Using global constraints for local search

Conventional ways of using local search are di cult to generalize Increased e ciency is the only goal generality often being disregarded This is manifested in the highly monolithic encodings of complex problems and the application of highly speci c satisfaction methods Other approaches take the general constraint programming framework as a starting point and try to introduce local search methods for constraint satisfaction These methods frequently fail because they have only a very limited view of the unknown search space structure The present paper attempts to overcome the drawbacks of these two ap proaches by using global constraints The use of global constraints for local search allows us to revise a current state on a more global level with domain speci c knowledge while preserving features like reusability and maintenance The proposed strategy is demonstrated on a dynamic job shop scheduling prob lem Introduction The use of local search has become very popular in recent years as applications have begun to tackle complex real world optimization problems for which complete re nement search methods are still not powerful enough Problem domains with nontrivial optimization functions large problems short computation time limits partial constraint satisfaction dynamic constraint satisfaction and only limited demands on optimality are promising candidates for local search Local search approaches perform a search by iteratively changing an initial as signment of variables In each iteration a neighborhood of potential successor states is generated The quality of the neighborhood states can be computed by a cost function This information is used by the successor choice criterion to determine the successor state Conventional ways of using local search are di cult to generalize Increased e ciency is the only goal generality often being disregarded It is quite common to de ne highly sophisticated and problem tailored representations with special ized neighborhoods and successor selection methods see Aarts for examples Mathematics Subject Classi cation Primary C Secondary T This work is supported by the German Research Foundation DFG NICOSIO Cross Plat form Research Germany CPR and Conitec Datensysteme GmbH c American Mathematical Society