Real-time autonomous distributed cooperation under dynamically changing environments

(1)开发了一种可以通过视觉观察来检测对手能力的足球代理。我们开发了各种短喊叫，即非常短的交流，以激活两个以上球员的较长时间的战略团队合作。在RoboCup足球模拟联赛中，我们研究并测试了如何对RoboCup足球模拟联赛中与其他球队比赛中经常观察到的不适当的、意外的代理行为进行补救。我们设计了一种机制，让另一个内部代理参与到足球代理中；也就是说，我们制作了一个代理，而这个代理又是一个多代理系统。新引入的内部代理通过参考其操作历史和当前情况来监视原始(基本)代理是否要发出不适当的操作命令。如果内部代理发现错误，它会中断基础代理更改命令，并将相关日志上报给开发人员进行调试。通过实验，我们可以为有反射的智能体和有反射的团队制定新的研究方案。(4)探索了将二维足球仿真扩展到三维足球仿真的新途径。我们的研究重点是将现有的3D代理模型改进为更自然、更逼真的3D模型，并提供一个良好的物理仿真模型。

(1) We developed a soccer agent which can detect opponent ability by visual observation. It is needed to make the agent action suitable for various type of opponent agents.(2) We developed various short shouts, namely very short communication, to activate rather a long term strategic team play of more than two players. This kind of short shouts are more advanced than those which invoke only reflective actions of other players.(3) We investigated and tested how to remedy inappropriate, unexpected agent actions which are often observed in matches with other teams in RoboCup soccer simulation league. We devised a mechanism to involve another inner agent in the soccer agent ; that is, we made an agent which is in turn a multi-agent system. Newly introduced inner agent monitors the original (base) agent whether it is to issue an inappropriate action command by referring to its action history and the current situation. If the inner agent finds something wrong, it interrupts the base agent to change the command, and reports the relevant log to the developer for debugging. Through our experiments, we could make a new research plan for agent with reflection and team with reflection.(4) We investigated a new way of the extension of 2D soccer simulation to 3D soccer simulation. Our research focused on the improvement of the current 3D agent model toward more natural and humanlike 3D model with a good physical simulation model.