Ant colony optimization with adaptive heuristics design

Abstract

Heuristics design, including definitions of heuristic information and parameter settings that control the impact of heuristic information, has significant influence on the performance of ant colony optimization (ACO) algorithms. However, in complex real-world problems, it is difficult or even impossible to find one heuristics design that suits all problem instances. Besides, static heuristics design biases ACO to search certain areas of the solution space constantly, which makes ACO less explorative and increases the risk of prematurity. This paper proposes a heuristics design adaptation scheme (HDAS) for addressing the above problems in ACO. With HDAS, each ant defines a profile of heuristics design to guide its solution construction procedure. Such profiles are adaptively adjusted towards the most suitable heuristic design according to the search experience of ants.The ACO with HDAS (HDA-ACO) is validated on a set of benchmarks of flexible job-shop scheduling problems (FJSP). Experimental results show that the HDA-ACO outperforms the original ACO.