【作者】 吴杰；

【导师】 许映秋； 徐洪彬；

【作者基本信息】 东南大学 ， 机械工程（专业学位）， 2016， 硕士

【摘要】 机器人救援仿真系统(RoboCup Rescue Simulation System)是典型的多智能体系统,它为促进多智能体技术、人工智能以及机器人等相关领域的研究提供了广泛的技术支持,在理论研究和实际应用方面都有巨大的价值。而警察智能体作为救援仿真系统中重要的智能体之一,其清障行为是整个救援团队高效运转的基础,对警察智能体的行为进行全面深入的研究就显得尤为重要。论文针对警察智能体的高效行为展开研究并提出一系列实现方案,具体工作如下：(1)警察智能体的行为研究对警察智能体的行为进行了分析和定义,将警察的行为归类成决策层的策略行为和执行层的动作行为,同时从面向任务的角度出发,将警察的行为分为任务获取、任务分配和任务执行三个过程。另外,从合理性、有效性和通畅性三个方面描述警察智能体行为高效性的特征。(2)警察智能体的人员分布策略根据救援仿真系统及警察智能体的特点,本文针对警察智能体的人员分布策略提出一种基于K-means算法的警察分区策略,适应不同情况下警力的有效分配实现。(3)警察智能体的任务分配策略根据救援仿真系统中智能体的通信情况,把警察智能体的任务分配分为两类：无中心智能体情境下的分布式任务分配和有中心智能体情景下的集中式任务分配;分别提出了基于任务优先级的任务分配策略和基于Hungarian算法的任务分配策略。(4)构建警察智能体的清障动作模型警察智能体的清障动作是贯彻其策略行为的保证,是实现警察智能体高效行为的基础。根据新的清障方式的特点,构建了“引导线”模型,该模型可使警察智能体沿着引导线清障前进,所走轨迹光滑平整,所清道路平滑顺畅。更多还原

【Abstract】 RoboCup Rescue simulation system is a typical multi-agent system, which provides a wide range of technical support to promote the research of AI (Artificial Intelligence), robotics and related fields as well as multi-agent collaborative. It is of great value both in theoretical research and practical application. As an important agent of rescue simulation system, the police force agent’s mission- clearing blockades is critical in the early simulation. So, it is particularly important to deeply study the behavior of the policeforce agent.The main jobs are as follows:(1) Research of the Policeforce Agents’BehaviorFirst, the behavior of the policeforce agents is analyzed and defined, it is categorized into strategy behavior in decision-making layer and action behavior in execution layer. And these two behaviors are introduced in detail. In addition, this paper deeply analysis the characteristics of the efficiency of the policeforce agent, and the relationship between the efficiency of the policeforce agent behavior and the methods to achieve the efficiency.(2) Distribution strategy of the policeAccording to the characteristics of rescue simulation system and policeforce agent, this paper proposes a partitioning strategy based on K-means algorithm. It is a distribution strategy of the police. The artical describes the principles and the application process of K-means algorithm.(3) Task allocation algorithm of the policeRegarding of the Rescue Simulation System communication situation, the allocation is defined in two categories:distributed task allocation in scenarios without agent center and centralized task allocation in scenarios with agent center and enough communication channels. This paper uses the strategy based on task priority and the strategy based on Hungarian algorithm.(4) Construction a clearing modelThe clearing action ensures the strategy and is the basis of efficient behavior for policeforce agent. According to the characteristics of the new clearing method, this paper builds a clearing model called "GuideLine". The police clears blockades and moves along the guideline, so that the path it traversed is smooth and clean.更多还原