City Research Online

Abstract

This thesis is concerned with constraint-based logical and computational frameworks for resolving and relaxing over-constrained systems. The context is provided by two such frameworks, Hierarchical Constraint Logic Programming (HCLP) and Partial Constraint Satisfaction Problem techniques (PCSP), both of which have been extensively discussed in the literature. Our work is driven by the reasons why over-constrained systems arise, the characteristics of the ‘ideal’ paradigm for resolving them, and the issue of compositionality which is very important in general if we wish to examine large systems by examining and then combining smaller parts. We abstract away from programming language issues in order to focus on constraint solving.

The main original work of this thesis is divided into three parts. Firstly we present a complete method for transforming between the HCLP and PCSP representations of a problem, thus showing that theoretically they have equivalent expressive power. Secondly, having discussed compositionality in general, we present a two-stage variant of HCLP; the first stage is compositional but calculates a superset of the solutions we expect from HCLP. The second stage removes precisely those solutions which are not acceptable to HCLP, but at the cost of re-introducing HCLP’s non-compositional behaviour. We also discuss the compositional aspects of PCSP.

The third part of this thesis presents Gocs, our system which allows the use of both the HCLP and PCSP approaches to problem relaxation and ordering. The Gocs integrated framework has HCLP and PCSP as special cases and also subsumes all of their separate advantages, when considering the characteristics of the ideal system for relaxing and resolving over-constrained problems. We present examples throughout the thesis, some of which are comparative and so may be used to substantiate our claims. Finally, we present conclusions and discuss further work.

Publication Type:	Thesis (Doctoral)
Subjects:	Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Departments:	School of Science & Technology > Computer Science School of Science & Technology > School of Science & Technology Doctoral Theses Doctoral Theses

Export

Downloads