The Application Of EC To Solve Search & Optimization Problems

Search & Optimization Problems are problems with very old historical background, very wide domain of application, and very hard difficulty. Search & Optimization Problems widely be used in machinery, chemical industry, administration, computer, economy, biology and military. Search & Optimization Problems proved be very important both in theory and application.By the means of simulating evolution of the nature, especially evolution of living things, evolutionary computation (EC) is a wonderful computation model to solve complex problems, Evolutionary computation not only has feature of auto-study, auto-organized, auto-adapt, but also, it has the advantage of simple, universal, and suit parallel press. Evolutionary computation is widely used in parallel search, associative memories, model reorganizations, acquire knowledge self, and so on. To solve some search & optimization problems, we put a deep study and discussion about generic algorithm, particle swarm optimization algorithm, and simulated annealing algorithm. About how to solve search & optimization problems, we proposed our methods and algorithms. We used our methods and algorithms in these problems: multivariate function optimization, minimal area circumscribing quadrilateral of convex n-polygons, TSP, Dynamic TSP, MINLP, soccer robot action selection algorithm. We propose the concept of Dynamic Traveling Salesman Problem (Dynamic TSP). According to the characteristic of Dynamic TSP, We used a hybrid algorithm of N-OPT (2-OPT, 3-OPT) and GA to solve it. We designed and realized the algorithm. The productivity and quality of solutions are evaluated by the experiment.Proposed and designed a new PSO algorithm for optimization of MINLP problems. In the algorithm, a migration tactics and a model of "backup particles warehouse" were proposed and the speed tactics was replaced by a new one. Experiments show that the new algorithm has better results and more effective than others.Some properties of minimal area circumscribing quadrilateral of convex n-polygons are proposed and proved. On the base of that, an algorithm for searching minimal area circumscribing quadrilateral of a convex n-polygon is designed. A soccer robot action selection algorithm based on PSO is proposed. In the algorithm, a set of soccer robot action is defined, and every soccer robot is assigned a role and a task according to the situation of the game, and then, using the PSO to search suitable action for every soccer robot. Experiments show that the soccer robot applied new algorithm is more flexible and has better results than others.