Interacted Multiple Ant Colonies for Search Stagnation Problem

Ant Colony Optimization (ACO) is a successful application of swarm intelligence. ACO algorithms generate a good solution at the early stages of the algorithm execution but unfortunately let all ants speedily converge to an unimproved solution. This thesis addresses the issues associated with search stagnation problem that ACO algorithms suffer from. In particular, it proposes the use of multiple interacted ant colonies as a new algorithmic framework. The proposed framework is incorporated with necessary mechanisms that coordinate the work of these colonies to avoid stagnation situations and therefore achieve a better performance compared to one colony ant algorithm. The proposed algorithmic framework has been experimentally tested on two different NP-hard combinatorial optimization problems, namely the travelling salesman problem and the single machine total weighted tardiness problem. The experimental results show the superiority of the proposed approach than existing one colony ant algorithms like the ant colony system and max-min ant system. An analysis study of the stagnation behaviour shows that the proposed algorithmic framework suffers less from stagnation than other ACO algorithmic frameworks.