Dynamic branching in qualitative constraint-based reasoning via counting local models

Abstract

We introduce and evaluate dynamic branching strategies for solving Qualitative Constraint Networks (QCNs), which are networks for representing and reasoning about spatial and temporal information in a natural manner, e.g., a constraint can be “Task A is scheduled after or during Task C”. Specifically, we propose heuristics that dynamically associate a weight with a relation in the branching decisions that occur during backtracking search, based on the count of local models that the relation is involved with in a given QCN. Experimental results with a random and a structured dataset of QCNs of Interval Algebra show that it is possible to achieve up to 5 times better performance for structured instances, whilst maintaining non-negligible gains of around 20% for random ones. Finally, we show that these results may be notably improved via a selection protocol algorithm that synthesizes the involved heuristics into an overall better performing meta-heuristic in the phase transition.