A DEPENDENCY-BASED CONSTRAINT RELAXATION SCHEME FOR OVER-CONSTRAINEPROBLEMS

Abstract

In this paper, we propose an incremental constraint solver in order to realise a constraint relaxatio system for over-constrainted problems. There are two requirements to be considered. Firstly, the constraint relaxation is based on constraint propagation techniques to maintain local consistency. Secondly, in the case of inconsistency, it proposes conflict candidates to retract only those constraints which are really involved in the contradiction. Thereby, no unnecessary modifications of the original CSP are executed. What is crucial si that the relaxation operatio is incremental, in that it maintains a dependency graph among variables and constraints and updates only subset of the graph which is the real responsibility for the failure. Moreover, this system borrows intelligent backtracking techniques in determining, upon the inconsistency, a pertinent choice point which can restore the satisfiability. Therefore, these considerable terms would support constraint relaxations and experiments have shown arbitrarily speedups compared with previous approaches.