Toward understandable co-simulations in model driven engineering

Size: px

Start display at page:

Download "Toward understandable co-simulations in model driven engineering"

Buck Floyd
5 years ago
Views:

1 Toward understandable co-simulations in model driven engineering University of Nice, I3S CNRS INRIA Aoste

2 Embedded cyber-physical systems Application Concurrent application State machine region Data flow process network... Heterogeneous Control flow Data flow Time driven Event driven 2

3 Embedded cyber-physical systems Application 10ms Concurrent Heterogeneous Constraints < 3ms safety-critical hard real-time Extra functional low power Cost Cyber-Physical related 3

4 Embedded cyber-physical systems Application 10ms Concurrent Heterogeneous Constraints safety-critical hard real-time Extra functional low power Cost Cyber-Physical related 4

5 Embedded cyber-physical systems Application 10ms Concurrent Heterogeneous Constraints < 3ms safety-critical hard real-time Extra functional low power Cost Cyber-Physical related Need for V&V 5

6 Embedded cyber-physical systems Application Execution Platform Parallel execution Heterogeneous Distributed... 6

7 Embedded cyber-physical systems Application deployment Execution Platform 7

8 Embedded cyber-physical systems Application deployment Execution Platform Constraints safety-critical hard real-time Extra functional Cyber-Physical related 8

9 Embedded cyber-physical systems Application deployment Different languages are used to address these different concerns Execution Platform Constraints safety-critical hard real-time Extra functional Cyber-Physical related 9

10 Embedded cyber-physical systems Application deployment Different languages are used to address these different concerns Execution Platform Need for V&V Constraints safety-critical hard real-time Extra functional of the global system despite the use of different languages Cyber-Physical related system ok? 10

11 Embedded cyber-physical systems Application deployment Different languages are used to address these different concerns Execution Platform Need for V&V Constraints safety-critical hard real-time Extra functional of the global system despite the use of different languages...all along the development cycle Cyber-Physical related system ok? 11

12 How to deal with that? Formal models MDE 12

13 Model-Driven Engineering MDE Pros Design in the large or in the small Domain Specific Languages (For a specific goal, the adequate concepts are reified) Multi-view modeling Formal definition of concepts and relations 13

14 Model-Driven Engineering MDE Pros Design in the large or in the small Domain Specific Languages (For a specific goal, the adequate concepts are reified) Multi-view modeling Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic reasoning 14

15 Model-Driven Engineering MDE Pros Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic reasoning Cons Discrepancies in interpretations 15

16 Model-Driven Engineering MDE Pros Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic reasoning Cons Discrepancies in interpretations A B 16

17 Model-Driven Engineering MDE Pros Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic reasoning Cons Discrepancies in interpretations Producer Consumer 17

18 Model-Driven Engineering MDE Pros Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic reasoning Cons Discrepancies in interpretations Client Server 18

19 Model-Driven Engineering MDE Pros Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic reasoning Cons Discrepancies in interpretations Lack of precise explicit behavioral semantics Model 1 Behaviorally Behaviorally Equivalent Equivalent?????? Refines Refines?????? Model 2 19

Model-Driven Engineering MDE Pros Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic

20 Model-Driven Engineering MDE Pros Formal definition of concepts and relations Tools take benefits of this: automatic generation of advanced editors, static checkers, automatic reasoning Cons Discrepancies in interpretations Lack of precise behavioral semantics Tools suffer from this No automatic generation of simulators, debuggers, animators, explorer, etc. 20

21 Formal models Models Models of of Computation Computation and and Communications Communications (MoCCs) (MoCCs) 21

22 Formal models Models Models of of Concurrency Concurrency and and Communications Communications (MoCCs) (MoCCs) 22

23 Formal Models (MoCC) Formal models Pros Mathematical semantics (i.e. no ambiguity) Tools benefit from that! 23

24 Formal Models (MoCC) Formal models Pros Mathematical semantics (i.e. no ambiguity) Tools benefit from that! Powerful analysis and algorithmic methods Optimization / verification Guaranteed equivalence between code and model Basis for well-founded transformations Model 1 equivalence equivalence refinement refinement Model 2 24

25 Formal Models (MoCC) Formal models Pros Mathematical semantics (i.e. no ambiguity) Tools benefit from that! Powerful analysis and algorithmic methods, Optimization / verification, Guaranteed equivalence between code and model, Basis for well-founded transformations Cons Distance from current mainstream engineering practice Exotic formalisms for designers Possibly far from a designer domain Not necessarily adapted for all the models in a system 25

26 Where are we going? Model Driven Engineering Formal models Tools, standards, engineering world effective environment for executable DSMLs Verification / validation effective mathematical mean to address a problem 26

27 So? Formal models MDE Transformations are often used from a world to another one... 27

28 So? MDE Transformations can distort models due to the abstraction gap between a DSL and a formal model Difficulty to understand the resulting behavioral semantics given by both the transformation and the semantics of the targeted language No direct way to play with the behavioral semantics 28

