Constraint directed variable neighbourhood search

Local search algorithms operate by making small changes to candidate solutions with the aim of reaching new and improved solutions. The problem is that often the search will become trapped at sub optimal states from where there are no improving neighbours. Much research has gone into creating schemes to avoid these local optima and various strategies exist mainly based around altering the acceptance function. Another approach is Variable Neighbourhood Search which aims to bypass optima by linearly switching through multiple search neighbourhoods. We propose a new method where the selection of neighbourhoods is dynamically decided dependant on the violations of the problem constraints, Constraint Directed Variable Neighbourhood Search. We compared Constraint Directed Variable Neighbourhood Search to Variable
Neighbourhood Search and show that the same search progress can be
achieved whilst exploring only a fraction of the states.