This paper explores the solution of the VT Sisyphus II challenge using a Constraint Satisfaction Problem (CSP) paradigm and is an extension of the ExtrAKTor work presented at EKAW 2006. ExtrAKTor takes a Protege KB describing a propose and-revise (PnR) problem, including both constraints & fixes. Subsequently, it extracts and transforms these components so that they are directly usable by the ECLiPSe CSP toolkit to solve a range of configuration tasks. It was encouraging to note that (a) the solver coped very well with constraints involving real variables even when using a generalised propagation technique and (b) the techniques needed no "fix"' information, yet successfully dealt with the "antagonistic constraints" and the associated "thrashing" problem that had been a key issue in the original Marcus, Stout & McDermott VT paper. Consequently, we believe this is a widely useable technique for automatically generating and then solving this class of constraint problems, when they are expressed as Protege ontologies.
|Title of host publication||Research and development in intelligent systems XXV|
|Subtitle of host publication||Proceedings of AI-2008, The Twenty-eighth SGAI International Conference on Innovative Techniques and Applications of Artificial Intelligence|
|Editors||Frans Coenen, Miltos Petridis|
|Place of Publication||Godalming|
|Number of pages||14|
|Publication status||Published - 2009|