Higher-Order Functional Constraint Networks

Abstract

This paper discusses value propagation on functional constraint networks. We first focus on conventional “first-order” networks, and then show how the technique generalizes for what we call higher-order networks. Value propagation is attractive because of its simplicity and efficiency. Although it cannot solve all satisfiable networks (since it cannot break loops of constraints) and it may suggest solutions to unsatisfiable networks (since it does not discover inconsistency), it is of significant practical value. We describe several algorithms for planning value propagation, and point out that planning can be regarded as proof search in intuitionistic propositional logic.