请输入检索词:   智能扩展

<<上一记录 下一记录>>
ID   23140103720544 [ 公开 ]
----------
论文名称   机器人三维足球仿真比赛中多智能体及建模方法的研究/Research of MAS and Modeling Methods in RoboCup3D Simulation
----------
作者   朱磊  <中华人民共和国>
----------
学位   硕士
----------
答辩日期  
----------
培养单位   上海大学.机电工程与自动化学院.控制理论与控制工程.
----------
研究方向   人工智能
----------
中文关键词   机器人足球; 卡尔曼滤波; A*搜索; 计算动词 
英文关键词   RoboCup; Kalman Filtering; A* Search; Computational Verb 
----------
指导老师   费敏锐 .上海大学  .  .
----------
论文总页码

  共75页
----------
馆藏号

  060211
----------
服务年限

  0 年后
----------
中文摘要     RoboCup(Robot World Cup)即机器人世界杯足球锦标赛,是一种典型的多智能体系统。它融合了多个领域的技术,已经成为研究多智能系统的一个标准实验平台。在实际比赛中,需要 结合多种技术:如Agent的设计理论、多Agent协作、策略获取、实时推理等。通过具有普遍意义的试验平台,可以深入研究和评价多智能体系统中的各种 理论和算法,并将结果推广到众多领域。 本文的主要研究工作和特色是围绕RoboCup3D多智能体体系结构及其核心模块建模展开,内容如下: 1)研究RoboCup3D仿真平台的内部模型和运行机制。以RoboCup3D仿真环境为基础,以智能模型作为智能体系统的软件开发模型,运用模块化设 计思想,建立了RoboCup3D仿真智能体体系结构。 2)针对RoboCup3D仿真环境建立世界模型。首先研究了基于卡尔曼滤波的智能体自定位系统,以此为基础建立了智能体的运动子模型、球的运动子模型和 智能体的踢球子模型。同时提出了智能体世界模型预测算法。 3)采用数值计算方法建立RoboCup3D仿真队伍的直接射门模型。提出用A*搜索算法建立射门路径寻优,针对A*搜索进行效率优化,减少入队出队的内 存分配操作。另外用一数组来记录节点位置,并据此来还原数据,能达到快速建立射门路径的目的。 4)针对不同水平的球队,研究了对手建模。提出了基于计算动词理论的RoboCup3D仿真足球机器人对手建模方法。通过动词理论中模糊扩展的方法建立了 计算动词控制器,得出对手控球能力的强弱程度模型,从而实现了在线实时对对手进行能力评估。 在以上研究的基础上,本文用面向对象的方法设计了一支仿真球队,并在全国RoboCup比赛中得到验证。
 
外文摘要     The Robot World Cup ( RoboCup ) is a typical MAS ( Multi-agent System ) in which multi-field technology has been involved and RoboCup has been a standard test platform to study MAS. Many theories are necessary to build a RoboCup team, including agent designing theory, multi-agent cooperation, strategy obtaining, real-time reasoning and so on. Based on this general test platform, various theories of MAS can be researched and applied to many fields. The main research and feature of this paper are as follows. 1) The internal model and function mechanism of RoboCup3D simulation server has been studied in this paper. On the basis of knowledge of RoboCup3D simulation environment, we establish RoboCup3D multi-agent system by using agent model as the software-developing model and adopting modularized designing ideas. 2) The world model is established in RoboCup3D environment. The agent self location system based on Kalman filtering is researched first and it is the basis of studying the movement submodel of agent, the movement submodel of ball, and the kick submodel of agent. The forecasting arithmetic of world model has been presented. 3) Using numerical calculation method for direct-shoot model and putting forward A* search arithmetic to search shooting path while establishing the shoot model of RoboCup3D simulated team. To improve the efficiency of A* search in shoot model, we reduce the manipulation for memory allocation when entering or exiting queue, and use an array to record the node which can be used to rebuild the path. 4) Opponent modeling is studied to adopt rivals at different levels. Based on Computational Verb Theory, an opponent modeling method for RoboCup has been presented. Fuzzy generalization is used to design the computational verb controller, by which the model of extent of opponent’s controlling ability can be obtained to estimate opponent on-line. Based on the above research, a RoboCup3D simulation team is designed with OOP method and this team has been validated in the competition of China RoboCup .
 
原文    
 
参考文献

 

  浏览参考文献
 
<<上一记录 下一记录>>