一种分布式检测系统的递推反馈自适应学习算法

分布式多传感器检测系统中的等概率假设在检测概率未知及时变情况下不能得到最优的检测状态．研究分布式检测系统的最优检测问题,考虑传感器虚警与漏报的概率未知,且概率不相等的情况,提出了一种递推的状态反馈自适应学习算法,通过在线的修正融合权值,最终使系统收敛于最佳权值,并对算法收敛性和误差进行了理论分析．仿真研究了概率未知、环境时变等情况下的算法性能,验证了所提方法的有效性．     

一种基于信息融合的安全信息管理模型

文章针对当前各种网络安全设备独立分散地工作带来的问题提出了一种安全信息管理模型,采用信息融合的概念和方法从误用检测和异常检测两个过程对信息进行处理.最后,利用安全信息关联和融合的结果进行量化的实时安全风险评估.     

目标识别中的信息融合技术

在信号检测中,单一传感器只能反映被测对象某一方面的特性,信息融合可融合多传感器信息,获取对对象较全面的认识。比较全面地阐述了目标识别中信息融合技术的结构和方法,分析了各种方法的优点及缺陷,提出了信息融合发展的一些问题。     

基于信息融合入侵检测系统的体系结构研究

从用户的角度描述了现有IDS存在的虚警太多的问题．提出用信息融合的方法解决这个问题。为此设计了一个基于信息融合技术的入侵检测系统的体系结构。     

信息融合技术在入侵检测系统中的应用

本文将入侵检测技术和信息融合技术相结合,针对大型异构网络提出了基于Internet的IDS即CyberIDS的概念,给出了CyberIDS的体系结构和相关融合问题,并提出了用于实现CyberIDS的相关融合算法,揭示了新一代IDS的发展趋势和特点.     

多传感器系统的延长周期全信息融合滤波

提出了基于系统状态预测信息、子系统状态预测信息及量测信息的多传感器系统全信息融合滤波方法.针对多传感器系统中子系统状态估计信息与系统状态预测信息之间存在相关性的问题,采用延长系统融合器融合周期的方法来弱化它们之间的相关性,再进行直接融合来获得系统状态估计.基于"GPS姿态敏感器+陀螺"模式组合的卫星姿态确定系统的仿真结果表明了延长周期全信息融合滤波方法的有效性.     

神经网络集成、多传感器融合——在机器人对障碍物的识别中的应用

近年来，基于传感器的机器人研究成为开发智能机器人的一个重要方面，它极大地改善了机器人的作业能力，具有重大的应用价值和发展前景。多传感器融合技术就是对同一检测对象，利用各种传感器检测的信息和不同的处理方法以获得该对象的全面检测信息，从而提高检测精度和可靠性在多传感器系统中，信息表现为多样性、复杂性以及大容量，信息处理不同于单一的传感检测处理技术，多传感器信息融合技术已成为当前的一个重要研究领域。目前信息融合方法利用多个信息源所获取的关于对象和环境的信息获得根据任务所需要的全面、完整的信息，主要体现在融合算法上。因此，多源信息融合的核心问题是选择合适的信息融合算法。     

多传感信息融合与自动化

综述多传感信息融合技术的概念、发展的起因、主要研究内容、现有的理论和方法,同时还就其在工业机器人、汽车自动导航与驾驶、身份识别、多目标跟踪、人工智能等自动化领域的应用进行了讨论.作者对于这一新兴研究方向未来的发展趋势作了展望,并就信息融合的一般理论、主要方法论、存在的问题等进行了探讨.     

信息融合理论的基本方法与进展

信息融合是现代信息技术与多学科交叉、综合、延拓产生的新的系统科学研究方向,由 于其在军事和民用领域已经展现出的有效与广阔的理论和应用前景,而备受国内外学者和众多 实际工程领域专家的高度关注.文章对近几年来信息融合国际年会和近百种国际学术期刊进行 了统计分析,对信息融合在军事和民用领域的应用分布情况、所采用的各种不同数学工具和研 究方法所占的比例、融合系统建模方法、算法、发展动向、存在的问题和解决这些问题的思路给 出了系统的综述.     

