Pre-scheduled colony size variation in dynamic environments

Abstract

The performance of the MAX -MIN ant system (MMAS) in dynamic optimization problems (DOPs) is sensitive to the colony size. In particular, a large colony size may waste computational resources whereas a small colony size may restrict the searching capabilities of the algorithm. There is a trade off in the behaviour of the algorithm between the early and later stages of the optimization process. A smaller colony size leads to better performance on shorter runs whereas a larger colony size leads to better performance on longer runs. In this paper, pre-scheduling of varying the colony size of MMAS is investigated in dynamic environments.