29 The finality? Formal MDE 29

30 The finality! Formal MDE 30

31 Formal Models (MoCC) Formal models Synchronous Data Flow Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 22 initial initial 'data' 'data' 31

32 Formal Models (MoCC) Formal models Synchronous Data Flow Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 2 executions of A causes 1 of B 22 initial initial 'data' 'data' 32

33 Formal Models (MoCC) Formal models Synchronous Data Flow Logical Time Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 22 initial initial 'data' 'data' 2 executions of A causes 1 of B A B 33

34 Formal Models (MoCC) Formal models Synchronous Data Flow Logical Time Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 22 initial initial 'data' 'data' 2 executions of A causes 1 of B schedules schedules A B 34

35 Formal Models (MoCC) Formal models Synchronous Data Flow Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 22 initial initial 'data' 'data' Logical Time Schedule of A Schedule of B Schedule of C 35

36 Formal Models (MoCC) Formal models Synchronous Data Flow Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 22 initial initial 'data' 'data' Logical Time Schedule of A 5 Schedule ofofab 4 Schedule Schedule ofofab 3 Schedule 2 Schedule ofofaof Schedule C Schedule B Schedule ofofaof 1 Schedule C Schedule B Schedule of C Schedule ScheduleofofBC Schedule of C 36

37 Formal Models (MoCC) Formal models Synchronous Data Flow Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 22 initial initial 'data' 'data' Logical Time Schedule of A 5 Schedule ofofab 4 Schedule Schedule ofofab 3 Schedule 2 Schedule ofofaof Schedule C Schedule B Schedule ofofaof 1 Schedule C Schedule B Schedule of C Schedule ScheduleofofBC Schedule of C Analysis Analysis 37

38 Formal Models (MoCC) Synchronous Data Flow Execution Execution of ofaa provide provide 11 'data' 'data' Execution Execution of of AA require require 22 'data' 'data' 22 initial initial 'data' 'data' Formal models Logical Time Schedule of A 5 Schedule ofofab 4 Schedule Schedule ofofab 3 Schedule 2 Schedule ofofaof Schedule C Schedule B Schedule ofofaof 1 Schedule C Schedule B Schedule of C Schedule ScheduleofofBC Schedule of C Analysis Analysis Still... (they are languages where) the semantics is either (mathematically) given on paper or encoded in tools and algorithms... 38

39 Formal Models (MoCC) Synchronous Data Flow Execution Execution of ofaa provide provide 11 'data' 'data' 22 initial initial 'data' 'data' Formal models Logical Time Schedule of A 5 Schedule ofofab 4 Execution Schedule Execution of of AA Schedule ofofab 3 Schedule require 2 'data' require 2 'data' 2 Schedule ofofaof Schedule C Schedule B Schedule ofofaof 1 Schedule C Schedule B So what is the essence of Schedule of C Schedule ScheduleofofBC what they wanted to Schedule specifyof?c Analysis Analysis Still... these are languages where the semantics is either (mathematically) given on paper or encoded in tools and algorithms... 39

40 Design Time Formal MDE Expliciting the MoCC within an MDE environment Adding explicit activation conditions Adding relations between these activation conditions formal semantics explicit within the model (not hidden in the simulator / any transformation) 40

41 Design Time Formal MDE Expliciting the MoCC within an MDE environment Adding explicit activation conditions Adding relations between these activation conditions formal semantics explicit within the model (not hidden in a simulator or a transformation) The MARTE time model and CCSL were a first step toward that 41

42 Design Time Formal MDE MDE is good for the structural concern Logical Time is good for the dynamic concern Adding explicit activation conditions Adding relations between these activation conditions 42

43 Design Time Formal MDE MDE is good for the structural concern Logical Time is good for the dynamic concern Adding explicit activation conditions Adding relations between these activation conditions User User Model Model MDE 43

44 Design Time Formal MDE MDE is good for the structural concern Logical Time MoCC is good for the dynamic concern Adding explicit activation conditions Adding relations between these activation conditions User User Model Model MDE MoCC 44

45 Design Time Formal MDE MDE is good for the structural concern Logical Time is good for the dynamic concern Adding explicit activation conditions Adding relations between these activation conditions User User Model Model Events Events MDE MDE MoCC 45

Time @ Design Time Formal MDE MDE is good for the structural concern Logical Time is good for the dynamic concern Adding explicit activation

46 Design Time Formal MDE MDE is good for the structural concern Logical Time is good for the dynamic concern Adding explicit activation conditions Adding relations between these activation conditions User User Model Model Events Events MDE MDE Constraints Constraints MoCC Formal MDE 46