基于多源信息融合的雷达装备故障诊断决策研究

针对现代雷达故障特征参数较多,一般诊断方法难以解决的问题,文章在分析雷达检测诊断中诊断信息、诊断知识、推理过程不确定性的基础上,结合信息获取、信息融合、故障诊断及趋势预测几个方面建立了基于多元信息融合的雷达故障诊断决策模型,研究了信息融合中不同信息的归一化方法;该方法充分利用信息之间的冗余性和互补性.使故障诊断决策结果更加科学和准确;实例分析证明了所提模型及方法在提高故障检测率、减小Bayes风险、改善故障诊断性能方面具有明显的作用.     

多源测量信息融合的弱点动目标跟踪算法

在复杂背景下,为提高微弱点动目标跟踪系统的抗干扰能力,从不同信噪比的多红外图像序列出发,基于新的融合跟踪结构（增加局部处理器）,提出了一种模糊逻辑的多源信息测量融合跟踪算法：各局部处理器对原始测量进行恒虚警和自适应检测,将判决后的测量送入融合中心,融合中心对测量作模糊逻辑判决融合,转化为虚拟单序列测量,采用PDA卡尔曼滤波算法跟踪。实验结果分析表明,该融合跟踪算法与单序列相比,具有较高的跟踪精度、稳定性,避免了单图像序列跟丢。     

基于贝叶斯理论的冗余开关量传感器信号的融合研究

基于贝叶斯理论提出了火电机组开关量检测系统融合规则的代价分析方法,针对多传感器冗余测量系统,采用等风险和不等风险系数进行了各种逻辑组合方法的误动和拒动代价的比较,得出了修正“三取二”方法的具体方案。按代价最小的原则,在被测对象正常概率P0的不同范围内采用分段融合方法,真实反映了对象状态,并提高了测量系统的正确性。     

多速率传感器不完全测量数据自适应融合算法

针对不完全测量的多速率传感器系统,建立了自适应的分级融合算法。该算法构建了传感器子系统故障检测方法,检测子系统故障,确认不完全测量,避免单点故障污染整个系统;建立子系统的切换控制逻辑,引入映射矩阵,构成新的信息分享系数,自适应地选择子系统参与中心融合;对于不完全测量信息的传感器子系统,利用子滤波预测值代替估计值进行融合;全局状态估计利用各子系统的当前与前一时刻的估计值,提高了融合性能。对建立的算法进行仿真,验证了该算法的有效性。     

基于几何－拓扑广域三维地图和全向视觉的移动机器人自定位

面向大规模室内环境, 研究了基于全向视觉的移动机器人自定位. 提出用分层的几何－拓扑三维地图管理广域环境特征, 定义了不同层次的三维局部环境特征及全局拓扑属性, 给出了分层地图的应用方法. 构建了全向视觉传感器成像模型及其不确定性传播方法, 使得地图中的概率元素能够在系统中有效应用. 采用随机点预估搜索的方法提取环境元素对应的曲线边缘特征. 用带反馈的分层估计方法在融合中心对多观测特征产生的相应估计状态进行总体融合. 以分层逻辑架构设计实现了移动机器人交互式自定位系统. 实验分析了真实环境中不同初始位姿和观测信息情况下定位系统的收敛性和定位精度, 在考虑动态障碍物的遮挡情况下完成了机器人的在线环境感知和运动自定位任务. 实验结果表明本文方法的可靠性和实用性.     

基于模糊评判的决策级信息融合算法的研究

文章针对水电故障诊断系统中普遍采用的传感器阀值判断方法引起的信息损失问题,将决策级信息融合技术应用于故障诊断系统中。在模糊综合评判技术和软判决融合结构下,提出了一种新的决策级信息融合算法。该算法以合成运算和全局决策融合来自多传感器的局部判决以获取所处理对象的综合决策分析,并通过在丰满水电仿真系统的故障诊断系统中的实际应用表明该算法优于传统的故障检测方法。     

指数分布的泛逻辑自相关性

指数分布是复杂控制系统中随机参数的一种常见分布.给出了广义区间[a,b]上的相关概念,研究了指数分布的泛逻辑自相关性,给出了指数分布对应的N范数、N性生成元,讨论了N范数不动点l及广义自相关系数k与指数分布参数θ之间的关系.当θ>0时,发现了广义自相关系数k的恒负性.设参数θ在(一oo, ∞)均有意义,发现了k是θ的中心对称奇函数,即k的中心对称性,以及不动点l的守1性.最后举例说明了求解k值的具体方法,为从泛逻辑角度来分析复杂系统中的控制参数提供┅了一种新的思路.     

