Boosting ACO with a Preprocessing Step

  • Authors:

  • Christine Solnon

  • Affiliations:

  • -

  • Venue:
  • Proceedings of the Applications of Evolutionary Computing on EvoWorkshops 2002: EvoCOP, EvoIASP, EvoSTIM/EvoPLAN
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

When solving a combinatorial optimization problem with the Ant Colony Optimization (ACO) metaheuristic, one usually has to find a compromise between guiding or diversifying the search. Indeed, ACO uses pheromone to attract ants. When increasing the sensibility of ants to pheromone, they converge quicker towards a solution but, as a counterpart, they usually find worse solutions. In this paper, we first study the influence of ACO parameters on the exploratory ability of ants. We then study the evolution of the impact of pheromone during the solution process with respect to its cost's management. We finally propose to introduce a preprocessing step that actually favors a larger exploration of the search space at the beginning of the search at low cost. We illustrate our approach on Ant-Solver, an ACO algorithm that has been designed to solve Constraint Satisfaction Problems, and we show on random binary problems that it allows to find better solutions more than twice quicker.