Improved Ant Colony Algorithm And Its Application In The Continuous Space Optimization Problems

Ant colony algorithm (ACA) has emerged recently as a new meta-heuristic which belong to the class of problem-solving strategies derived from nature. It is based on the research on the ant finding the shortest road from the nest to the destination. It simulate the cooperating process which ant colony search route from nest to food. The characteristic of parallel, positive feedback and robust is strongly showed in the running of ACA. It has showed a great deal of salient character and performed great value in its application special for solving the combination-optimize questions.Since the ant colony algorithm has very good performance in the solution of discrete optimization question, then people naturally think that take the algorithm to apply in continuous optimization whether also can obtain the good result. Many scholars do the research of this aspect, but the algorithm has a particularity that it is stem from the discrete optimization question. So it must carry on certain modification before the solution of continuous problem. They introduce the idea of other methods to change the continuous space into discrete space. The findings show that this method applies in continuous function optimization also to wait for further development, this paper is precisely in view of this sufficiency, carries on the research of continuous function optimization by ant colony algorithm, proposes some new improvement measure, causes the ant colony algorithm to more adapt the continuous function optimization.First: Imitates the real ant wok mechanism, divides the artificial ant into two types: reconnaissance ant and exploration ant. Two kinds of ant complete respectively the different task with different method, but also mutually cooperate to complete the optimization goal together. The reconnaissance ants use rapid iterative searching and simple pheromone communication scheme to effectually carry on the reconnaissance wok on the continuous function space. The exploration ants carry on the stochastic optimization on the base of a pheromone element which is formed by the reconnaissance ants. This is determinism and stochastic disposition union. This method not only guarantees the ant move to the spot of high pheromone density, but also guarantees the stochastic process. The algorithm doesn't sedulously change the continuous into discrete space, the ants can move freely in the optimization space. Only the every time optimization result and the pheromone are preserved by discrete space. The result of experiment showed that the improved algorithm has better performance of function optimization than other continuous function optimization ant colony algorithm.Second: The improved algorithm has more parameter, so according to other scholar's the experience of using the Genetic Algorithm to optimize the parameters of this improved algorithm. But uses some improved strategies, such as real number code,arithmetic crossover,non-uniform mutation, to carry on optimization. This method can find a group of better parameters to make the improved ant colony algorithm had better performance.The paper uses the improved ant colony algorithm to solve the problem of parameter's optimization design of PID controller. The result of simulation experiment show that the improved algorithm has better performance, and manifested the value of the ant colony algorithm in the practice.Third: The improved algorithm has unsatisfactory performance when optimizing the higher dimensional space function. So in light of this problem, this paper uses the artificial fish-swarm algorithm to previously optimize the higher dimensional space function. The solution domain of the artificial fish-swarm algorithm can become the solution space of the improved ant colony algorithm. The simulation experiment indicated that this kind of improvement measure can improve the performance of the higher dimensional space function optimization of the ant colony algorithm. But still need further improvement to adapt the higher dimensional space function optimization.