【作者】 初旭新；

【导师】 韩光胜；

【作者基本信息】 北京工业大学 ， 检测技术与自动化装置， 2006， 硕士

【摘要】 机器人足球比赛是近年来国际上迅速开展起来的高科技对抗活动。足球机器人涉及机器人学、计算机技术、数字通讯、图像处理、传感器数据融合和人工智能等多个领域,为多智能体系统、自主机器人等理论研究以及各种相关学科的研究提供了良好的实验平台,是理论密切联系实际极富生命力的成长点。MiroSot足球机器人系统由视觉子系统、决策子系统、无线通讯子系统和足球机器人等四个部分组成。要想在机器人足球比赛中战胜对手,取得比赛胜利,足球机器人系统的每个部分都必须具有优良的性能,并能协调可靠地工作。本文着重对MiroSot机器人足球比赛系统的决策子系统相关技术进行研究。主要研究工作包括:(1)首先对足球机器人系统的决策特点进行分析,介绍足球机器人中决策子系统的层次结构,包括基本动作层、运动规划层、协调层。针对机器人的动作特点,设计了简单动作、复杂动作。提出守门员和定位球的设计方案。(2)针对机器人足球比赛环境的动态性和不确定性,提出一种基于模糊逻辑确定机器人防守和进攻的决策方法。将角色划分与运动规划结合在一起,运动规划效率高。讨论进攻和防守情况下的角色划分及运动规划问题。(3)提出一种基于局部加权回归预测球运动轨迹的学习方法,尤其是球与边界碰撞后的运动轨迹。对预测进行模糊调整,即根据球的状态生成机器人的动作。(4)介绍仿真足球机器人的比赛平台,提出利用黑板结构解决角色分配和机器人之间协作的方法。本研究是在北京工业大学所拥有的机器人平台上进行测试和实验的。更多还原

【Abstract】 Robot soccer competition is a High-Tech activity that is developed rapidly in recent years. Soccer robot relates to many fields such as robotics, computer, data communication, image processing, sensor and artificial intelligence etc. Therefore, it provides a favorable experimental flat-form for the research on multi-agent, robot and other related fields, and it is helpful to theory’s combining with practice.MiroSot of robot soccer system consists of four components vision subsystem, expert decision-making subsystem, wireless communication subsystem and robot. In order to conquer opponent and win in soccer robot game, each parts of soccer robot system must be excellent capability and cooperative work reliably. We emphasize our researches on some relative technologies in decision-making subsystem. The main contents and related original achievements are as follows:Firstly, this dissertation analyses decision-making character of robot soccer system and introduces arrangement of configuration which is Basic Movement Layer, Path Planning Layer and Harmonizer Layer. Simple and complex actions are designed according action characteristic of robot. Design of goalie and placement-ball is proposed.Secondly, a robot soccer strategy based on fuzzy logic is presented for the decision-making subsystem of robot in case when robot player is in dynamic environment. The strategy puts together the role assignment and motion planning to achieve high efficiency. The role assignment and motion planning of robot in offensive and defensive cases are discussed separately.Thirdly, a learning method based on Locally Weighted Regression is proposed to forecast the motion of the ball, especially after the ball bump into wall. The error is adjusted with fuzzy logic, that is, the robot moves based on the forecasted condition of the ball.Fourthly, the game platform of Middle League SimuroSot is introduced. This dissertation proposes that coordination problem among multi-robot and role assignment are resolved using blackboard.更多还原