The Research And Application Of Process Based Ant Colony Algorithm

The Ant Colony Algorithm (ACA) is a new type of simulated evolutionary algorithm in the early 90's, which absorbs the idea of the behavior of ants, by simulating the real ant colony's searching for food to complete the process of solving problems. It has some characteristics, such as systematicity,distributed computation,self-organization,positive feedback and has been widely used in communication,transportation,chemical engineering,data analysis and artificial intelligence. The ACA has many advantages, but there are also weaknesses in practice: (1) The searching efficiency is rather low in the initial phase of the implementation. Additionally, the pheromones distribute randomly and the paths exhibit chaotic. (2) The process is easy to get stuck into local optimal solution and involved into premature and stagnation. (3) It is difficult to solve continuous domain function optimization problems.Systematic analysis and research of the ACA was described in this paper. Aimed at the disadvantages of the ACA, the process optimized based ACA was also proposed, the key of which was optimization at each stage of the process. Detailed improvements included:Firstly, adopting other optimized design methods in the implementation, such as orthogonal design,uniform design and creating orthogonal discrete process.1. Orthogonal discrete process optimization based ACA effectively improved pheromone distribution and optimized the initial stage through orthogonal discretion, additionally utilized strategies such as dynamic transition probability to arrange and choose paths, which greatly improved the performance of the algorithm. Formula design mathematic model was further created. Tested in pigeon fodder formula design, it greatly improved performance efficiency better than traditional optimization methods, as well as found out a new way of the ACA in solving continuous domain problems.2. Uniform discrete process optimization based ACA generated local and global optimal solutions, through depth-uniform discrete sub-populations search and global uniform discrete search process, and overcame the low level and less flexibility of orthogonal discretion. Algorithm performance results showed that the uniform discrete process optimization based ACA had good optimization values and stable optimization performance. It also showed certain advantages on the number of iterations and execution time, which added useful complementation to the orthogonal discrete process optimization based ACA.Secondly, creating a hybrid algorithm of the Shuffled Frog Leaping Algorithm (SFLA) fused with the ACA.Integration of the SFLA and ACA obtained the best initial global solution group with the SFLA at an early stage and produced a precise solution search with the ACA at a later stage, which effectively solved the problems of low searching precision of the SFLA and slow searching speed of the ACA. Algorithm performance test and simulating application in animated path planning has proved that this fused algorithm was an effective way in solving continuous domain problems.Compared with the traditonal ACA, through the above changes, the process optimized based ACA achieved an improved implementation efficiency and optimization-searching ability. It also provided a valuable and practical model and problem-solving method in addressing continuous domain problems.