47 Design Time Formal MDE Sketchy example of its use User model Producer c1 Consumer 47

48 Design Time Formal MDE Sketchy example of its use User model Producer c1 sendevent Consumer receiveevent 48

49 Design Time Formal MDE Sketchy example of its use User model Producer c1 on Send : Event Consumer on Receive : Event CCSL model 49

50 Design Time Formal MDE Sketchy example of its use User model Producer c1 on Consumer on c1.sendevent Event c1.receiveevent: Event CCSL model left right R1 : Precedes 53

51 Design Time Formal MDE Sketchy example of its use User model Producer c1 on Consumer on c1.sendevent Event c1.receiveevent: Event CCSL model this is a symbolic left representation of all the acceptable schedules for the user model right R1 : Precedes 54

52 Implementation in TimeSquare 55

53 Implementation in TimeSquare 56

54 Implementation in TimeSquare, Frédéric Mallet, Charles André. On the Formal Execution of UML and DSL Models. [Short paper]. WIP of the 4th International School on Model-Driven Development for Distributed, Realtime, Embedded Systems, Apr 2009, Aussois, France Frédéric Mallet, Charles André,. Executing AADL models with UML/Marte. Int. Conf. Engineering of Complex Computer Systems - ICECCS'09, Jun 2009, Potsdam, Germany. pp < Frédéric Mallet,, Charles André, Robert De Simone. The Clock Constraint Specification Language for building timed causality models. Innovations in Systems and Software Engineering, Springer London, 2010, 6 (1-2), pp < 57

55 Implementation in TimeSquare Marie-Agnès Peraldi-Frati,. Scheduling Multi Clock Real Time Systems: From Requirements to Implementation. International Symposium on Object/Component/Service-oriented Real-time Distributed Computing, Mar 2011, Newport Beach, United States. IEEE computer society, IEEE international Symposium on Object/Component/service Oriented Real-Time Distributed Computing, pp. 50; 57. < Frédéric Mallet, Charles André. On the Formal Execution of UML and DSL Models. [Short paper]. WIP of the 4th International School on Model-Driven Development for Distributed, Realtime, Embedded Systems, Apr 2009, Aussois, France Frédéric Mallet, Charles André,. Executing AADL models with UML/Marte. Int. Conf. Engineering of Complex Computer Systems - ICECCS'09, Jun 2009, Potsdam, Germany. pp < Frédéric Mallet,, Charles André, Robert De Simone. The Clock Constraint Specification Language for building timed causality models. Innovations in Systems and Software Engineering, Springer London, 2010, 6 (1-2), pp < 58

56 Implementation in TimeSquare Calin Glitia,, Frédéric Mallet, Jean-Vivien Millo, Pierre Boulet, Abdoulaye Gamatié. Progressive and explicit refinement of scheduling for multidimensional data-flow applications using uml marte. Design Automation for Embedded Systems, Springer Science+Business Media, LLC 2012, 2012, 16 (2), pp < < Calin Glitia,, Frédéric Mallet. Logical Work: Capturing Data Dependencies and Platform Constraints. Kaźmierski, Tom J. J. and Morawiec, Adam. System Specification and Design Languages, 106, Springer New York, pp , 2012, Lecture Notes in Electrical Engineering, < Marie-Agnès Peraldi-Frati,. Scheduling Multi Clock Real Time Systems: From Requirements to Implementation. International Symposium on Object/Component/Service-oriented Real-time Distributed Computing, Mar 2011, Newport Beach, United States. IEEE computer society, IEEE international Symposium on Object/Component/service Oriented Real-Time Distributed Computing, pp. 50; 57. < Frédéric Mallet, Charles André. On the Formal Execution of UML and DSL Models. [Short paper]. WIP of the 4th International School on Model-Driven Development for Distributed, Realtime, Embedded Systems, Apr 2009, Aussois, France Frédéric Mallet, Charles André,. Executing AADL models with UML/Marte. Int. Conf. Engineering of Complex Computer Systems - ICECCS'09, Jun 2009, Potsdam, Germany. pp < Frédéric Mallet,, Charles André, Robert De Simone. The Clock Constraint Specification Language for building timed causality models. Innovations in Systems and Software Engineering, Springer London, 2010, 6 (1-2), pp < 59

