Consistency Check in Modelling Multi-Agent Systems

1
Consistency Check in Modelling Multi-Agent
Systems

• Lijun Shan Hong Zhu
• Dept of Computer Science Dept of Computing
• National Univ.of Defense Tech. Oxford Brookes
  Univ.
• Changsha, 410073, China Oxford OX33 1HX, England
• lijunshancn_at_yahoo.com hzhu_at_brookes.ac.uk

2
Outline

• Motivation and related work
• Agent-oriented software development (AOSD)
• Model-driven software development (MDSD)
• Roles and problems of consistency checking in
  MDSD
• Definition of consistency constraints
• Consistency checkers in a modelling environment
• Case studies
• Conclusion

3
Motivation

• Agent technology
• Widely recognized as a viable solution for
  applications in dynamic environments such as the
  Internet
• Agent-oriented methodologies, such as Gaia,
  Tropos, AUML, etc.
• CAMLE Caste-centric Agent-oriented Modelling
  Language and Environment
• as a method of requirements analysis and
  specification
• as the base for the design and implementation of
  MAS
• to generate formal specifications in SLABS
• to write programs in AO programming languages

4
Model-driven SW development

• Models play the central role in SW development
• modelling real world systems
• as a means of requirements analysis
• modelling software/information systems
• as specifications of software systems
• to analyse systems properties, etc.
• as high level designs
• to derive or generate code
• even as executable representation of software
  systems
• as bases of software testing, etc.
• Characteristics of modern modelling languages
• Semiformal diagrammatic notation with defined
  semantics
• Facilities to support separation of concerns
• Multiple views
• Hierarchically structured collection of diagrams

5
Consistency as a basic quality attribute

• Consistency
• Between different views
• Between models at different levels of abstraction
  
• Consistency is crucial in MDSD
• A model must be consistent before it is
  transformed into other forms, e.g.
• the generation of code,
• the derivation of test cases,
• the uses as software documentation,
• A model must be consistent before the derivation
  and proof of the properties of the system
• Ensuring consistency is very difficult
• Related to the semantics of the model
• Complexity due to multiple views and multiple
  levels

6
Consistency check a challenging problem

• The aim of consistency check is to provide a
  partial but effective solution to model
  consistency problem.
• Defining a set of consistent constraints that can
  detect a large number of common errors in models
• Design algorithms that can automatically check if
  a model satisfies the consistency constraints

• The idea of consistency check in modeling
  languages is similar to type check in programming
  languages
• defining a type system to prevent type errors
• device a static type checking mechanism

7
Consistency constraints

• What are consistent constraints?
• Restrictions on the uses of diagrammatic
  notations, variables and names, types and symbols
  in a modeling language to reduce the possibility
  of inconsistency
• Related to the semantics of the diagrams
• Must be able to be syntactically checked
• Typical examples
• The same identifier that occurs at different
  places must refer to the same entity
• An entity should be referred to by the same
  identifier if it occurs at different diagrams

8
Types of consistency constraints

• Intra-model consistency
• Within one type of model
• Intra-diagram
• Within one diagram
• Inter-diagram
• Between different diagrams of the same model
• Inter-model consistency
• Between different types of models, hence also
  different diagrams

• Horizontal consistency
• Between models/diagrams of the same level of
  abstraction
• Vertical consistency
• Between models/diagrams at different levels of
  abstraction
• Local
• Between two levels
• Global
• With respect to the overall structure

9
Related work

• Defining consistency constraints for UML
• (Andre, P. 2000), (Pap, Z. S. 2001), (Nentwich,
  C. 2001), (Paige, R. F. 2002), (Astesiano, E.
  2003)
• Generic tools
• Xlinkit language for defining consistency
  engine for automated checking, (Nentwich, C.,
  Emmerich, W., Finkelstein, 2003)
• Formal methods (e.g. SPIN)
• (Inverardi, P. 2001), (Schafer, T. 2001)
• As a part of modelling environment
• NDRASS (structured modelling) (Xu, J., Jin, L.,
  Zhu, H. 1996)

