An Object-Oriented Framework with Corresponding Graphical User Interface for Developing Ant Colony Optimization Based Algorithms

This paper describes GRAF-ANT (Graphical Framework for Ant Colony Optimization), an objectoriented C# framework for developing ant colony systems that we have developed. While developing this framework, abstractions that are necessary for ant colony optimization algorithms were analyzed, as well as the features that their implementing classes should have. During creation of these classes, several problems were solved: implementation of individual ants and ant colonies, connection between visualization and problem spaces, creation of a multithread application in which multiple ant colonies can communicate, creation of a problem independent graphical user interface (GUI), establishing an opportunity for hybridization of ACO (Ant colony optimization). Effects of this hybridization to different variations of ant colony systems is analyzed. The use of the GRAF-ANT and its suitability is illustrated by few instances of the Traveling Salesman Problem (TSP). We also present a concept of escaping ACO stagnation in local optima, named suspicious path destruction, that is also a part of GRAF-ANT. Key-Words: Ant colony system, Evolutionary computing, Combinatorial Optimization, Swarm Intelligence