Synthetic Genes for Artificial Ants. Diversity in Ant Colony Optimization Algorithms

Abstract

Inspired from the fact that the real world ants from within a colony are not clones (although they may look alike, they are different from one another), in this paper, the authors are presenting an adapted ant colony optimisation (ACO) algorithm that incorporates methods and ideas from genetic algorithms (GA). Following the first (introductory) section of the paper is presented the history and the state of the art, beginning with the stigmergy and genetic concepts and ending with the latest ACO algorithm variants as multiagent systems (MAS). The rationale and the approach sections are aiming at presenting the problems with current stigmergy-based algorithms and at proposing a (possible - yet to be fully verified) solution to some of the problems ("synthetic genes" for artificial ants). A model used for validating the proposed solution is presented in the next section together with some preliminary simulation results. Some of the conclusions regarding the main subject of the paper (synthetic genes: agents within the MAS with different behaviours) that are closing the paper are: a) the convergence speed of the ACO algorithms can be improved using this approach; b) these "synthetic genes" can be easily implemented (as local variables or properties of the agents); c) the MAS is self-adapting to the specific problem that needs to be optimized.