57 Implementation in TimeSquare Kelly Garcés,, Frédéric Mallet. A Model-Based Approach for Reconciliation of Polychronous Execution Traces. SEAA th EUROMICRO Conference on Software Engineering and Advanced Applications, Aug 2011, Oulu, Finland. IEEE. < Calin Glitia,, Frédéric Mallet, Jean-Vivien Millo, Pierre Boulet, Abdoulaye Gamatié. Progressive and explicit refinement of scheduling for multidimensional data-flow applications using uml marte. Design Automation for Embedded Systems, Springer Science+Business Media, LLC 2012, 2012, 16 (2), pp < < Calin Glitia,, Frédéric Mallet. Logical Work: Capturing Data Dependencies and Platform Constraints. Kaźmierski, Tom J. J. and Morawiec, Adam. System Specification and Design Languages, 106, Springer New York, pp , 2012, Lecture Notes in Electrical Engineering, < Marie-Agnès Peraldi-Frati,. Scheduling Multi Clock Real Time Systems: From Requirements to Implementation. International Symposium on Object/Component/Service-oriented Real-time Distributed Computing, Mar 2011, Newport Beach, United States. IEEE computer society, IEEE international Symposium on Object/Component/service Oriented Real-Time Distributed Computing, pp. 50; 57. < Frédéric Mallet, Charles André. On the Formal Execution of UML and DSL Models. [Short paper]. WIP of the 4th International School on Model-Driven Development for Distributed, Realtime, Embedded Systems, Apr 2009, Aussois, France Frédéric Mallet, Charles André,. Executing AADL models with UML/Marte. Int. Conf. Engineering of Complex Computer Systems - ICECCS'09, Jun 2009, Potsdam, Germany. pp < Frédéric Mallet,, Charles André, Robert De Simone. The Clock Constraint Specification Language for building timed causality models. Innovations in Systems and Software Engineering, Springer London, 2010, 6 (1-2), pp < 60

58 Implementation in TimeSquare Kelly Garcés,, Frédéric Mallet. A Model-Based Approach for Reconciliation of Polychronous Execution Traces. SEAA th EUROMICRO Conference on Software Engineering and Advanced Applications, Aug 2011, Oulu, Finland. IEEE. < Calin Glitia,, Frédéric Mallet, Jean-Vivien Millo, Pierre Boulet, Abdoulaye Gamatié. Progressive and explicit refinement of scheduling for multidimensional data-flow applications using uml marte. Design Automation for Embedded Systems, Springer Science+Business Media, LLC 2012, 2012, 16 (2), pp < < Calin Glitia,, Frédéric Mallet. Logical Work: Capturing Data Dependencies and Platform Constraints. Kaźmierski, Tom J. J. and Morawiec, Adam. System Specification and Design Languages, 106, Springer New York, pp , 2012, Lecture Notes in Electrical Engineering, < We were able to reason on the scheduling state space of a model Marie-Agnès Peraldi-Frati,. Scheduling Multi Clock Real Time Systems: From Requirements to Implementation. International Symposium on Object/Component/Service-oriented Real-time Distributed Computing, 2011, Newport Beach, United States. IEEE computer society, IEEE at different places inmarthe development cycle international Symposium on Object/Component/service Oriented Real-Time Distributed Computing, pp. 50; 57. < Frédéric Mallet, Charles André. On the Formal Execution of UML and DSL Models. [Short paper]. WIP of the 4th International School on Model-Driven Development for Distributed, Realtime, Embedded Systems, Apr 2009, Aussois, France Frédéric Mallet, Charles André,. Executing AADL models with UML/Marte. Int. Conf. Engineering of Complex Computer Systems - ICECCS'09, Jun 2009, Potsdam, Germany. pp < Frédéric Mallet,, Charles André, Robert De Simone. The Clock Constraint Specification Language for building timed causality models. Innovations in Systems and Software Engineering, Springer London, 2010, 6 (1-2), pp < 61

59 TimeSquare in one slide Formal MDE Consumer c1 on on c1.sendevent Event c1.receiveevent: Event CCSL model left right R1 : Precedes TimeSquare TimeSquare User model Producer Simulation Model animation Timing Diagram Code execution State space exploration maxdiff: Integer, Frédéric Mallet. TimeSquare: Treat your Models with Logical Time. Carlo A. Furia, Sebastian Nanz. TOOLS - 50th International Conference on Objects, Models, Components, Patterns , May 2012, Prague, Czech Republic. Springer, Objects, Models, Components, Patterns, 7304, pp , Lecture Notes in Computer Science - LNCS. < 62

60 TimeSquare in one slide Formal MDE Consumer c1 on on c1.sendevent Event c1.receiveevent: Event CCSL model left right R1 : Precedes TimeSquare TimeSquare User model Producer Simulation Model animation Timing Diagram Code execution State space exploration maxdiff: Integer Agnostic of the kind of user model Use the CCSL specification as a configuration of the model simulator Amenable to some V&V activities without knowledge of the user model syntax, Frédéric Mallet. TimeSquare: Treat your Models with Logical Time. Carlo A. Furia, Sebastian Nanz. TOOLS - 50th International Conference on Objects, Models, Components, Patterns , May 2012, Prague, Czech Republic. Springer, Objects, Models, Components, Patterns, 7304, pp , Lecture Notes in Computer Science - LNCS. < 63

61 Design Time Good but: The CCSL specification is not a MoCC since it depends on a specific model how can we provide a meta-language so that the MoCC can be defined on the syntax of a language. 64

62 Design Time Good but: The CCSL specification is not a MoCC since it depends on a specific model how can we provide a meta-language so that the MoCC can be defined on the syntax of a language. 65

