Search Algorithms for Minimal Cost Repair Problems

Many scheduling scenarios involve altering or repairing an initial, candidate schedule, sometimes as a result of unexpected changes to the problem. Given a constraint satisfaction problem and an initial assignment of values to variables that violates some constraints, we consider the problem of finding a solution which differs minimally from the initial state. We consider two search spaces for optimally solving this problem, a commitment-based search space and a difference-based search space. Two search strategies are evaluated – a depth-first branchand-bound strategy as well as an iterative-deepening strategy. We present preliminary empirical investigations of these search spaces and strategies are on a simple resource-scheduling repair problem, as well as a resource allocation repair problem related to server consolidation. An iterative-deepening strategy applied to a commitment-based search space is shown to have the best performance on these two problems.