A constraint describes a relation to be maintained

Constraint Logic Programming (CLP) is a general scheme for extending logic programming to include constraints. It is parameterized by D, the domain of the constraints. However, CLP(D) languages, as well as most other constraint systems, only allow the programmer to specify constraints that must hold. In many applications, such as interactive graphics, page layout, and decision support, one needs to express preferences as well as strict requirements. If we wish to make full use of the constraint paradigm, we need ways to represent these defaults and preferences declaratively, as constraints, rather than encoding them in the procedural parts of the language. We describe a scheme for extending CLP(D) to include both required and preferential constraints, with an arbitrary number of strengths of preference. We present some of the theory of such languages, and an algorithm for executing them. To test our ideas, we have implemented an interpreter for an instance of this language scheme with D equal to the reals. We describe our interpreter, and outline some examples of using this language.

The constraint logic programming paradigm CLP(X) (CLP for short) has been proposed by Jaffar and Lassez in order to integrate a generic computational mechanism based on constraints with the logic programming framework. This paradigm retains the semantic properties of logic languages, namely the existence of equivalent operational, model theoretic and fixpoint semantics. Moreover, since computation is performed over the particular domain of computation X , CLP(X) programs have an equivalent "algebraic" semantics, i.e. a semantics which is defined directly on the algebraic structure of the domain X . In this paper we propose an extension of such a semantics, for the success set case, in order to fully characterize the operational behaviour of programs. We introduce a framework for defining various notions of models, each corresponding to a specific operationally observable property. The construction is based on a new notion of interpretation (set of constrained atoms), on a natural exten...

Model-based diagnosis is the activity of locating malfunctioning components of a system solely on the basis of its structure and behavior. Diagnostic systems usually rely on qualitative models and reason by local constraint propagation methods. However, there is a large class of applications where ATMS-like systems or pure logic programs are unpractical since they are unable to solve simultaneous equations. In particular, modeling real-valued system parameters with tolerances requires some degree of numerical processing, and feedback loops in general cannot be resolved by Appears in Proc. 4th Intl. GI Congress (W. Brauer, D. Hernandez, Eds.), pp. 343-353, Munchen, October 23-24, 1991, Springer-Verlag (IFB 291). local constraint propagation methods. Examples of such systems are analogue circuits, e.g., amplifiers or filters. In the paper we describe the role of Constraint Logic Programs over the domain of reals (CLP(!)) in representing both, qualitative and numerical models. CLP(!)...

Graphics and modeling systems allow users to place objects or primitives in space and permit the individual editing of these objects. A more powerful paradigm allows relationships expressed by constraints to be established and maintained between pairs of these primitives. For example, two lines might be constrained to be parallel or two points constrained to be coincident. These geometric constraints lead to a \live " model in which design changes to one part of the model may induce changes elsewhere thus permitting the encoding of the designer's intent into the model. Constraints may describe complex non-linear algebraic relationships between the parameters of a model. To provide constraint satisfaction and maintenance in an interactive environment requires extremely fast and general equation solving techniques. We have developed a system called MechEdit for the interactive editing and animation

For many combinatorial problems, it is often useful to use specialized OR algorithms that allow to solve a subproblem or a simplified version of the problem. Among these, Linear Programming algorithms, and most importantly the Simplex, are quite efficient and widely used. Many applications using Constraint Programming can take benefit from using such algorithms. It is therefore very useful to integrate Linear Programming algorithms with the propagation-based algorithms found in Constraint Programming, as soon as care is taken about the communication between both kinds of algorithms. This article describes the principles and advantages of integrating a solver for linear constraints with propagation. It also demonstrates the benefits of implementing a linear constraint solver as a global constraint. Finally, experimental results on a prototype implementation with Ilog Solver are given that show the validity of the approach. Keywords: Constraint Programming. Combinatorial pr...

A widely used property of Prolog is that it is possible to write Prolog programs to construct and manipulate other Prolog programs in a very general manner. Unfortunately, this property is not carried over to richer languages such as CLP(R) -- the manipulation of CLP(R) programs in CLP(R) is quite limited. The reason is that the equality of terms in CLP(R) is not based on their syntactic structure. We propose an extended language, CLP(R+M), in which programs may be represented and structurally manipulated. Importantly, CLP(R+M) is not just a meta-language for CLP(R), but it can also be used as its own meta-language. We present a decision algorithm for R+M constraints, discuss implementation issues, and describe the implementation of a sublass of R+M constraints. Finally, by building on the extended language, we present an integrated set of system predicates and a methodology for practical meta-programming. An earlier version of this paper appeared in the Proceedings of the N...

ion Operators Igor Mozetic Austrian Research Institute for Artificial Intelligence Schottengasse 3, A-1010 Vienna, Austria igor@ai-vie.uucp Christian Holzbaur Austrian Research Institute for Artificial Intelligence, and Department of Medical Cybernetics and Artificial Intelligence University of Vienna Freyung 6, A-1010 Vienna, Austria christian@ai-vie.uucp Abstract We present two contributions to the explanation-based generalization techniques. First, the operationality criterion is extended by abstraction operators. These allow for the goal concept to be reformulated not only in terms of operational predicates, but also allow to delete irrelevant arguments, and to collapse indistinguishable constants. The abstraction algorithm is presented and illustrated by an example. Second, the domain theory is not restricted to variables with finite (discrete) domains, but can deal with infinite (e.g., real-valued) domains as well. The interpretation Appears in Machine Learning --- EWSL-91 (Y...

Model-based diagnosis is the activity of locating malfunctioning components of a system solely on the basis of its structure and behavior. In the paper we describe the role of Constraint Logic Programming (CLP) in representing models and the search space of minimal diagnoses. In particular, we concentrate on two instances of the CLP scheme: CLP(B) and CLP(!). CLP(B) extends the standard computational domain of logic programs by boolean expressions, while CLP(!) comprises a solver for systems of linear equations and inequalities over real-valued variables. 1 Introduction There are two fundamentaly different approaches to diagnostic reasoning. In the first, heuristic approach, one encodes diagnostic rules of thumb and experience of human experts in a given domain. In the second, model-based approach, one starts with a model of a realworld system which explicitly represents the structure and components of the system (e.g., Genesereth 1984, Davis 1984, de Kleer & Williams 1987, Reiter 198...