ARJ解释方法及其目标求解算法的研究

本文研究了规则相关性理论,给出了“递归时刻”、“完成时刻”定义及相关性定理,在此基础上提出了一种新的演绎方法——ARJ(AdvancedRelational Join)解释方法。它比RJ解释方法具有更大的优越性,从而有效地提高了逻辑程序设计系统效率。     

Preface

Over recent years, the notion of agency has claimed a major role in defining the trends of modern research. Influencing a broad spectrum of disciplines such as Sociology, Psychology, among others, the agent paradigm virtually invaded every sub-field of Computer Science, not least because of the Internet and Robotics.Multi-agent Systems (MAS) are communities of problem-solving entities that can perceive and act upon their environments to achieve their individual goals as well as joint goals. The work on such systems integrates many technologies and concepts in artificial intelligence and other areas of computing. There is a full spectrum of MAS applications that have been and are being developed; from search engines to educational aids to electronic commerce and trade.Although commonly implemented by means of imperative languages, mainly for reasons of efficiency, the agent concept has recently increased its influence in the research and development of computational logic based systems.Computational Logic, by virtue of its nature both in substance and method, provides a well-defined, general, and rigorous framework for systematically studying computation, be it syntax, semantics, procedures, or attending implementations, environments, tools, and standards. Computational Logic approaches problems, and provides solutions, at a sufficient level of abstraction so that they generalise from problem domain to problem domain, afforded by the nature of its very foundation in logic, both in substance and method, which constitutes one of its major assets.The purpose of this workshop is to discuss techniques, based on computational logic, for representing, programming and reasoning about multi-agent systems in a formal way. This is clearly a major challenge for computational logic, to deal with real world issues and applications.The first workshop in this series took place in Las Cruces, New Mexico, USA, in 1999, under the designation Multi-Agent Systems in Logic Programming (MASLP'99), and affiliated with ICLP'99. In the following year, the name of the workshop changed to Computational Logic in Multi-Agent Systems (CLIMA'00), taking place in London, UK, and affiliated with CL'2000. The subsequent edition, CLIMA'01, took place in Paphos, Cyprus, affiliated with ICLP'01. The present edition, CLIMA'02, takes place in Copenhagen, Denmark, on August the1st of 2002, and is affiliated with ICLP'02 and part of FLOC'02.We would like to thank the authors of the submitted papers, the members of the program committee and the additional reviewers for their contribution to both the meeting and this volume. We would also like to thank Michael Mislove for his help with the editing of the proceedings.

Programme Committee

Jürgen Dix (The University of Manchester, UK)

Thomas Eiter (Vienna University of Technology, Austria)

Klaus Fischer (DFKI, Germany)

Michael Fisher (University of Liverpool, UK)

James Harland (Royal Melbourne Institute of Technology, Australia)

Wiebe van der Hoek (Utrecht University, The Netherlands)

Katsumi Inoue (Kobe University, Japan)

João Alexandre Leite (New University of Lisbon, Portugal)

Luís Moniz Pereira (New University of Lisbon, Portugal)

Ken Satoh (National Institute of Informatics, Japan)

V. S. Subrahmanian (University of Maryland, USA)

Francesca Toni (Imperial College, UK)

Paolo Torroni (University of Bologna, Italy)

Additional Reviewers

José Alferes

Koji Iwanuma

Andrea Schalk

Tadashi Araragi

Gerhard Lakemeyer

Michael Schroeder

Alastair Burt

Wei Liu

Kenji Taguchi

Anna Ciampolini

Seng Loke

Hans Tompits

Pierangelo Dell'Acqua

Hidetomo Nabeshima

Mirosaw Truszczynski

Uwe Egly

Naoyuki Nide

Mathijs de Weerdt

Michael Fink

Inna Pivkinna

Michael Winikoff

Chiara Ghidini

Fabrizio Riguzzi

Cees Witteveen

Hisashi Hayashi

Chiaki Sakama

Full-size table

For this edition of CLIMA, we have received 25 submissions of which 12 were selected for presentation, after a careful review process where each paper was independently reviewed by three members of the Program Committee.The workshop consisted of five sessions: four devoted to the oral presentation of the selected papers and subsequent discussion; and one devoted to a panel discussion, Paolo Torroni being the invited moderator. There follows a brief overview of the workshop.

