Machine Learning of Multi Agents for RoboCup Soccer Domain： A Survey

Machine learning has been widely applied to deal with problems in complex environment such as RoboCup, which is assumed as the ideal platform for research on AI and robotic. In RoboCup simulation league,software agents play soccer games on an official soccer server over the network. When constructing these software agents, issues in area of agents learning techniques arise to satisfy the properties specified by agent theorists. This paper presented an overview of the agents learning used in the simulator teams. Many kinds of agents learning techniques were reported and compared. It also provided open questions for discussing and pointed out some possible answers to verify in near future.     

基于多智能主体的高速公路控制系统结构

高速公路交通量的增加对高速公路交通控制提出了更高的要求。传统的基于精确数学模型的交通控制方法很复杂并且在实际应用中往往达不到预期的控制效果，因此寻求新的有效的交通控制方法也就成为必然。多智能主体(Multi—Agent)技术是人工智能领域中的一个研究热点，单个Agent能够自主地、主动地对对象进行控制，并且Agent之间可以进行协作从而使整个系统达到更大的智能性。介绍多智能主体系统(Multi-Agent System，MAS)的概念，然后描述基于MAS的高速公路交通控制系统结构以及Agent的内部模块的功能，最后对Agent之间的协作进行了简要的分析。基于多智能主体的高速公路交通控制系统拥有一种全新的智能交通控制结构，它能够较好地克服传统交通控制系统所存在的缺陷，适应高速公路交通系统的复杂性和随机性。MAS技术的关键在Agent之间的协调，因此下一步的研究应该集中在Agent之间如何建立有效的协调机制。     

基于变精度Rough集的对象甄选优化模型

定义了对象甄选信息系统,并根据属性值序列间的距离,给出了对象的相似关系,并利用Rough集理论讨论了对象的分类问题,最后给出了甄选的优化模型.     

TOWARDS AN ONTOLOGY-BASED MULTI-AGENT MEDICAL INFORMATION SYSTEM BASED ON THE WEB

Introduction　　 Medical information systems cover a numberof medical domains to satisfy different medical ap-plications and characterized by the rich functionali-ties and valuable medical case data[1,2 ] .But infor-mation systems are heterogeneous and lack of effi-cient knowledge organization mechanisms.So theredundant information is existing,and real time in-formation acquisition and updating is difficult andmedical information systems are hard to interoper-ate and implementcomplex medical a…     

Object tracking and tracing: Hidden semi-Markov model based probabilistic location determination

The enhancement of radio frequency identification (RFID) technology to track and trace objects has attracted a lot of attention from the healthcare and the supply chain industry. However, RFID systems do not always function reliably under complex and variable deployment environment. In many cases, RFID systems provide only probabilistic observations of object states. Thus, an approach to predict, record and track real world object states based upon probabilistic RFID observations is required. Hidden Markov model (HMM) has been used in the field of probabilistic location determination. But the inherent duration probability density of a state in HMM is exponential, which may be inappropriate for modeling of object location transitions. Hence, in this paper, we put forward a hidden semi-Markov model (HSMM) based approach for probabilistic location determination. We evaluated its performance comparing with that of the HMM-based approach. The results show that the HSMM-based approach provides a more accurate determination of real world object states based on observation data.     

�������彨ģ�����ڻ�������������Ŧѡַ���Ż���Ӧ��

本文提出一种用于评价公路－铁路联合运输货运枢纽选址方案的整数模型。模型中的四个主体为：货运枢纽经营者、交通运输网络设施规划者、客户、社会公众。介绍了可使全局目标优化的基于主体建模方法(ABM)在货运枢纽选址中的应用。最后，本文以澳大利亚昆士兰州东南部的货物联合运输枢纽为例，介绍了其原始选址过程、个体目标函数测试及基于主体的模型设计过程。     

Optimizing the Location of Intermodal Freight Hubs: An Overview of the Agent Based Modelling Approach

This paper aims at developing an integral model for the evaluation of road-rail intermodal freight hub location decisions. The model comprises four dominant agents, namely, hub owners or operators; transport network infrastructure providers; hub users; and communities. An agent based modelling approach is introduced to allow such negotiation to happen to achieve a global objective. The paper outlines the methodology to be used. It also presents an initial location selection process, a testing with individual objective functions, and a design for an agent based model using a case study of intermodal freight hub location decisions in South East Queensland of Australia.     

