Abstract :
Constraint propagation is an important inference engine for different Al applications that reason about quantities. Many systems [4, 6, 9, 14] represent the knowledge they reason about in terms of mathematical relations on qualitative or quantitative values of quantities, normally expressed as constraints. Constraint propagation has to be solved efficiently to allow Al applications to react in a reasonable time. This has been shown to be a hard problem [3]. Constraints can be classified into disjoint sets for which efficient propagation algorithms can be implemented. This paper is a report on the design and implementation of HRCP (Hybrid Representation Constraint Propagation), a constraint propagation engine that separates constraints into sets with different representations. HRCP can be extended to accept other implementations to fulfill the application needs. All this, of course, must be transparent to the final user of the propagation engine, whio sees the constraints in a uniform representation.
