LNS is a framework that combines the search power of CP with the scaling performance of local search. As in local search, we modify an existing solution to the problem. However, instead of making small changes to a solution, as is typical with local search move operators, we select a subset of variables from the problem, Once variables are selected, we unassigned them, lock the remaining variables to take the values assigned in the current solution, and them search for a better solution by reassigning only the unassigned variables.