基于遗传算法的磁性目标磁模型

磁性目标形体复杂,磁化不均匀,目前一般采用逆方法拟合测量数据,从而建立其磁模型.文中以系数矩阵的条件数为目标函数,利用遗传算法搜索磁偶极子的位置,同时监测拟合误差,使得目标函数达到最优的同时,拟合误差也满足要求.实测数据的建模结果表明,用该方法建立的磁场模型精度高,对测量误差不敏感,具有很强的鲁棒性.     

无线传感网络中QoS约束多播路由算法

分析了无线传感网络中具有时延、带宽和剩余能量等多QoS约束的多播路由问题.描述了一种适应于研究QoS多播路由的网络模型.提出了一种具有多QoS约束的能量有效的多播路由算法(EMRA),该算法成功地解决了无线传感网络中多播节点动态变化时的QoS路由选择问题.仿真实验证明了EMRA的有效性和优越性.     

双注意力引导的跨层优化交通场景语义分割

针对交通场景目标分割边缘不平滑以及小目标难以准确分割等问题，本文提出一种双注意力引导的跨层优化交通场景语义分割算法。首先，构建多分支特征提取编码网络，并利用串行非比例式空洞卷积实现空间上下文信息提取，进而改善小目标信息的丢失；其次，构建基于空间对齐的跨层特征融合解码网络，实现语义信息和细节信息的融合，增强不同尺度目标的表达能力；最后，提出通道和空间注意力机制，建模全局通道相关性和长距离位置相关性，增强网络对关键特征的学习能力。交通场景数据集Cityscapes和CamVid上的实验结果表明，所提特征提取编码网络、跨层特征融合解码网络以及注意力机制模块是有效的；所提语义分割算法获得了77.79%和78.66%的平均交并比，能够平滑目标分割边缘，尤其对细长条形目标具有鲁棒性。     

OBSTACLE-AVOIDANCE NAVIGATION OF MULTIPLE MOBILE ROBOTS

IntroductionRobo Cup,which involves multiple robots in acomplicated environment to achieve specific objec-tives,poses a common standard problem for aboard spectrum of specialize sub- fields in AI andintelligentrobotic agents[1] .The majorcharacteris-tic of the Robo Cup Soccer competition is the dy-namic nature of environment surrounding robots,which includes the problem addressed in this pa-per.In the current research of it[2 ,3 ] ,when the re-searchers think about the problem of avoiding ob…     

应用小波模历史图像的运动车辆视频检测

为提高车辆目标检测的稳定性和准确性,提出了基于背景减除和小波分解模历史图像的运动车辆检测算法.首先对原始图像进行小波分解,对低频分量用混合高斯模型和纹理特征相结合的方法,自适应更新背景并标记运动目标初始区域;然后,基于高频分量计算模值,并通过逐帧历史累积得到模历史图像;最后,利用车辆目标与阴影相比富含边缘细节的特点,对目标进行倾斜校正后,将目标边缘分别沿图像x和y方向投影,利用投影曲线将边缘信息与目标初始区域信息迭代融合,得到最终检测结果.实验结果表明,用本文方法检测车辆的捕获率达到99.0%,有效率为92.5%;与使用单一自适应背景提取方法相比,在实际交通场景中可有效处理阴影导致的多目标粘连问题,检测结果更准确.     

概念格的分层及逐层建格法

给出了分层概念格重要性质:同层概念不可比和每个概念至少被1个上层的概念覆盖.得到了概念格上的对象映射和属性映射的一般计算公式和在对象化简过程中格结构不变的定理.由上述性质和定理,给出建格算法.此方法以概念格分层为基础,逐层建格.因为在建每层格的节点的过程中同时进行筛选,所以不会生成冗余重复的格节点.     

基于UML的MAS环境下Agent的一般性分层框架

研究有时间、资源等约束的开放MAS环境下的Agent模型，提出了混合型Agent的一般性分层框架．该框架包括特性说明层、心智状态层、计算处理层和信息交互层．分析了Agent的体系结构、其在组织中承担的角色和部件组成等．采用UML技术设计Agent的概念类图，给出了Agent的形式化表达和通用程序的设计方法，并阐述了综合调度引擎、并发性消息处理的实现机制．将该框架应用于青藏铁路贯通线广域线路Agent保护研究。设计了多个Agent实现协同保护功能，解决了传统保护的误动问题，表明采用该Agent框架的设计对解决这一复杂问题是有效的．     

