Positioning Model Of Local Vision In RoboCup Middle-Size Soccer Robot And Parameter Identification

Robot soccer is a standard platform of researching artificial intelligence, robot and the related problem. Robot soccer is used to verify theory algorithm and technology by the form of match. RoboCup middle-size soccer robot is a kind of many typical competition items. It provides confrontation environment of multi-robots with distribution, local sensing and narrow communications.The vision system of RoboCup middle-size soccer robot, which is consisted of local vision and omnidirectional vision system, is an important part of soccer robot system. The main task of vision system is to percept the information of surrounding environment, obtain vision images to analysis, deal with and understand. It solves the problems of self-positioning, mutual positioning and behavior understanding of opponent and teammate.The local vision system of Mid-size soccer robot in RoboCup is monocular and local system. The local vision system can get precise location information in its scope. So the local vision is often used for précis positioning of ball and robots'self-positioning. In the process of robots'self-positioning, it needs identification on gate and gatepost. Because gate and gatepost are three-dimensional, so it need remodeling replace of the plane model which is not fit for them.On the base of local vision system's study, we developed a model of local vision location. When modeling, it is proposed that local camera is ideal linear, used pinhole imaging theory to model, don't consider its distortion. According to the identification of model parameter, the Paper compares several solutions, then we used improved genetic algorithm (IGA) to identify the model's parameters. We use the model in the basic movement and the global location of soccer robot, and gain a good result. Compared with traditional genetic algorithm, the improved genetic algorithm (IGA) has many improvements, like: the hamming distance, the variable precision cross operation, the orthogonal test design, the hybrid mutation, the feedback mutation, the dynamic encoding and so on.At last, experiment of comparison and analysis of data explain advantage of the model and identification algorithm. The competition system which has used the model proves its effectiveness in the real competition.