Quiescence in Internal Propagation

Abstract Concurrent design often involves a large number of variables related through a complex network of constraints. This not only makes it difficult to find a design solution, but also to understand the design parameter trade-offs, trends and interactions. Precise estimates of all the design parameters which would facilitate the exploration of tradeoffs and interactions are generally not available, however, the allowable ranges or intervals of some design parameters are often known. Narrower, or more refined intervals, facilitate the inference of stronger conclusions and the more robust application of mathematical optimization and constraint reasoning tools. Refinement of intervals can be carried out conveniently by constraint propagation, however, direct constraint propagation is often a computationally slow process. In this paper we study constraint propagation at a fundamental level and identify certain properties of the process. We then propose an alternative method in which these properties are utilized to increase considerably the speed and accuracy with which the final quiescent state of constraint propagation can be computed.