论文部分内容阅读
为试验和比较各种先进的群体Agent合作求解智能算法,给群体Agent合作策略提供一个比较与测试的平台,该文针对传统的测试平台过分强调具体动作实施细节,忽略合作策略重要性的问题,给出了一种基于信念愿望意图(BDI)逻辑推理的群体Agent合作求解仿真系统——MAS-Soccer。在设计过程中简化Agent具体行为的执行细节,突出整体合作策略以及BDI逻辑推理在比赛中的重要性,并以此作为测试平台,以机器人足球赛任意球战术配合为实验内容,验证和比较了基于特征向量提取的再励学习算法与传统的再励学习算法在学习效果以及学习时间上的优劣。实验在验证算法的先进性的同时,也表明MAS-Soccer测试床能够准确和方便地进行合作策略的验证与比较。
In order to test and compare the intelligence algorithms of all kinds of advanced cooperative agents, this paper provides a comparison and testing platform for the cooperation strategy of group Agent. This paper focuses on the traditional test platform too much emphasis on the implementation details of specific actions, ignoring the importance of cooperation strategy, This paper presents a MAS-Soccer simulation solution based on belief reasoning intension (BDI) logical reasoning. In the process of design, the implementation details of Agent’s concrete behavior are simplified, and the importance of overall cooperation strategy and BDI’s logical reasoning are highlighted. The test platform is used as the test platform, and the robot soccer matches the free-kick tactics are taken as experimental contents to verify and compare Re-excitation algorithm based on eigenvector extraction and traditional re-learning algorithm in learning effectiveness and learning time advantages and disadvantages. While verifying the advanced nature of the algorithm, the experiment also shows that the MAS-Soccer testbed can verify and compare the cooperation strategy exactly and conveniently.