【作者】 甄宗雷；

【导师】 赵臣；

【作者基本信息】 天津大学 ， 机械电子工程， 2004， 硕士

【摘要】 足球机器人是当前机器人研究中的一个热点，涉及机器人学、人工智能、智能控制、计算机视觉等多个领域。足球机器人系统作为一个典型的多智能体协作系统，为机器人学、多智能体系统理论研究和技术应用提供了一个理想的实验平台。本文以基于全局视觉的足球机器人系统为研究对象，研究内容包括多智能体协作框架、差动轮式机器人路径规划与运动控制、足球机器人行为设计及实时决策推理等问题。本文首先针对基于全局视觉的足球机器人系统的特点，综合人工智能自上而下（慎思式）和自下而上（反应式）两种研究方法，提出了一种混合式足球机器人决策结构，即通过一个全局性高层决策模块来决定机器人团队当前队形、战术以及智能体与角色间的映射关系，然后将高层决策结果通知各个队员智能体；队员智能体根据高层决策模块的指示及自己的知识进行反应式的思考和行动。随后，文中建立了差动轮式足球机器人的运动学模型，并给出了一种路径规划与运动控制方法。根据文中提出的一种便于多智能体协作的体系结构，对足球机器人的进攻行为、防守行为以及各种战术行为进行了分析与设计。最后，基于战术、队形、角色等概念，对足球机器人团队协作框架进行了详细的讨论和分析，并给出了形式化描述结果。在该协作框架下，任务分解问题通过特定队形的角色集定义来完成，而任务联结问题通过特定角色的行为选择来完成。更多还原

【Abstract】 Robotic soccer is a new interdisciplinary research area, which encompassesdiverse fields such as robotics, artificial intelligence, intelligent control, computervision, machine learning and so on. Soccer robot system is a typical multi-agentcooperative system. This system provides the theoretical research of robotics,multi-agent system and relative fields a standard test-bed. Vision-Based Remote Brainless Soccer Robot System is investigated in thisthesis and several problems are researched, including multi-agent system teamworkstructure, motion control and path planning for differential drive mobile robot, soccerrobot’s behavior design, real time decision-making etc. Firstly, based on the characteristics of Vision-Based Remote Brainless SoccerRobot System, a kind of centralized-distributed soccer robot decision-making controlstructure is put forward by a high level decision module processing centralizeddecision-making and robot agent can respond to the environment under high-leveldecision module coaching, while acting autonomously. Secondly, the differential drive mobile robot kinematics model is set up and amotion control algorithm is developed, which allows a general differential-drivenmobile robot to accurately reach a target point with a desired orientation in anenvironment with multiple moving obstacles. Thirdly, a kind of agent architecture that allows agents to sense the environment,reason about, and take actions in the real world is proposed. Meanwhile individualagent’s diverse behaviors are designed and implemented, such as attacking behavior,defending behavior, tactical behavior. Finally, based on concept of tactics, formations and roles, the teamworkarchitecture for multi-agent system is proposed. Collaboration between agents throughthe introduction of tactics has been achieved. Tactics is made up of a series offormations, and the task space is decomposed through defining a set of roles.Homogeneous agents can flexibly switch roles within a formation, and team canchange formations dynamically, according to judgement of current conditions.更多还原