63 Reifying language behavioral semantics Executable language Language Syntactic Domain mapping Semantic Domain 66

64 Reifying language behavioral semantics Executable language Language Syntactic Domain mapping Semantic Domain Executable meta-modeling Providing languages to define the language elements, i.e. providing some meta languages 67

65 Reifying language behavioral semantics Executable language Grammar languages (BNF), Markup Languages (DTD for SGML/XML), Object Oriented grammar languages (Meta Object Facilities),... Language Syntactic Domain mapping Semantic Domain definedby Executable meta-modeling Providing languages to define the language elements, i.e. providing some meta-languages 68

66 Reifying language behavioral semantics Executable language Grammar languages (BNF), Markup Languages (DTD for SGML/XML), Object Oriented grammar languages (Meta Object Facilities),... Language Syntactic Domain mapping Semantic Domain Transformation languages (XSLT, QVT, etc), Interpreter language (Kermeta, java EMF, etc),... definedby definedby Executable meta-modeling Providing languages to define the language elements, i.e. providing some meta-languages 69

67 Reifying language behavioral semantics Executable language Grammar languages (BNF), Markup Languages (DTD for SGML/XML), Object Oriented grammar languages (Meta Object Facilities),... Language Syntactic Domain mapping Semantic Domain Transformation languages (XSLT, QVT, etc), Interpreter language (Kermeta, java EMF, etc),... definedby definedby Executable meta-modeling Providing languages to define the language elements, i.e. providing some meta-languages How can the behavioral semantics be manipulated? we miss a concrete formal entity to represent it. 70

.. Language Syntactic Domain mapping definedby definedby Semantic Domain We need meta-languages for the behavioral semantic

68 Reifying language behavioral semantics Executable language Grammar languages (BNF), Markup Languages (DTD for SGML/XML), Object Oriented grammar languages (Meta Object Facilities),... Language Syntactic Domain mapping definedby definedby Semantic Domain We need meta-languages for the behavioral semantic specification Executable meta-modeling Providing languages to define the language elements, i.e. providing some meta-languages How can the behavioral semantics be manipulated? we miss a concrete formal entity to represent it. 71

69 Reifying language behavioral semantics Executable language Grammar languages (BNF), Markup Languages (DTD for SGML/XML), Object Oriented grammar languages (Meta Object Facilities),... Language Syntactic Domain mapping Semantic Domain We need meta-languages for the behavioral semantic specification definedby definedby This is the first question we addressed in the GEMOC ANR project 72

70 look at a simple example 73

71 look at a simple example Data management Structuring Computing 74

72 look at a simple example Data management Structuring Computing Conditional control If (boolean condition) 75

73 look at a simple example Data management Structuring Computing Conditional control If (boolean condition) When Internal control (tokens) External control (events) 76

74 look at a simple example Data management Structuring Computing Language Theory Conditional control If (boolean condition) When Internal control External control Concurrency Theory 77

75 from what ingredients the semantic domain is made up with? sequential semantics specification Operational Axiomatic Translational Specifies evolution of the model state Language Theory Concurrency specification Specifies (possibly timed) Concurrency Tagged signal causal relations and Theory Event structure synchronizations Model of Computation 78

76 proposition Providing meta-languages for the semantic specification Language Mapping Syntactic Domain Model State specification and evolution Semantic Domain Model of Concurrency, causalities and synchronizations 79

77 Meta-languages for executable modeling DSML < Abstract Syntax, semantics > Specify the grammar of the language 80

78 Meta-languages for executable modeling DSML < AS, Domain Specific Actions,, > currentstate fire() evaluate() Specify the execution state and its evolution 82

79 Meta-languages for executable modeling DSML < AS, DSA,, MoCC> Defines the possible schedules of the previously defined actions 84

80 Meta-languages for executable modeling DSML < AS, DSA,, MoCC> Defines the possible schedules of the previously defined actions in a symbolic way 85

81 Meta-languages for executable modeling DSML < AS, DSA, Domain Specific Events, Domain MoCC > Specific Events Realizes a mapping between the event from the MoCC and the action calls and returns 87

82 Meta-languages for executable modeling DSML < AS, DSA, Domain Specific Events, Domain MoCC > Specific Events Event Constraint Language from inria Aoste team ( 88

83 Meta-languages for executable modeling Reifying Concurrency for Executable Metamodeling (Benoit Combemale,, Matias Vara Larsen, Frédéric Mallet, Olivier Barais, Benoit Baudry, Robert France), In 6th International Conference on Software Language Engineering (SLE 2013) (Richard F. Paige Martin Erwig, Eric van Wyk, eds.), Springer-Verlag, [bibtex] [pdf] 89

84 The GEMOC Studio Language Workbench Automatic generation Modeling Workbench 93

85 Modeling workbench: 95

86 Modeling workbench: A generic execution engine Only configured by the semantics! 96

87 Modeling workbench: A generic scheduling state space explorer Only configured by the semantics! 97

Modeling workbench: http://gemoc.org/studio/ A generic scheduling state space explorer, Papa Issa Diallo, Ciprian Teodorov, Joël Champeau, Benoit Combemale. a Meta-Language the Concurrency Concern in!

