Facilitating MAS Complete Life Cycle through the Protégé-Prometheus Approach

Please download to get full document.

View again

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
 37
 
 

Humor

  The approach of this paper aims to support the complete multi-agent systems life cycle, integrated by two existing and widely accepted tools, Protégé Ontology Editor and Knowledge-Base Framework, and Prometheus Development Kit. A general sequence of
Share
Transcript
  Facilitating MAS Complete Life Cycle throughthe Prot´eg´e-Prometheus Approach Marina V. Sokolova 1 , 2 and Antonio Fern´andez-Caballero 1 1 Universidad de Castilla-La Mancha, Escuela Polit´ecnica Superior de Albacete &Instituto de Investigaci´on en Inform´atica de Albacete, 02071-Albacete, Spain 2 Kursk State Technical University, Kursk, ul.50 Let Oktyabrya, 305040, Russia smv1999@yandex.ru,caballer@dsi.uclm.es Abstract.  The approach of this paper aims to support the completemulti-agent systems life cycle, integrated by two existing and widely ac-cepted tools, Prot´eg´e Ontology Editor and Knowledge-Base Framework,and Prometheus Development Kit. A general sequence of steps facili-tating application creation is proposed in this paper. We propose that itseems reasonable to integrate all traditional software development stagesinto one single methodology. This view provides a general approach forMAS creation, starting with problem definition and resulting in programcoding, deployment and maintenance. The proposal is successfully beingapplied to situation assessment issues, which has concluded in an agent-based decision-support system for environmental impact evaluation. Keywords:  Multi-agent systems, Software life cycle, Methodologies. 1 Introduction Nowadays there are many works and approaches dedicated to multi-agent sys-tems (MAS) development, which pay attention to internal MAS functionality,reasoning and its coding. Creation, deployment and post-implementation of MASas software products is a complex process, which passes through a sequence of stages forming its life cycle [13][21]. Every step of the life cycle process has to be supported and provided by means of program tools and methodologies. Incase of MAS development, in our opinion there is still no solution to a unifiedapproach to cover all the stages. However, there are some previous works dedi-cated to this issue [2][12]. For instance, de Wolf and Holvoet [2] have introduced a methodology in the context of standard life cycle model, with accent to decen-tralization and macroscopic view of the process. The authors offer their approachon the assumption that the research task has already been defined, omitting theproblem definition and domain analysis stages of MAS development process.But, a complete software development in case of MAS should be based on thefollowing steps: (1) domain and system requirements analysis, (2) design, (3)implementation, (4) verification, and, (5) maintenance [9][12].Some well known alternative agent-oriented software engineering methodolo-gies, including MaSE [1] , Gaia [23], MASDK [6], Prometheus [15], Tropos [4], N.T. Nguyen et al. (Eds.): KES-AMSTA 2008, LNAI 4953, pp. 63–72, 2008.c  Springer-Verlag Berlin Heidelberg 2008  64 M.V. Sokolova and A. Fern´andez-Caballero INGENIAS [5], among others, support some of the cited stages of MAS life cycleprocess. Nonetheless, these methodologies often work under the condition thatthe developer has already defined the problem and determined the goals and thetasks of the system. However, domain analysis is a crucial stage and has to bescrutinizingly examined and planned. Indeed, the whole deployed system func-tionality and efficiency depends on how precisely the problem was defined andthe domain ontology was elaborated. In the most general case, when a MAS isdistributed and has to deal with heterogeneous information, the domain analysisbecomes even more important.Therefore, it seems reasonable to integrate all the software development stagesinto one single methodology, which should provide a general approach to MAScreation, starting with the problem definition and resulting in program coding,deployment and maintenance. As a tool for the system and domain require-ments, we suggest using an OWL-language based toolkit, as OWL has becomea standard for ontologies description [14]. The Prot´eg´e Ontology Editor andKnowledge-Base Framework [16] complies a set of procedures for ontology cre-ation and analysis, offering a set of plug-ins covering viewers, problem-solvingmethods, knowledge trees, converters, etc. According to our proposal, ontolo-gies can be represented by means of Prot´eg´e and later may be incorporated intoMAS. In order to provide the following stages with tools we have tested differentmethodologies. We came to the conclusion to use the Prometheus DevelopmentTool (PDT) [17], which provides a wide range of possibilities for MAS planningand implementation: the system architecture, the system entities, their internalsand communications within the system and with outer entities. The most im-portant advantages of PDT are an easy understandable visual interface and thepossibility to generate code for JACK TM  Intelligent Agents [11]. The proposalis summed up in Fig. 1. 1. Domain and System Requirements Analysis 2. Design 3. Implementation 4. Verification 5. Maintenance Protégé Ontology Editor (creation of metaontology and private ontologies) Prometheus Design Toolkit (system elements analysis, MAS design, skeletton code generation) JACK  (MAS coding, testing and support) Fig.1.  The Prot´eg´e-Prometheus approach applied to the MAS life cycle  Facilitating MAS Complete Life Cycle 65 The paper is organized as follows. In section 2 the metaontology creationrealized in Prot´eg´e is described and in section 3 the MAS designed in PDT isintroduced. In section 4 our intention to implement the ideas for further usageof the integrated methodology are briefly explained. 