Solution bias in ant colony optimisation: Lessons for selecting pheromone models