88 Modeling workbench: A generic scheduling state space explorer, Papa Issa Diallo, Ciprian Teodorov, Joël Champeau, Benoit Combemale. a Meta-Language the Concurrency Concern in! OnlyTowards configured byforthe semantics DSLs. Design, Automation and Test in Europe Conference and Exhibition (DATE), Mar 2015, Grenoble, France. IEEE proceedings of Design, Automation and Test in Europe Conference and Exhibition (DATE), Amani Khecharem, Carlos Gomez,, Frédéric Mallet, Robert de Simone. Execution of Heterogeneous Models for Thermal Analysis with a Multi-view Approach. FDL 2014 : Forum on specification and Design Languages, Oct 2014, Munich, Germany. IEEE Arda Goknil,, Marie-Agnès Peraldi-Frati, Frédéric Mallet. Tool Support for the Analysis of TADL2 Timing Constraints using TimeSquare. ICECCS' th International Conference on Engineering of Complex Computer Systems, Jul 2013, Singapore. 98

89 We were able to define the behavioral semantics of a single language Good but... The previous definition only address a single language... What about system modeling and heterogeneous execution? Globalizing Modeling Languages. Benoit Combemale,, Benoit Baudry, Robert France, Jean-Marc Jézéquel, Jeff Gray. Computer, IEEE, 2014, pp

91 Mechanical Structure Airlines HumanMachine Interaction Avionics Aerodynamics Several Several concerns concerns for for aa single single system system Propulsion System Communications Environmental Impact Safety Regulations Authorisations Navigation

92 Mechanical Structure Airlines HumanMachine Interaction Avionics Aerodynamics At At least least one one expert expert for for each each concern concern Propulsion System Communications Environmental Impact Safety Regulations Authorisations Navigation

93 At At least least one one model model for for each each concern concern

94 At At least least one one model model for for each each concern concern

95 At At least least one one model model for for each each concern concern

96 At At least least one one model model for for each each concern concern Heterogenous because they have different Model of Computation

97 Carlos Ernesto Cardena Gomez Multi-View Approaches IEEE (updated version of IEEE-1401) View 1 The System is composed of different views Correspondences The view are expressed in domain specific languages Each domain specific language have its own behavioral semantics (and usually its own tooling) View 3 View 2 System [M. Nassar, VUML: a viewpoint oriented UML extension, 2003] 107

98 Carlos Ernesto Cardena Gomez Multi-View Approaches IEEE (updated version of IEEE-1401) View 1 The System is composed of different views The view are expressed in domain specific languages Correspondences Each domain specific language have its own behavioral semantics (and usually its own tooling) View 2 View 3 System 108

99 Carlos Ernesto Cardena Gomez Multi-View Approaches IEEE (updated version of IEEE-1401) View 1 The System is composed of different views The view are expressed in domain specific languages Correspondences Each domain specific language have its own behavioral semantics (and usually its own tooling) View 2 View 3 System There is a need to coordinate/orchestrate all these model executions to understand the global behavior of the system 109

The The System is composed of different views View 1 The

Correspondences Each domain specific language have its

Modeling (and usually its own tooling) Workbench

coordinate/orchestrate all these model executions to

100 Carlos Ernesto Cardena Gomez Multi-View Approaches IEEE (updated version of IEEE-1401) GEMOC Studio The The System is composed of different views View 1 The view are expressed in domain specific languages Correspondences Each domain specific language have its own behavioral semantics Automatic generation Language Modeling (and usually its own tooling) Workbench Workbench 93 View 2 View 3 System There is a need to coordinate/orchestrate all these model executions to understand the global behavior of the system By taking advantage of explicit behavioral semantics 110

101 model behavioral coordination DSML < AS, DSA, Domain Specific Events, MoCC > fuml merge TFSM i:=0 After 10 /! ping i:=0 doit s1 s2 merge I > 10 /! finish [true]? pong / i++ i==10 done i:= i+1 111

102 model behavioral coordination DSML < AS, DSA, Domain Specific Events, MoC > MyFUMLModel MyTFSMModel i:=0 merge p1 p1 After 10 /! ping i:=0 p1 p1 s1 s2? pong / i++ doit p2 p2 I > 10 /! finish merge p2 p2 [true] i==10 p3 p3 done i:= i+1 Automatically extracted model interface (Instances of Domain Specific Events acts as coordination interfaces) 112

