A Best First approach for solving over-constrained dynamic problems

Many real-life problems tackled using Constraint Programming are dynamic. Addition and deletion of constraints may lead to inconsistencies. When a solution is required, constraints may be violated in order to ful ll the user requirements. In this paper, we propose a bestrst search method (parameterized by the domain of constraints and the associated solver) for solving over-constrained problems arising in dynamic environments. We propose a specialization of this approach for csp using arc-consistency. We address complexity issues to show the tractability of our approach and describe our implementation: the decorum system.