2 Domain and MAS Requirements in Prot´eg´e Ontology creation may be viewed as a crucial step in MAS design as it determinesthe system knowledge area and potential capabilities [7]. In the first part of this article a model of distributed metaontology that serves as a framework forMAS design is proposed. Its components - private ontologies - are described inextensive with respect to an application area and in terms of the used semantics.When defining an ontology  O  in terms of an algebraic system, we have thefollowing three attributes: O  = ( C,R,Ω  ) (1)where  C   is a set of concepts,  R  is a set of relations among the concepts, and  Ω  a a set of rules. The principal point of MAS is to determine the rules  Ω   and toevaluate them. Formula (2) proposes that the ontology for the domain of interest(or the problem ontology) may be described by offering proper meanings to  C  , R  and  Ω  .The model of the metaontology that we have created consists of five compo-nents, or private ontologies: the “Domain Ontology”, the “Task Ontology”, the“Ontology of MAS”, the “Interaction Ontology” and the “Agent Ontology”.In first place, the “Domain Ontology”, includes the objects of the problemarea, the relations between them and their properties. It determines the compo-nents  C   and  R  of expression (2), which is detailed as: OD  = < I,C,P,V,Rs,Rl >  (2)where the set  C   (see formula (2)) is represented by two components: Individuals( I  ) and Classes ( C  ), which reflect the hierarchical structure of the objects of theproblem area;  P   - are class properties;  V    - are the properties values;  Rs  - arevalues restrictions;  Rl  embodies the set  R , and includes rules which state howto receive new individuals for the concrete class.The “Task Ontology” contains information about tasks and respective meth-ods, about the pre-task and post-task conditions, and informational flows forevery task. The “Task Ontology” has the following model: OT   = < T,M,In,Ot,R >  (3)where  T   is a set of tasks to be solved in the MAS, and  M   is a set of methodsor activities related to the concrete task,  In   and  Ot   are input and output dataflows,  R  is a set of roles that use the task. Component  R  is inherited fromthe “Ontology of MAS” through the property  belong to role . The tasks areshared and accomplished in accordance with an order.  66 M.V. Sokolova and A. Fern´andez-Caballero The “Ontology for MAS” architecture is stated as: OA  = < L,R,IF,Or >  (4)where  L  corresponds to the logical levels of the MAS (if required),  R  is a set of determined roles,  IF   is a set of the corresponding input and output informationrepresented by protocols. Lastly, the set  Or   determines the sequence of executionfor every role (orders).The interactions between the agents include an initiator and a receiver, ascenario and the roles taken by the interacting agents, the input and outputinformation and a common communication language. They are stated in the“Interaction Ontology” as: OI   = < In,Rc,Sc,R,In,Ot,L >  (5)Actually, as  In   and  Rc   Initiator and Receiver, respectively, of the interactionwe use agents. The component  Sc   corresponds to protocols.  R  is a set of rolesthat the agents play during the interaction.  In   and  Ot   are represented by in-formational resources, required as input and output, respectively. Language  L determines the agent communication language (ACL).In our approach BDI agents [3], which are represented by the “Agent On-tology”, are modeled. Hence, every agent is described as a composition of thefollowing components: Agent  = < B,D,I >  (6)Every agent has a detailed description in accordance with the given ontology,which is offered in a form of BDI cards, in which the pre-conditions and post-conditions of agent execution and the necessary conditions and resources for theagent successful execution are stated. Evidently,  B ,  D  and  I   stand for Believes,Desires and Intentions, respectively.Metaontology is a specification for further MAS coding; it provides the neces-sary details about the domain, and describes system entities and functionality.It includes five components: MO  = < OD,OT,OA,OI,Agent >  (7)where  OD  stands for the ”‘Domain Ontology”’,  OT   for the “Task Ontology”, OA  “Ontology for MAS” architecture,  OI   is the “Interaction Ontology”, and, Agent  is the “Agent Ontology”.Private ontologies mapping is made through slots of their components. So,the “Agent Ontology” has four properties:1.  has intentions  - which contains individuals of the methods “M” class fromthe “Task Ontology”.2.  has believes  - which contains individuals from the “Domain Ontology”.3.  has desires  - which contains individuals from the “Task Ontology”.4.  has type  - which contains variables of   String  type.  Facilitating MAS Complete Life Cycle 67 There is a real connection between the “Task Ontology” and the “DomainOntology”. The  OT  , in turn, refers to the “Ontology of MAS” ( OA ), which isformally described by four components. The first two –  level value –  order contain values of   Integer  type, which determine the logical level number andthe order of execution for every role. Roles ( R ) are the individuals of the namedontology. The next two properties –  has input –  has output refer to individuals of the “Interaction Ontology”; in particular, to protocols,which manage communications. Their properties are of type  String : –  has scenario , –  language , –  roles at scenario .The “Interaction Ontology” slots named  has initiator  and  has receiver are the individuals of the “Agent Ontology” ( Agent ). Thus, agents are linked Fig.2.  Metaontology as a result of private ontologies mapping
Related Search
Similar documents
View more
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x