Knowledge-based automation of a design method for concurrentsystems

摘要

This paper describes a knowledge-based approach to automate ansoftware design method for concurrent systems. The approach usesnmultiple paradigms to represent knowledge embedded in the design method.nSemantic data modeling provides the means to represent concepts from anbehavioral modeling technique, called Concurrent Object-Based Real-timenAnalysis (COBRA), which defines system behavior using data/control flowndiagrams. Entity-relationship modeling is used to represent a designnmetamodel based on a design method, called COncurrent Design Approachnfor Real-Time Systems (CODARTS), which represents concurrent designs asnsoftware architecture diagrams, task behavior specifications and modulenspecifications. Production rules provide the mechanism for codifying anset of CODARTS heuristics that can generate concurrent designs based onnsemantic concepts included in COBRA behavioral models and on entitiesnand relationships included in CODARTS design metamodels. Together, thensemantic data model, the entity-relationship model, and the productionnrules, when encoded using an expert system shell, compose CODA, annautomated designer's assistant. CODA is applied to generate 10nconcurrent designs for four real-time problems. The paper reports thendegree of automation achieved by CODA. The paper also evaluates thenquality of generated designs by comparing the similarity between designsnproduced by CODA and human designs reported in the literature for thensame problems. In addition, it compares CODA with four other approachesnused to automate software design methods