The Optimization Of Ant Colony Algorithm And Its Application

Ant Colony Algorithm is a kind of bionics algorithm, proposed by Drigo in 1991. The Ant Colony Algorithm is derived in ant colony foraging behavior, when ants search for food, walking in front of ants release pheromone, and walking in back of ants accord to the concentration of pheromone as a basis, also release the pheromone. After a large number of ants repeated walking, a high concentration of pheromone path will appear, the path is the shortest path from the ant nest to the food source. The application of Ant Colony Algorithm has become one of the hot spots in distributed artificial intelligence research, and many enterprise operation mode is applied to the Ant Colony Algorithm. Ant Colony Algorithm is easy to combine with other methods, it has strong robustness, but, it runs long time, it is easy to fail into local optimum. Analyzing the shortcomings of the algorithm, and the following points are studied in is paper:1. Ant Colony Algorithm running time is long. When the maximum cycle is relatively large, after iteration to a certain number times, no matter how many times iterations, the solution is the same as the before optimal solution. In this paper, to shorten the running time of the algorithm by using the task termination strategy in advance.2. In the basic Ant Colony Algorithm, the next path of the ant's choice is to obtain the maximum value of the probability. But the study found that ants traverse the next path is not necessarily obtained the maximum probability path, but relatively large probability of collection. In this paper, Ant Colony Algorithm joins the roulette algorithm, in order to get a better solution, this paper based on the basic Ant Colony Algorithm to add the roulette algorithm.3. The pheromone evaporation coefficient has a direct impact on the search ability and convergence speed of ant colony algorithm. In this paper, a dynamic adaptive strategy is used to assign a larger initial value, each iteration decreases its value dynamically. The constant reduction of the pheromone volatilization coefficient can effectively avoid falling into local convergence.