Study On Multi-Objective Optimization Problem Based On Ant Colony Optimization Algorithm

Multi-Objective Optimization Problem a very important research topic in the field of scientific research and engineering applications. They not only have a certain academic value but also a wide range of social applications. Their results are widely used in engineering, economics, management, military and other fields.The technology of Modern intelligent bionic algorithm in multi-objective evolutionary algorithm (such as the Vector Evaluated Genetic Algorithm, and the genetic algorithm genetic algorithm, etc.) to solve multi-objective optimization problems is relatively mature. However, so far, the ant colony algorithm to solve the problem of multi-objective optimization is also very little at home and abroad, I only found in articles of [30] and [31]. Therefore, the purpose of this paper is to do some exploratory work on ant colony algorithm to solve multi-objective optimization problems.Preliminary studies have shown that ant colony algorithm has the advantages of solving complex optimization problems, especially on discrete optimization problems [4]. However, for solving continuous optimization problems and discrete domain optimization problems are very different. The technology of solving the continuous problem of ant colony algorithm is not very mature, and to solve multi-objective optimization problem with the ant colony algorithm is in its infancy.On this basis, the author of literature has also done in-depth analysis at [30] and [31]. to do the tests by using of commonly used benchmark functions by programming the algorithm, finds that the results have several shortcomings:1) only very low probabilities appeared ideal results (the poor of stability); 2) the results appear better (less efficient)after a long time (more iterations); 3) the diversity is poor; 4) the distribution is bad.In response to these problems, this exploratory study was applied to continuous multi-objective optimization problems with the ant colony algorithm. Two steps of algorithm are approached:1) Change "the constructed way of initial solution of ant colony algorithm"; 2) increase the step of" increase the initial feasible solution of ant in the iteration". Test results of two specific examples show that these efforts are fruitful. The following brief account of the next test results.The Binh problem is partial constraint problems with a two-variable. It's cutting edge is continuous, easy to solve. The results obtained with the old algorithm is missing or sparse in the Pareto front. Changing to "Uniform random generation of initial ant" algorithm greatly improves the density of the Pareto front.Difficulty of problem Tanaka is that it's cutting edge is not continuous curve, and ratio of non-convex curve. Obtained with the old algorithm is very sparse, only a few points. Improved algorithm for the Binh test gets the results better, but the Pareto front is still relatively sparse. Then increase a step. with the tests, the quality of Pareto front is greatly improved.