in collaboration with the CRITT-MDTS' (2008)


Rehabilitation devices used for rehabilitation are controlled by classical control laws. Although the results obtained with these controllers are satisfying in terms of rehabilitation specifications, they are restrictive in terms of control performance, mainly because they do not theoretically guarantee good behavior in the entire state space and do not ensure the rejection of external disturbances such as patient efforts. Thus, a robust nonlinear control structure based on a coupled trajectory generator with a TS trajectory tracking controller (based on a descriptor modeling) has been proposed to ensure the users' safety [J10].

For more details on this project, one can also refer to the PhD dissertation of Lynda Seddiki (in french) supervised by Kevin Guelton.



Projects and grants :


Grants and award :


- Recipient of the French award for scientific excellence 'PES' (2010-2013, ranked A by the national research evaluation committee in 2010).

- PhD grant (2000-2003, 75k€). Founded by the French ministry of Research.


Current projects :


- CPER MOSYP
(2008-2011, 300k€). Hold by Pr A. Hamzaoui, Founded by the "Région Champagne-Ardenne"

- DICOP
(2011-2012, 20k€). Hold by Pr M. Djemaï, LAMIH-UVHC, Founded by the GIS 3SGS.


Past projects :

- CPER SYS-REEDUC (
2005-2007, 421k€). Held by Pr J. Zaytoon, CReSTIC-URCA, Founded by the 'Région Champagne-Ardenne".

- Partnership CReSTIC (Reims) / LMS (Poitiers)
(2007, 9k€). Held by K. Guelton, Founded by the "Région Champagne-Ardenne".

- COSMOS I
(2008, 59k€). Held by Pr J. Ragot, CRAN-INPL, Founded by the GIS 3SGS.

- COSMOS II (2010-2011, 59k€). Held by Pr N. Manamanni, CReSTIC-URCA, Founded by the GIS 3SGS.




Research associates and alumni :


Current PhD students :


- Cuong Duong-Chin (PhD student since June 2010). Toward reliable control and diagnostic of Takagi-Sugeno and Hybrid systems. University of Reims Champagne-Ardenne, France.

Former PhD students :

- Dalel JABRI (PhD defended on December the 1st 2011). Control law synthesis for switched nonlinear interconnected systems based on Takagi-Sugeno modeling. University of Reims Champagne-Ardenne, France / Ecole Nationale d’Ingénieurs de Gabes (ENIG), Tunisia, Automatic Control.

- Dr Tahar Bouarar (PhD defended on December the 8th 2009). LMIs Stability conditions for various classes of Takagi-Sugeno descriptors. University of Reims Champagne-Ardenne, Automatic Control.
Current position: Post doc at CRAN, SURFDIAG Team, University of Nancy 1.

- Dr Lynda SEDDIKI (PhD defended on October the 28th 2008). Design and T-S control of a closed muscular chain lower limbs rehabilitation device. University of Reims Champagne-Ardenne, Automatic Control.
Current position: Associate Prof., University Paris 8, LIASD.


Former MS students :

- Darius Koumba-Emianiwe (Master Degree June 2010). Stability analysis of polynomial Takagi-Sugeno systems via Sum-Of-Squares techniques. University of Reims Champagne-Ardenne, Applied Mathematics.

- Abdelhafidh Jaadari
(Master Degree June 2009). Robust switched stabilization of Takagi-Sugeno systems. University of Reims Champagne-Ardenne, Automatic Control.

- Tahar Bouarar (Master Degree June 2006). LMIs Stability conditions for uncertain Takagi-Sugeno descriptors. University of Reims Champagne-Ardenne, Automatic Control.

- Ludovic Delporte (Master Degree June 2001). A comparative study between radiography and motion capture towards the evaluation of spine mobility. University of Valenciennes, Biomechanics.



Committee memberships and Reviewing activities :

- International Program Committee (IPC)
member of many research events: IEEE ETFA'11 (Track 6), HIS 2011, MISC 2011, CICA 2011, NABIC 2010, SoCPaR 2010, ICINCO 2010, INC-2009, NABIC 2009, SoCPaR 2009, IEEE CICA 2009, IFAC MCBMS’09), IFAC MCBMS’06.

- Referee
for numerous journals: IEEE Transactions on Fuzzy Systems, Fuzzy Sets and Systems, Control Engineering and Practice, Automatica, Int. J. Robust Nonlinear Control, Information Sciences, Int. J. System Sciences, J. Of the Franklin Institute, Control and Intelligent Systems, Physics Letters A, Biomedical Signal Processing and Control, IEEE Transactions on BioMedical Engineering, IEEE Sensor Journal....

- Referee
for numerous conferences: IEEE CDC’08, IEEE FUZZ’08, IFAC WC2008, IFAC AFNC’07, IEEE FUZZ’07, ECC’07, IEEE CDC’06, IEEE CCA’06, IFAC MCBMS’06, CIFA’06, JDMACS’05...

- National Organizing Committee (NOC)
member of research events: WS3SGS'10, JD-JN MACS’07, IFAC MCBMS’06 (Vice chair of the NOC), IFAC CHAOS’06.

- Special Session and Technical Committees
member, Special session organizer, IEEE WCCI’08 / IEEE Fuzz’08, 1-7 June, 2008, Hong-Kong, Chine.

- Member
of the IFAC TC8.2 Technical Committee on Modeling and Control of Biomedical Systems (IFAC TC8.2).

- Member
of the IEEE Technical Committee on Haptics (IEEE TCH).

- Member
of GDR MACS, GDR STIC-Santé and GDR Robotique.