基于退火进化算法的机器人足球策略研究

针对一般遗传算法存在的一个显著的问题：“早熟收敛”与“快速收敛”之间的矛盾，解决早熟收敛的基本思想就是保持群体中个体的多样性，而模拟退火接受准则（即Metropolis准则）可以解决这方面的问题，可以避免搜索陷入局部极值，确保找到问题的整体最优懈。因此在遗传算法中引入退火操作，定义了一个足球机器人的动作集合，根据赛场上的实际情况为足球机器人分配角色与任务，然后利用退火进化算法为足球机器人选择合适的动作，采用了一种高效的适应性评价方法，实验表明，应用退火进化算法的仿真足球机器人动作准确、效果更好。

Study on Soccer Robot Strategy Based on Annealing Evolution Algorithm

Aim at existence the prominent problem of the general genetic algorithm: inconsistency of between"precocious convergence" and "fast convergence", the basic thought that the solution precocious convergence is maintains in individual multiplicity of the community,moreover simulation annealing accepts rule (namely the Metropolis rule) can solve this problem,may avoid searching falls into the partial extremum,insure to find the overall optimal solution of the problem.So introduce annealing operation in the genetic algorithm,a set of soccer robot actions is defined,roles and tasks are assigned to every soccer robot according to the situation of the game,then make use of annealing evolution algorithm for the soccer robot to choose appropriate actions,used one highly effective adaptability estimate method.Experiments show that the soccer robots applied annealing evolution algorithm own more accurate actions than others.