When do bounds and domain propagation lead to the same search space

Abstract

This paper explores the question of when two propagation-based constraint systems have the same behaviour, in terms of search space. We categorise the behaviour of domain and bounds propagators for primitive constraints, and provide theorems that allow us to determine propagation behaviours for conjunctions of constraints. We then show how we can use this to analyse CLP(FD) programs to determine when we can safely replace domain propagators by more efficient bounds propagators without increasing search space.