һ�����͵��˶�Ŀ��ʶ��������㷨�о�

提出了一种应用于智能交通监控系统的运动目标识别和跟踪方法。针对帧间差分提取运动目标的缺陷与不足,提出了一种基于冗余小波变换的运动目标识别算法,即直接在冗余小波变换域提取运动区域,从而检测出运动目标。对于检测出来的运动目标,本文对mean-shift算法进行了改进,采用自适应mean-shift算法,对目标进行跟踪。实验结果表明,本文提出的算法可以有效地提取运动目标,即使目标与背景具有较高的相似度,也可以较准确的提取出前景运动信息,效果要好于传统的帧差法;跟踪目标准确度高,不受目标大小变化的影响。本算法具较高的实用价值和应用前景。     

客运专线综合运营调度系统Agent结构和规则描述

为提高客运专线综合运营调度系统的智能性,采用智能体(Agent)技术,将运输调度决策过程看作是Agent之间的协商过程,建立了客运专线综合运营调度系统的多智能体(MAS)体系结构。将列车运行调度子系统、动车组调度子系统、牵引供电调度子系统和车站分别抽象为Agent,所有Agent构成了Agent社会,给出了Agent社会规则。在体系结构的基础上,给出了二路四层垂直型结构的公司行调Agent结构和反应式车站Agent结构,并阐述了其工作原理。仿真结果表明:在遵循社会规则下,Agent能在10s内调整列车运行计划,5s内选取调度预案,因此,MAS体系结构可行,能够提高运营调度系统的智能性。     

顾及小目标特征的视频人流量智能统计方法

人流量统计对智能安防等领域具有重要研究价值. 针对视频监控系统中人流量统计准确率较低的问题,提出一种顾及小目标特征的视频人流量智能统计方法,重点研究用于小目标检测的Faster R-CNN （Faster region-convolutional neural network）改进算法、运动目标关联匹配以及双向人流量智能统计等关键技术,根据人头目标的小尺度特点对Faster R-CNN网络结构进行适应性改进,利用图像浅层特征提高网络对于小目标的特征提取能力,通过基于轨迹预测的跟踪算法实现运动目标的实时追踪,同时设计双向人流量智能统计算法,以实现人流量高准确率统计;为证明方法的有效性,在密集程度不同的场景中进行了测试. 测试结果表明:改进的目标检测算法相较于原始算法,在Brainwash测试集和Pets2009基准数据集上的平均准确率分别提高了7.31%、10.71%,视频人流量智能统计方法在多种场景下的综合评价指标F值均能达到90.00%以上,相较于SSD-Sort算法和Yolov3-DeepSort算法,其F值提高了1.14% ~ 3.04%.      

基于Q-Learning的智能体训练

针对机器人足球比赛的多智能体环境下智能体的训练问题,提出了一种将模糊控制与Q-Learning相结合的学习方法,并在学习过程中自动调节回报函数以获得最优策略,此方法的有效性在中型组的仿真平台上得到了验证,并取得了较好效果,还可将它改进应用于其他多智体环境。     

Calculation of 3D sensitivity matrix for electromagnetic tomography system using boundary elements and the perturbative approach

Electromagnetic tomography （EMT） is a non-invasive imaging technique capable of mapping the conductivity and permeability of an object. In EMT, eddy currents are induced in the object by the activation coils, and the receiving coils can measure the EMT voltages. When the activation frequency is significantly large, we can treat the metallic targets as electrically perfect conductors （EPCs）. In this situation, a thin skin approximation is reasonable and this type of scattering problem can be effectively treated by the boundary element method （BEM） formulated through integration equations. In this study, we compute three-dimensional （3D） sensitivity matrix between the sensors due to an EPC perturbation. Efficiency improvement is achieved through the utility of scalar magnetic potential. Two EPC objects, one sphere and one cube shaped, are simulated. The results agree well with the H dot H formula. Overall, we conclude that BEM can be used to calculate the 3D sensitivity matrix of an EMT system efficiently. This method is a general one for any shaped objects while the H dot H solution is only capable of producing the response for a small ball.     

A multiscale superpixel-level salient object detection model using local-global contrast cue

The goal of salient object detection is to estimate the regions which are most likely to attract human’s visual attention. As an important image preprocessing procedure to reduce the computational complexity, salient object detection is still a challenging problem in computer vision. In this paper, we proposed a salient object detection model by integrating local and global superpixel contrast at multiple scales. Three features are computed to estimate the saliency of superpixel. Two optimization measures are utilized to refine the resulting saliency map. Extensive experiments with the state-of-the-art saliency models on four public datasets demonstrate the effectiveness of the proposed model.