Curriculum vitæ Benoˆıt GAUDIN Born the 4th May 1977 in Saint Malo (France) Professional Address Competence Center: MOTION FOKUS Fraunhofer Kaiserin Augusta Allee 31 10589 Berlin Tel : (+49) 30 34 63 72 42 Fax : (+49) 30 34 63 80 00 E-mail :
[email protected] www : http://benoit.gaudin1.free.fr
1
Personal Address Iranische Strasse 6, 22103 13347 Berlin Tel : (+49) 30 43 73 86 61
Professional Currently: Since the 1st of September 2005, I hold an Ercim fellowship at the FOKUS lab, in Berlin (Germany). I make part of the MOTION competence center, and I work on designing control theory to be applied to supervision of components of autonomic systems/networks. Oct2004-Aug2005 One year research and teaching position (ATER) at the university Rennes I. Concerning research, I made part of the Vertecs team (Verification, testing, control) of the IRISA computer sciences lab, in Rennes (France). Oct2001-Sept2004 Ph.D student at the university Rennes I. Concerning research, I made part of the Vertecs team (Verification, testing, control) of the IRISA computer sciences lab, in Rennes.
2
Education 2001-2004 Ph.D in Computer Sciences at the university Rennes I (France). This Ph.D dealt with system validation and more especially with supervisory control of concurrent discrete event systems. 2001 Pre-doctorate diploma in computer sciences at the university Rennes I (France). 2001 Magist`ere in mathematics and computer sciences at university Rennes I (France). 2000 Master in mathematics at the university Rennes I (France). 1999 Bachelor in mathematics at the university Rennes I (France).
3
Research topics
Key-words : Validation, test, verification, supervisory control, autonomic systems, discrete event systems, composition, hierarchical systems.
Systems are now so complex than they are hard to analyze. They are most of time composed of several subsystems which interact with each others. The behaviors of the global system then depend on the ones of the subsystems as well as on their interaction. Because of this complexity, avoiding undesired behaviors is difficult, even during the design process. Therefore, it becomes of great interest to be able to constrain systems in order that some good behaviors are ensured on them. Undesired behaviors can not only be introduced while implementing the system, but also while designing the system. Specification described from several specifications of subsystems are quite complex. This complexity entails that bad behaviors are unfortunately introduced in the specification itself (enabling deadlock states from being reachable, disabling some desired synchronizations, enabling reachability of states from which integrity of some shared resources is not ensured,...). Supervisory control is a validation technique on system specifications or systems themselves. Acting at the modeling phase, supervisory control does more than verification. Indeed, given a specification and a desired property, verification (and in particular Model-Checking) aims at giving a verdict such as: the property holds on the specification, or the properties does not hold on it. If the answer is NO, then a new specification has to be designed. The verifying algorithm provides most of time some informations about why the property does not hold (or a counter-example). However, the modifications have to be made by hand, what facilitate the input of new errors. The purpose of supervisory control is precisely to provide an automatic way to perform modification on the specification in order to ensure a property. To do this, the behaviors of the new specification are assumed to be some of the initial specification. Therefore, supervisory control consists in transforming specifications by restraining their behaviors. This restriction entails that some specified events must be prevented from occurring. However, preventing certain events from occurring is not always relevant. For example, it is not the case for events modeling failure, timer expiration, someone who press a button, etc... This kind of events are said to be uncontrollable and input difficulties in supervisory control. Supervisory control can also be seen as a way to automatically compute specification of supervisors for a given system. In that case, the systems under consideration are assumed to be implemented and the design phase can no more be modified. Such systems may possess undesired behaviors which come from specifications. It means that the system acts as described in its specification (this can be validated using testing methods for example). In this case, the purpose of supervisory control is to compute the model of a system called supervisor (and which could also be validated using testing methods). The supervisor has then to interact with the initial system so that it ensures a given property. In that case, the supervisor prevents some events from occurring by disabling some synchronizations with the system. Supervisory control was first introduced by people from automatic community. System were then specified with Finite State Machines (FSM). Some operators over FSM exist to model synchronizations and parallel execution of subsystems. Given a set of FSM modeling subsystems which have to run in parallel, the computation of one unique FSM modeling the global system is needed to perform the computation of a supervisor. Unfortunately this computation is often not feasible in practice because it requires too much time and space. Finding methods in order to tackle this problem is then challenging. To that purpose, one approach consists in optimizing the data structure which encodes the FSM and the algorithms which perform supervisory control. For example, Binary Decision Diagrams (BDD) can be used to that goal. An other approach consists in finding methods to apply supervisory control, avoiding the computation of a FSM modeling the whole system. In my Ph.D, I was interested in this approach. In particular, I considered the classical Basic Supervisory Control Problem (BSCP), as well as the State Avoidance Control Problem (SACP). BSCP consists in computing a supervisor which restrains the behaviors of the system to the ones modeled by a FSM. SACP consists in computing a supervisor which prevents a given set of states of the system from being reachable. Since September, in the FOKUS lab (Berlin, Germany), I have worked to extend and apply supervisory control to a new kind of systems called Autonomic Systems. Autonomic systems are composed of distributed components which are able to self-adapt, autonomously self-organized, and control themselves together. The goal of autonomic systems is to deal with most of problems in order to provide by itself satisfactory services to the user.