10
Overview of CAMLE

• Conceptual model of multi-agent systems 2
• Agent
• an active computational entity that encapsulates
  data, operations and behaviours and situates in
  its designated environment
• state space
• a set of operations/actions
• designated operating environment
• a set of behaviour rules
• Caste
• a set of agents that have the same structural
  and behavioural characteristics
• agents are instances of, but have dynamic
  memberships of castes
• inheritance relationship between caste

11
Overall Structure of CAMLE Models
Collaboration Models and Behavior models

• A CAMLE model consists of
• a caste model
• collaboration models
• behaviour models

Caste Model with Whole-Part Relations
12
Caste model

• Caste model
• models the organizational structure
• Caste diagram
• Defines the castes in the system
• Specifies their relations

Notation
Example
13
Constraints on Caste Model

• Constraint 1a) A caste diagram defines a naming
  space in which each node defines a caste with a
  unique name.
• Constraint 1b) Each link defines a binary
  relation on castes by linking two nodes in the
  diagram.
• Constraint 1c) An inheritance relation and a
  migration relation must be associated to two
  different caste nodes.
• Constraint 1d) Inheritance relations must not
  form any loops.

14
Collaboration models

• Collaboration Diagram
• Describes the dynamics of a system from
  communication perspective
• The structure of collaboration models
• A hierarchic of collaboration models as
  refinement of castes
• A collaboration model for each compound caste
• a general collaboration diagram a set of
  scenarios specific collaboration diagrams

Notation
Example
15
Constraints on Collaboration Model

• Well-formedness conditions
• Constraint 2a) Each caste / agent node must have
  a unique name.
• Constraint 2b) The number assigned to an action
  indicating its temporal order must be unique, if
  any.
• Consistency between general and specific
  diagrams, e.g.
• Constraint 2c) Every agent node in the general
  collaboration diagram G must appear in at least
  one scenario-specific collaboration diagram.
  Formally,
• ?n?ANode (G).?D?S. (n?ANode (D))
• Consistency between models at different levels,
  e.g.
• Constraint 2j) For all castes C in a
  collaboration model M, Cs environment described
  in M must be equal to Cs environment described
  in Cs collaboration model MC.
• ?n.(n?Env(MC) ? ?a?Interaction(G).(nBegin(a)?C
  End(a)))

16
Behaviour models

• Scenario diagram
• Describes a typical situation in the operation of
  a system from an agents view.
• used in behaviour diagrams
• behaviour diagram
• define behaviour rules of an agent

17
Constraints on Behaviour Model

• Well-formedness conditions, e.g.
• Constraint 3a) The temporal order between events
  must be linear
• Consistency between behaviour diagrams and
  scenario diagrams, e.g.
• Constraint 3d) In a behaviour diagram, every
  scenario reference node must refer to a scenario
  defined by a scenario diagram.
• ?n? ScenarioNode(DC). ?S? SC.(Name(n)Name(S)

18
Inter-Model Consistency

• Between collaboration model and caste model
• Constraint 4a) Every caste in a collaboration
  model CD must be a caste in the caste model C.
  Formally,
• ?D?CD.?n?Node(D).?n?Node(C).(CName(n)Name(n))
• Constraint 4b) The hierarchical structure of the
  collaboration models must be consistent with the
  whole-part relations between castes defined in
  caste diagram C.
• ?MA,MB?CM.(MB lt MA??R? Aggr(C).(R(B,A))

19
Between behaviour model and caste model

• Constraint 4c) For each behaviour model BM, the
  caste that BM defines its behaviour must be in
  declared in the caste model C.
• ?B?BM.?n?Node(C).(Caste(B) n ).
• Constraint 4d) Every agent occurs in a scenario
  in a behaviour model BM must have its caste
  defined in the caste model C.
• ?B?BM.?a?Agents(B).?n?Node(C).
• (Caste-of(a)Name(n))
• Constraint 4e) The join, quit and move
  actions occur in the behaviour model of a caste c
  must be consistent with the migration relation
  described in the caste model.

