【作者】 于海燕；

【导师】 郭宝龙；

【作者基本信息】 西安电子科技大学 ， 控制理论与控制工程， 2010， 硕士

【摘要】 足球机器人涉及到智能控制、数据融合、计算机技术、无线通讯、图像工程、微机电系统、电气传动等多个研究领域,为控制理论研究提供了一个较好的实验平台,并推动了机器人技术的迅速发展和多智能体的深入研究。本文以RoboCup足球机器人比赛为背景,以DF-II(东风II代)全自主足球机器人为平台,对足球机器人的图像处理、目标预测和协调决策三个关键技术进行了研究。(1)介绍了DF-II全自主足球机器人平台的构成,给出了视觉子系统、决策子系统、执行子系统和通讯子系统的体系结构及工作原理。(2)研究了视觉系统的图像平滑及图像分割技术。采用高斯滤波对视觉系统采集图像进行图像平滑,采用基于YUV颜色模型的图像分割算法进行图像分割,并分别进行了仿真实验,为机器人下一步的目标预测和决策方案的制定提供可靠信息。(3)在分析比较几种目标预测算法的基础上,提出了采用基于自适应强跟踪滤波的目标预测算法,并与卡尔曼滤波算法进行了比较,仿真实验验证了算法的有效性。(4)在前期图像处理和目标预测的基础上,提出了基于无线网络的协调决策方案,系统采用双层决策模型,通过对场上机器人进行静态角色分配和动态角色转换,灵活应对比赛环境的变化,采用仿真软件AI-RCJ4.0实验平台,通过改变足球机器人比赛中团队决策和队员的个体决策验证了协调决策方法的有效性。更多还原

【Abstract】 The soccer robot is involved in many research fields, including intelligent control, data fusion, computer technology, wireless communication, image engineering, micro electromechanical system and electrical driver and so on. It provides a good test platform for the control theory and promotes the rapid development of robot technology and embedded research on Multi-Agent System.Based on the platform of“DF-Ⅱ”autonomous soccer robot, three key technologies are studied in the background of Robocup, including image processing, object prediction and decision of the autonomous soccer robot system.(1) The composition of“DF-Ⅱ”autonomous soccer robot is introduced. The working principle and architecture of subsystems are presented, including vision subsystem ,decision-making subsystem, wireless communication subsystem, and executive subsystem.(2) The technologies of image smoothing and image segmentation of vision system are studied. The Gauss Filtering is adopted to accomplish image smoothing and the algorithm based on YUY color model is adopted to accomplish image segmentation. The simulation experiments are processed so as to provide reliable information for making projects of the robot next object prediction and decision.(3) On the basis of analysis and comparison of some object prediction algorithms, an algorithm based on self-adapting strong tracking filtering is brought forward. The simulation results from the algorithm are effective to object prediction by comparing Kalman filtering algorithm.(4) According to the research of image processing and object predication, the coordination decision project based on wireless network is proposed. The double-layer decision model is used to neatly accommodate the change of the game environment through the static role assignment and dynamic role conversion of robots. The coordination decision method is validated to be effective through changing group decision and individual decision on the robot game by AI-RCJ4.0 simulation software.更多还原