Confidence-based Reasoning in Stochastic Constraint Programming

In this work we introduce a novel approach, based on sampling, for finding policies that are likely to be solutions to stochastic constraint satisfaction problems and constraint optimisation problems. Our approach reduces the size of the original problem being analysed and it guarantees that, with a given confidence probability, the policies produced by solving this reduced problem satisfy the chance constraints in the original model within prescribed error tolerance thresholds. To do so, we blend concepts from stochastic constraint programming and statistics. The strategy introduced can be immediately employed in concert with existing approaches for solving stochastic constraint programs. We illustrate our novel approach on a number of stochastic combinatorial optimisation problems. A thorough computational study demonstrates the effectiveness of our approach.