Session 1 - Agents: Arguments and Updates

Schroeder and Schweimeier present a framework based on logic programming with 3-valued multi-agent argumentation and fuzzy unification, for knowledge representation and reasoning in agents, to accommodate arguments for negotiating agents when agent communication is subject to uncertainty.Leite et al. extend the language LUPS introducing MLUPS, an update command language designed for specifying the flexible evolution of hierarchically related groups of agents, based on logic programming, thus assigning them declarative semantics.Kakas and Moraïtis present a modular argumentation framework for modelling agent deliberation, where object level arguments can be made conditional on agents' roles and the priority relation amongst such roles can, in turn, be made conditional on contents, on top of which a simple form of abduction allows dealing with incomplete knowledge.

Session 2 - Logics for Agents

Toyama et al. introduce a translation of multi-agent autoepistemic logic (MAEL), a logic for multi-agent systems based on Moore's autoepistemic logic, into logic programming, and show the correspondence between MAEL extensions and the stable models of the corresponding logic program.Dell'Acqua et al. extend their previous work on abductive logic programming based multi-agent systems, in which agents can update themselves and each other, eliminate contradictory update rules, abduce hypotheses to explain observations, and use them to generate actions, with asynchronous based communication through the use of buffers.Harland and Winikoff discuss the formalisation and implementation issues of BDI-type agents, using a Linear Logic based calculus that allows a mixture of forward- and backward-chaining techniques.

Session 3 - BDI Agent Systems

Bordini and Moreira investigate how far the Asymmetry Thesis Principles formulated by Rao and Georgeff are actually met by the abstract agent specification language AgentSpeak(L), hence contributing to the reconciliationBetween practice and theory of BDI-based agents.Araragi et al. formalise and propose a method to solve a verification problem that arises in implementing a commitment strategy for the BDI architecture, namely the verification of the suitability and/or feasibility of the intentions of an agent.Nide et al. extend with mental state consistency features their previously presented deduction system for CTL-based propositional BDI Logics using sequent calculus, as a step towards the use of the expressive power of BDI Logics as executable specification languages of rational agents.

Session 4 - Agents: Speculative Computation and Introspection

Hayashi et al. address the issue of integrating speculative computation and action execution through logic programming, namely by devising a method for plan modification when speculative computation fails or actions are executed.Iwanuma and Inoue refine the first-order consequence-finding procedure based on clausal tableaux SOL, with conditional answer computation and skip-preference, to formalise speculative computation in a master-slave multi-agent system.Bolander investigates on finding consistent classes of formulas under the syntactical treatment of knowledge and belief, identifying three maximal sets of introspective beliefs that strong introspective agents can consistently maintain so as to avoid the paradoxes of self-reference.

Session 5 - Panel Discussion

Torroni moderates a panel discussion entitled "Logics and Multi-agents: towards a new symbolic model of cognition".This volume constitutes the proceedings of CLIMA'02.September 2002, Jürgen Dix, João Alexandre Leite and Ken Satoh (Guest Editors)     

命题对象的空间逻辑运算模型

在现有逻辑系统中,各连接词的运算模型都可以归结为某些“代数算子”,其共同特征是仅考虑了命题所描述集合的代数测度大小,而没有考虑它们在几何空间中的位置关系。文章以“空间位置相关性”为中心,提出了“摸天花板问题”,分析了逻辑运算中存在的几何位置相关性。在“命题对象”、“真位向量”、“空间图像”等概念的基础上,提出了命题对象的空间逻辑运算模型,并结合格分维理论给出了在几何图像中的具体应用形式。本文工作拓展了泛逻辑学中广义相关性的含义,为连接词的运算形式提供了一种新的模型。     

多层BP网络的研究及应用

在本文中，我们用统计模式识别的方法分析了目前在模式识别中得到广泛应用的多层ＢＰ神经网络，揭示了具有线性输出单元的多层ＢＰ神经网络用作特征提取器和分类器时具有良好性能的原因。同时，我们设计了一个使用ＢＰ网络作为特征提取器和分类器时，普适的工件识别系统，对不能识别的样本，采用模糊推理技术，把传统的直观特征识别结果和多层ＢＰ网络结果特征级上融合，提高系统的性能。