103 model behavioral coordination DSML < AS, DSA, Domain Specific Events, MoC > MyFUMLModel MyTFSMModel i:=0 merge p1 p1 After 10 /! ping = i:=0 p1 p1 coordination s1 s2? pong / i++ doit p2 p2 I > 10 /! finish merge = p2 p2 [true] i==10 p3 p3 done i:= i+1 coordination interfaces synchronizes interface events 113

104 model behavioral coordination DSML < AS, DSA, Domain Specific Events, MoC > MyFUMLModel MyTFSMModel i:=0 merge p1 p1 After 10 /! ping = i:=0 p1 p1 s1 s2? pong / i++ doit p2 p2 I > 10 /! finish merge = p2 p2 [true] i==10 p3 p3 done i:= i+1 coordination interfaces synchronizes interface events 114

105 model behavioral coordination DSML < AS, DSA, Domain Specific Events, MoC > MyFUMLModel MyTFSMModel i:=0 merge p1 p1 After 10 /! ping < i:=0 p1 p1 s1 s2? pong / i++ doit p2 p2 I > 10 /! finish merge > p2 p2 [true] i==10 p3 p3 done i:= i+1 coordination interfaces synchronizes interface events 115

106 model behavioral coordination DSML < AS, DSA, Domain Specific Events, MoC > MyFUMLModel MyTFSMModel i:=0 merge p1 p1 After 10 /! ping i:=0 p1 p1 s1 s2? pong / i++ doit p2 p2 I > 10 /! finish merge p2 p2 [true] i==10 p3 p3 done i:= i+1 coordination interfaces synchronizes interface events 116

107 Railroad Crossing Heterogeneous Model (paper, slides) Matias Ezequiel Vara Larsen, Arda Goknil, model behavioral coordination 2013 Miami DSML < AS, DSA, Domain Specific Events, MoC > MyFUMLModel MyTFSMModel i:=0 merge p1 p1 After 10 /! ping i:=0 p1 p1 s1 s2? pong / i++ doit p2 p2 I > 10 /! finish merge p2 p2 [true] i==10 p3 p3 done i:= i+1 coordination interfaces synchronizes interface events 117

108 Matias Vara Larsen Heterogeneous development DSML 2 <<Conforms To>> DSML 1 DSML 3 Language designer 2 <<Conforms To>> <<Conforms To>> Language designer 3 Language designer 1

109 Matias Vara Larsen Heterogeneous development DSML 2 <<Conforms To>> DSML 1 DSML 3 Language designer 2 <<Conforms To>> <<Conforms To>> Language designer 3 Language designer 1 Coordination

110 S Matias Vara Larsen An integrator must specify how the models communicate/coordinate /interact one each other DE F IN E Integrator Integrato r DSML 2 <<Conforms To>> DSML 1 DSML 3 Language designer 2 <<Conforms To>> <<Conforms To>> Language designer 1 Coordination Language designer 3

111 Matias Vara Larsen BCOoL BCOol is a language for the integrator to express its coordination skills Integrator Integrato r DSML 2 <<Conforms To>> DSML 1 DSML 3 Language designer 2 <<Conforms To>> <<Conforms To>> Language designer 1 Coordination Language designer 3

112 Matias Vara Larsen BCOoL ConformsTo BCOol spec DSML 1 Behavioral Behavioral Interface Interface DSML 2 Behavioral Behavioral Interface Interface DE FI NE S BCOoL captures Coordination Patterns between languages DSML 3 Conforms To Conforms To Conforms To Coordination Behavioral Behavioral Interface Interface

between models is automatically generated

FI NE S BCOoL DSML 3 Conforms To Conforms To

113 Matias Vara Larsen By instantiating a Coordination Pattern, the coordination between models is automatically generated ConformsTo BCOol spec DSML 1 Behavioral Behavioral Interface Interface DSML 2 Behavioral Behavioral Interface Interface DE FI NE S BCOoL DSML 3 Conforms To Conforms To Conforms To GENERATES Coordination Behavioral Behavioral Interface Interface

114 Matias Vara Larsen Sum up on our approach DSML 1 Behavioral Behavioral Interface Interface Behavioral Behavioral Interface Interface DSML 2 Language Conforms To Conforms To Model parametrizes parametrizes Runtime Execution Engine Execution Engine

115 Matias Vara Larsen Sum up on our approach DSML 1 Behavioral Behavioral Interface Interface Behavioral Behavioral Interface Interface DSML 2 Language Conforms To Conforms To generates generates Model parametrizes parametrizes Runtime Execution Engine Execution Engine

116 Matias Vara Larsen Sum up on our approach DSML 1 Behavioral Behavioral Interface Interface Behavioral Behavioral Interface Interface imports imports DSML 2 Language BCOol spec Conforms To generates generates imports generates Model imports Coordination parametrizes parametrizes Synchronizations / Exchanges Execution Engine Conforms To parametrizes Synchronizations / Exchanges Orchestration/Coordination engine Runtime Execution Engine

