Consider the example already described, where the solution 11 is a local minimum (when a single value flip is the move used), and 00 is the global minimum. Solver provides support for metaheuristics through the IloMetaHeuristic class. We can use a metaheuristic provided in Solver to get around the problem that none of the two available moves reduce the cost. To do this, we allow cost-degrading moves to be accepted. A metaheuristic that allows such moves to be accepted is simulated annealing.

We use simulated annealing in our small example to attempt to move the search from a start state of 11 to a final state of 00.