Use of ontologies for validating mas analisys models

• Autores: Antonio Alejandro López Lorca
• Directores de la Tesis: Rodrigo Martínez Béjar (dir. tes.)
• Lectura: En la University of Wollongong ( Australia ) en 2012
• Idioma: inglés
• Número de páginas: 257
• Texto completo no disponible (Saber más ...)
• Resumen

  • Ontologies, as a knowledge representation tool, rely on formal de- scriptions of semantics. They have often been used in software devel- opment to support various activities and generally improve the value of the systems produced. However, their use during requirements engineering activities to validate that the product being developed complies with the client’s conceptualisation is largely unexplored. In the field of Agent Oriented Software Engineering (AOSE), a few of the extant methodologies contemplate the use of ontologies but only for modelling the domain problem to support agent communication.

    Due to the complexity of Multi Agent Systems (MAS), errors in modelling activities during their development can be costly. Early validation of MAS models can prevent rework or the building of a system that is non-compliant with the client’s specification. This thesis produces an automatic validation and verification MAS mod- els process that can be applied to the model development process defined by any of the extant AOSE methodologies. An ontology- based validation and verification MAS models process add-on and an associated automatic support tool are developed. The develop- ment of the process and the tool is interleaved with three case studies to evaluate and refine them. The process is conducted iteratively to accommodate the lifecycle defined by most AOSE methodolo- gies, and it concurrently incrementally validates the MAS models produced.

    A key contribution of this research is automating the bulk of the complex tasks in the ontology-based validation process, harnessing the automatic reasoning opportunities offered by the use of formal ontologies. The process utilises an AOSE metamodel that describes the most common features of MAS identified in the literature. The process validates the MAS models for completeness against the cli- ent’s conceptualisation and verifies their structure for consistency.

    When problems are detected, the process supports their rectification by suggesting changes to the requirements that may have remained v undetected to the developers. The process effectiveness is validated using different case studies. Validity threats are mitigated by ensur- ing that different ontology engineers and AOSE modellers are used, that various problem domains are selected for the applications being developed, and that different AOSE methodologies are applied; this is achieved by varying the case studies.