【作者】 陈万米；

【导师】 费敏锐； 胡豁生；

【作者基本信息】 上海大学 ， 控制理论与控制工程， 2009， 博士

【摘要】 多机器人系统是在多学科综合的基础上得到发展,涉及到计算机、通信、控制、智能机器人、多传感器及信息融合、视觉及图像处理、人工智能和仿生等相关学科。目前研究多机器人系统的单位与学者在逐年增加,多机器人系统研究水平和研究领域也在不断发展。本文构建了以小型足球机器人系统为对象的多机器人系统研究平台,并在此平台上进行了理论研究、仿真验证及比赛实践,主要取得了以下创新性的研究成果:1.提出了计算动词理论结合区域搜索方法解决小型足球机器人系统的实时图像信息获取问题,对机器人标识进行模板匹配和相似度的求解,将计算动词分解为行动词和列动词减轻系统的计算量,尤其研究了光照不均匀情况下难以识别的问题,提高了图像信息的准确性、稳定性,为路径规划等研究打下了基础。2.将卡尔曼滤波与人工神经网络理论用于小型足球机器人与球的位置预测,可以避开复杂的小型足球机器人建模过程,通过预先采集的输入、输出数据,预测出小型足球机器人的运动轨迹,为小型足球机器人的路径规划和系统决策提供了重要的支撑。3.在对小型足球机器人进行基本运动轨迹规划的基础上,提出了以Delaunay三角形网格理论指导的优化路径规划,用关键点及其姿态简化对路径的优化思想,使小型足球机器人运动规划能快速找到较优化的路径。4.提出了将蒙特卡洛理论应用于小型足球机器人系统上,实现系统的增强学习功能,包括个体能力学习和总体策略学习功能,实现了多个小型足球机器人的动态角色分配,避免了运行时的相互冲突,取得了良好配合的效果。5.研究并运用多传感器和数据融合理论,综合解决了小型足球机器人系统中遇到的小型足球机器人运动的直线校正、运动避障、运动正确性判断、“受伤球员”的智能判断等许多实际问题。以上研究成果已应用在上海大学小型足球机器人系统上,进行了实验室试验,并在国际国内RoboCup机器人足球赛事上,取得了2004--2007年RoboCup中国赛连续进入4强以及2007、2008年蝉联伊朗机器人(国际)公开赛冠军的成绩。更多还原

【Abstract】 The development of collaborative multi-robot systems is multi-disciplinary, involving computer science, communication, control theory, intelligent robots, multi-sensor technology, information fusion, computer vision, bionics, artificial intelligence, etc. Today, the number of scholars and researchers who study collaborative multi-robot systems is increasing rapidly. The research level and scope of collaborative multi-robot systems are also continuously increasing.In this thesis, small-size robot soccer competition has been as a research platform to conduct the fundamental research, simulation verification and practice competition of a collaborative multi-robot system.The major contributions of the thesis include:1. The computation verb theory is combined with the region reconnaissance method to solve the difficulty associated with real-time image acquisition in the small-size soccer robot system. Through matching robot marks, resolving similarity, and decomposing computation verb into a‘low verb’and a‘column verb’, the real-time system computation, as well as the accuracy and stability of image information acquisition under the non-uniform illumination condition are greatly improved, laid a solid foundation for path planning research.2. Based on the study of motion trajectory of small size soccer competition, the Delaunay triangle grid is used to optimise path scheduling. The key points and their simplified postures are adopted to optimize the robot path so that the small-size robots are able to plan their paths with a high degree of system stability and fast speeds.3. The Extended Kalman-Bucy Filter and the RPF neural network are used in small-size soccer robots and their position prediction so that the complexity of modelling small size soccer robots is greatly reduced. Through the pre-collected input and output data, motion trajectories of small size soccer robots are predicted. This provides an important foundation of path planning and system decision of small size soccer robots.4. The Monte-Carlo theory is applied to the small-size robot soccer system to study its reinforcement learning, including the overall strategy learning and the individual ability learning. As a result, it has realized many small soccer robot dynamic role assignments, avoided the movement mutual conflict, and obtained the good coordinate effect.5. Data fusion theory is investigated to solve many practical problems, such as The straight line adjusts, the movement evades bonds, the movement accuracy judgment, the intelligent judgment of“injured player”in the multi-robot system comprehensively based on the foundation of the multi-sensor system.The research results in this thesis have applied to the small-size robot system of Shanghai University, including many laboratory experiments, participations of many domestic and international robot-soccer competitions. The team continuous entered top 4 in 2004-2007 RoboCup China, became the league champion in 2007 RoboCup Iran Open, as well as the league champion in 2008 RoboCup Iran Open.更多还原