Supervisory control seems to be an interesting formal tool to help autonomic system to reach their goals. In order to be closer to system design models, I am currently extending my previous works to a more powerfull model than FSM: Extended Finite State Machines (EFSM). In particular, composition of EFSM allows to model set of process running in parallel, each possessing a finite set of variables taking values in a possibly infinite domain, which can synchronized together and exchange informations when synchronizing.
4
Teaching • Practical works in initiation to functional programming (with Scheme) for first year students in MIPE 1,IFSIC, University of Rennes I. • Practical works in initiation to imperative programming (with Java) for second year students in MIPE 1,IFSIC, University of Rennes I. • Directed and practical works in operating systems for third year students in IUP, IFSIC, University of Rennes I. • Directed and practical works in database for Master degree students (M2CCI), IFSIC, University of Rennes I. • practical works in Unix commands for Master degree students (M2CCI), IFSIC, University of Rennes I. • Practical works in protocol testing and validation for Master degree students (M2IR), IFSIC, University of Rennes I.
5
Publications
Ph.D thesis [1] B. Gaudin. Synth`ese de contrˆ oleurs sur des syst`emes a ` ´ev´enement discrets structur´es. University Rennes I, November 2004. Academic Revues [1] B. Gaudin, H. Marchand. Modular Supervisory Control problems of asynchronous and Hierarchical Finite State Machines. European Journal of Control Vol. 10(2), EJC 2004. International Conferences [1] B. Gaudin, H Marchand. Supervisory Control and Deadlock Avoidance Control Problem for Concurrent Discrete Event Systems. 44nd IEEE Conference on Decision and Control and Control (CDC’05) and European Control Conference ECC 2005, Sevilla (Spain), December 2005. [2] J. Komenda, J. H. van Schuppen, B. Gaudin, H. Marchand, Modular supervisory control with general indecomposable specification languages. 44nd IEEE Conference on Decision and Control and Control (CDC’05) and European Control Conference ECC 2005, Sevilla (Spain), December 2005. [3] B. Gaudin, H Marchand. Efficient Computation of supervisors for loosely synchronous Discrete Event Systems: A State-Based Approach. 16th IFAC world Congress, IFAC’05, Pragues, 4-8 July 2005. [4] B. Gaudin, H Marchand. Safety Control of Hierarchical Synchronous Discrete Event Systems: A State-Based Approach. Mediterranean Control Conference, Med’05, Cyprus, 27-29 June 2005. [5] B. Gaudin, H. Marchand. Modular Supervisory Control of a class of Concurrent Discrete Event Systems. Workshop On Discrete Event Systems, WODES’04, Reims, France. September 2004
[6] B. Gaudin, H Marchand. Modular Supervisory Control of Asynchronous and Hierarchical Finite State Machines. European Control Conference, ECC’03, Cambridge, UK. September 2003 [7] H. Marchand, B. Gaudin. Supervisory Control Problems of Hierarchical Finite State Machines. 41th IEEE Conference on Decision and Control, CDC’02, Las Vegas, USA. December 2002 National Conferences [1] B. Gaudin, H. Marchand. Une approche modulaire pour le contrˆ ole de syst`emes a ` ´ev´enements disi` e me crets concurrents 5 Colloque Francophone sur la Mod´elisation des Syst`emes R´eactifs, MSR’05, Grenoble, France, October 2005 [2] B. Gaudin, H. Marchand. Contrˆ ole de syst`emes a ` ´ev´enements discrets hi´erarchiques, 4 i`eme Colloque Francophone sur la Mod´elisation des Syst`emes R´eactifs, MSR’03, Metz, France, October 2003 Technical reports [1] B. Gaudin, H Marchand. Supervisory Control of Concurrent Discrete Event Systems, IRISA technical report, No 1593. February 2004 [2] B. Gaudin, H Marchand. Supervisory Control of Structured Discrete Event Systems, IRISA technical report, No 1569. November 2003
