ACO - Ant Colony Optimization

V naravi so mravlje sposobne najti najkrajšo pot od vira hrane do gnezda brez uporabe vizualnih informacij. Poleg tega so se zmožne prilagoditi spremembam v okolju. na primer najti novo naj krajšo pot. ko trenutno pot preseka ovira. Pri tem nastane zamisel, da bi lahko bilo posnemanje takšnega obnašanja mravelj učinkovito tudi v diskretnem svetu. V članku bomo prikazali reševanje problema trgovskega potnika s pomočjo optimizacije mravelj.Ant colony optimization is a relatively new approach to solving NP-Hard problems. It is based on the behavior of real ants, which always find the shortest path between their nest and a food source. Such behavior can be transferred into the discrcte world, were real ants are replaced by simple agents. Such simple agents are placed into the environment where different combinatorial problems can be solved In this paper we describe an artificial ant colony capable of solving the travelling salesman problem (TSP). Artificial ants successively generate shorter feasible tours by using information accumulated in the form of a phermone trail deposited on edges of the TSP graph [1]. The basic ant behavior can be improved by adding heuristic information, e.g. local search. We describe several different algorithms used in solving the TSP (and similar) problems. We start from the first algorithm that was first used in ant optimization named Ant System. This algorithm has been followed by many others approaches resulting in better performance of ant colony optimization. The main job is to test the ant behavior on different graphs, taken from the TSPLlJJ95 library. At the end we show a comparison of ant algorithms on several instances of TSP