In order to solve the problem of robot self-localization in structure environment, a fast and robust approach for self-localization implemented in robotic soccer was presented.In the approach, the probability of robot localization problem was converted to a minimum value in the range of a limited searching space.To achieve high searching efficiency and avoid the local optimum of particle filter method, a self-adaptive search algorithm was provided.The experiment results show that the localization method proposed has the advantages of real time and precision and is superior to other current robot localization methods.