20
Between collaboration model and behaviour model

• Constraint 4f) Every visible action of caste C
  defined in the collaboration model CM must occur
  in Cs behaviour model BMC or at least one of Cs
  components as a result action.
• ?a?VisibleActions(C).(?r?Rules(BC)?
  (?M?Components(C). ?r? Rules(BM)). (aAction(r))
• Constraint 4g) For each scenario used in the
  definition of caste Cs behaviour, the agents
  and/or castes that the scenario referred to must
  occur in the collaboration model CM as Cs
  collaborators.
• ?Sc?Scenarios(BC).?G?Ref(Sc).G?Collaborators(C)
• Constraint 4h) The agents/castes referred to in a
  scenario must be in the environment of the caste
  as described by the collaboration model.
• ?Sc?Scenarios(B).?G?Ref(Sc).(G?Env(C))
• Constraint 4i) Every action that a scenario
  refers to in a behaviour diagram must be a
  visible action of the caste of the scenario.
• ?Sc? Scenarios(BC).?a? ReferredActions(C, Sc).
  (a?VisibleActions(C)).

21
Summary
Horizontal Consistency Vertical Consistency Vertical Consistency
Horizontal Consistency Local Global
Intra-model Intra- diagram 1a, 1b, 1c, 1d, 1e, 2a, 2b, 3a, 3b, 3c - -
Intra-model Inter-diagram 2c, 2d, 2e, 2f, 2g, 2h, 2i, 3d 2j, 2k -
Inter-model Inter-model 4f, 4g, 4h, 4i 4e 4a, 4b, 4c, 4d
22
Consistency check tools in CAMLE

• Well-formedness conditions are checked during
  model construction in the process of interactive
  diagram editing
• Other consistency constraints are checked by a
  set of tools each for one specific constraint
• Consistency checking tools are called in the
  following order to reduce the complexity of error
  handling
• Intra-model intra-diagram consistency checkers
• Intra-model inter-diagram consistency checkers
• Inter-model local consistency checkers
• Inter-model global consistency checkers

23
Architecture of CAMLE Environment
24
Case studies

• United Nations Security Council
• the organisational structure and the working
  procedure to pass resolutions
• Amalthaea
• an evolutionary multi-agent system developed at
  MITs Media Lab for retrieving information from
  the Internet
• University
• a partial model of the university organisation
  and working procedures
• Web Services
• a model of the architecture of web services and
  an application of web services of online auctions

25
Observations on case studies

• In all cases, automated consistency check
  detected a large number of errors
• Models that passed consistency checks are of good
  quality
• Explicitly defined consistency constraints helped
  modeler to think more carefully in the
  construction of models, hence made fewer mistakes
  during modeling than without constraints
• The design and implementation of the transformer
  for generating formal specifications become much
  simpler with the assumption that certain
  inconsistency can be ruled out

26
Conclusion

• Roles of consistency constraints in MDSD
• Quality assurance in modelling process
• Consistency constraints can be defined to
  effectively improve the quality of models and the
  productivity in model construction
• Consistency checking can be implemented
  efficiently as a very useful tool in a modeling
  environment
• Partial specification of the correctness of the
  transformation rules between models
• The diagram generators in CAMLE can guarantee
  that a generated model (or partial model) is
  consistent with the original diagram(s).
• Further work
• Statistical analysis of the effectiveness of
  consistency constraints
• Further development of CAMLE environment

27
References

1. Shan, L. Zhu, H. CAMLE A Caste-Centric Agent
   Modelling Language and Environment. Proc. of
   SELMAS at ICSE04, Edinburgh, Scotland (UK), May
   2004.
2. Zhu, H. SLABS A Formal Specification Language
   for Agent-Based Systems, Journal of Software
   Engineering and Knowledge Engineering 11(5),
   529-558, 2001.
3. Pap, Z. S., Majzikl, I., Pataricza, A, Szegi,
   A. Completeness and Consistency Analysis of UML
   Statechart Specifications. Proc. of IEEE Design
   and Diagnostics of Electronic Circuits and
   Systems Workshop, 83-90. 2001.
4. Nentwich, C., Emmerich, W., Finkelstein, A.
   Flexible Consistency Check. ACM Transactions on
   Software Engineering and Methodology 12 (1),
   28-63, 2003.
5. Inverardi, P., Muccini, H., Pelliccione, P.
   Automated check of architectural models
   consistency using SPIN. Proc. of 16th IEEE Int.
   Conf. on Automated Software Engineering, San
   Diego, California, p.346, 2001.