Improvement And Application Study Of Ant Colony Optimizationm Algorithm

When people had serious trouble in solving complex Combination Optimization Problem apliying some traditional methods,some researchers had a fresher way and constructed some new algorithms by imitating ecological mechanisms of organism or ecosystem.These new algorithms include Evolutionary Algorithm(Genetic Algorithm,Genetic Programming,Evolution Strategy),Artifical Neural Network,Artificial Immune Systems,Particle Swarm Optimization, and so on.We generally call they Intelligent Optimization Algorithms.After nearly half a centruy of practice,It is proven that Intelligent Optimization Algorithms have lots of advantages that other traditional algorithms have not when they are used to solve some complex Combination Optimization Prolem and Control Problem.These advantages include good robustness,good global property of search,no details of issues themselves,and so on.Ant Colony Algorithm(ACO) were presented by Italy's scholar M.Dorigo et al in 1991.It has some advantages that Intelligent Optimization Algorithms have,and it also have some immanent disadvantage, which converges too slowly and falls into local extrema easily and so on.Based on the above disadvantages,the following parts will be researched on this paper.(1) Directed towards the issue that parameter values have a great effect on the overall performance of ACO when to implement it,we make an experimental study on reasonable parameter value range.By means of making other parameter value remain unchanged,changing only a parameter value,we obtain best parameter value that can make ACO's performance superior.(2) In this paper,we analyzed and proved the convergence of ACO simply and obtain several methods by which the overall performance of ACO could be improved.For example,adjusting the update mode of Pheromone,adjusting the mode that ant select paht.(3) In algorithm performance,the value of coefficientρhas a great effect on ACO's searching ability.To improve the searching ability,we introduce the conception of information entropy.Based on information entropy,we present a new improved ACO-Information Entropy-based Adaptive Ant Colony Algorithm.By applying it to resolve TSP problem,we prove the new algorithm is more effective than the original algorithm.(4) We present a greedy ant colony algorithm(GAOA) and use it to resolve the Agent Coalition Generation Problem.By contradistinctive experiment,we prove it is feasible and more effective than some other algorithm.The new algorithm provide a new way for resovling Agent Coalition Generation Problem.