Search Space Reduction as a Tool for Achieving Intensification and Diversification in Ant Colony Optimisation

The aim of adding explicit intensification/diversification measures to ant colony optimisation is so that it better samples the search space. A new and novel method of achieving this is based on the idea of search space reduction in which solution components are fixed during an intensification stage and certain values for some components are excluded during diversification. These phases are triggered as required throughout the search process. In comparison to an existing intensification/diversification scheme and a control ant colony solver, the results using the travelling salesman problem reveal that the new scheme offers a substantial improvement. It often achieves optimal results for benchmark problem instances. Additionally, this scheme requires few extra computational resources and is applicable to a range of combinatorial problems.