117 Sum up on our approach DSML 1 Behavioral Behavioral Interface Interface Behavioral Behavioral Interface Interface imports imports DSML 2 Language BCOol spec Conforms To generates generates imports generates Model imports Coordination parametrizes parametrizes Synchronizations / Exchanges Execution Engine Conforms To parametrizes Synchronizations / Exchanges Orchestration/Coordination engine Runtime Execution Engine

Only Onlyfor for discrete discretesemantics semantics DSML 1 Behavioral Behavioral Interface Interface Sum up on our approach Behavioral Behavioral Interface Interface imports imports DSML 2 Language

118 Only Onlyfor for discrete discretesemantics semantics DSML 1 Behavioral Behavioral Interface Interface Sum up on our approach Behavioral Behavioral Interface Interface imports imports DSML 2 Language BCOol spec Conforms To generates generates imports generates Model imports Coordination parametrizes parametrizes Synchronizations / Exchanges Execution Engine Conforms To parametrizes Synchronizations / Exchanges Orchestration/Coordination engine Runtime Execution Engine

119 Current work DSML 1 Behavioral Behavioral Interface Interface imports Language BCOol spec Conforms To generates generates imports Existing tool executedby parametrizes Synchronizations / Exchanges Model?? Coordination parametrizes Execution Engine?? imports?? Runtime Orchestration/Coordination engine

120 Current work DSML 1 Behavioral Behavioral Interface Interface imports Modelica abstraction Language BCOol spec Conforms To generates generates imports Existing tool executedby parametrizes Synchronizations / Exchanges Model?? Coordination parametrizes Execution Engine Behavioral Behavioral Interface Interface imports?? Runtime Orchestration/Coordination engine

121 Current work DSML 1 Behavioral Behavioral Interface Interface Behavioral Behavioral Interface Interface imports imports BCOol spec Conforms To generates Conforms To generates Execution Engine Model executedby parametrizes Synchronizations / Exchanges generates Abstracted model Coordination parametrizes Language Existing tool imports imports Modelica abstraction?? Runtime Orchestration/Coordination engine

122 Current work We Wedid didhome homemade madeconnectors connectorswith withscilab... scilab... DSML 1 Behavioral Behavioral Interface Interface imports CPU Temperature Evolution Conforms To (Scilab) Behavioral Behavioral Interface Interface imports BCOol spec generates Conforms To Modelica abstraction Language generates Existing tool generates Coordination Control evolution TimeSquare parametrizes Abstracted model imports imports Model executedby parametrizes Amani Khecharem, Carlos Gomez,, Frédéric Mallet, Robert de Simone. Synchronizations Execution of Heterogeneous Models for Thermal Analysis with a Multi-view Approach. FDL 2014 : Forum on specification and Design Languages, Oct / 2014, Munich, Germany. IEEE Exchanges Execution Engine?? Runtime Orchestration/Coordination engine

123 Michele Fabbri FMI / FMU Functional Mock-up Interface (FMI) is a tool independent standard to support both model exchange and co-simulation of dynamic models using a combination of xml-files and compiled C-code

124 Michele Fabbri FMI / FMU Need for an handcraft orchestrator on the bus... Functional Mock-up Interface (FMI) is a tool independent standard to support both model exchange and co-simulation of dynamic models using a combination of xml-files and compiled C-code

125 Current work Michele Fabbri DSML 1 Behavioral Behavioral Interface Interface Behavioral Behavioral Interface Interface imports imports BCOol spec Conforms To generates Conforms To generates generates Coordination parametrizes Abstracted model parametrizes Language Existing tool imports imports Modelica abstraction Model executedby Runtime Execution Engine Orchestration/Coordination engine

126 Michele Fabbri TFSM generates Preliminary results Behavioral Behavioral Interface Interface Conforms To imports imports fake model imports parametrizes parametrizes Execution Engine executedby Orchestration/Coordination engine 136

127 Michele Fabbri Preliminary results TFSM generates Conforms To imports imports imports fake model tion Behavioral Behavioral Interface Interface parametrizes parametrizes Execution Engine executedby Orchestration/Coordination engine 137

128 Conclusion 138

129 Conclusion 139

130 Conclusion Languages Languagesshould shouldreify reifytheir theirbehavioral behavioralsemantics... semantics... so sothat thatwe wecan canunderstand understandand andautomate automatereasoning... reasoning... on onthe theexecution executionand andcoordination coordinationof ofmodels... models... at atdifferent differentlevel levelof ofabstraction... abstraction

131 This is the end......thanks... University of Nice, I3S CNRS INRIA Aoste 141

Composition and concurrent execution of heterogeneous domain-specific models

Composition and concurrent execution of heterogeneous domain-specific models A work part of the GEMOC initiative Benoit Combemale Associate Professor, University of Rennes 1 Research Scientist, INRIA benoit.combemale@irisa.fr