【作者】 刘再群；

【导师】 丁永生；

【作者基本信息】 东华大学 ， 控制理论与控制工程， 2009， 博士

【摘要】 随着网络应用需求日益朝着高性能、大规模、多样性的方向发展,对Internet网络提出了更高的分布式要求。对网络系统逐渐提出自扩充性、可移动性、可生存性、简单易操作性的要求。并且,随着用户和网络环境变化的自适应性等特点,有必要进一步优化Internet网络体系结构,并设计其应用。生物系统是分布式自治系统,能提供给科学和工程领域各种富有成效的技术和方法。在生物界中,像人体系、蜂群和人类社会这样的大规模系统,已形成许多重要的原理和机理,正好可以满足以上对Internet的未来需求。本课题将生物细胞因子网络的一些相关原理和机理,应用到Internet网络中,提出新颖的生物网络计算模型和相关智能算法语应用。首先,本文对人工神经网络、人工免疫系统和人工内分泌系统等生物智能理论及各种智能技术发展进行了综述,指出了目前发展存在的问题以及将来的发展方向。并介绍了细胞因子网络的生理基础及其调控机理和网络模型,为本文智能网络模型及算法的研究设计奠定了生物理论基础。介绍了智能移动Agent的特性及应用领域,为生物网络模型的设计提供技术支撑。其次,从细胞因子网络的理论和研究结果中抽象出一种新颖的生物网络集成计算框架,设计了一种细胞因子网络结构及其中间件,该中间件包括网络平台和生物实体。生物网络平台由通用基础组件、网络服务模块和关联模块三部分组成,生物实体是网络应用中的最小组件。并设计了生物结构的通信机制。接着,对初步激活辅助性T细胞前体提出了一个“两步双信号”模型,用来了解外周的自我-非我区分的机制。基于这个“两步双信号”模型,我们提出了一种新颖的人工免疫算法用来产生不同的“两步双信号”生物实体,而且设计了基于一个“两步双信号”生物实体的分布式入侵检测系统用来保护生物网络结构的安全,对它的运行机制进行了讨论。借鉴造血细胞因子模型的协同调整模型,扩展生物实体的设计,提出了一种Web服务的协同调度。造血细胞因子模型研究了细胞因子之间的促进与抑制机制,以及对于造血祖细胞的控制调节作用。在细胞因子网络基础上,对Web合成服务请求与调度的建模,融入基于造血细胞因子调节模型,设计了一种资源的实时计算方法。并结合两种服务选择策略,在扩展的生物网络平台上研究系统的响应能力和服务质量。仿真实验表明基于造血细胞因子网络的Web服务实时调度方法是有效的,具有很好的性能。借鉴生物神经内分泌免疫系统的协同进化机制,基于细胞因子网络平台提出了Web服务的合成方法。在细胞因子网络调控下的生物实体代理Web服务,构成为一个带有条件的米兰机单元,Web服务的合成可以转换为米兰机进化过程。生物实体通过消息匹配和条件约束形成细胞因子网络,Web服务的合成通过生物实体细胞因子网络支持。在服务协同进化过程中可以动态调整其合成的服务,完成服务的动态自组织合成和管理等工作。仿真结果表明该方法在环境动态变化时具有适应性。然后,借鉴生物细胞因子网络的协同进化机制,为生物实体设计协同感知的属性与计算能力,提出了Web服务合成中的方法。在细胞因子网络调控下的生物实体代理Web服务,具有细胞因子网络实体的基本功能,并且扩展了对于分布感知的支持。通过考虑连接亲和度、多亲和度服务质量模型,以及信任亲和度的计算,提出了服务感知计算的具体方法。仿真结果表明该方法具有良好的性能和可扩展性。最后,对全文研究内容进行了总结,指出研究工作中存在的不足,明确了下一步的研究方向。更多还原

【Abstract】 With the development of network application requirements for high performance, large scale,and diversity,Internet should be highly distributed,that is to say,the network with the core of users should be extensibility,mobility,survivality,simple operation,and adaptability to the long or short change of users and network environments.Hence,it is necessary to optimize further the architecture of Internet and design its applications. Biological information systems can be regarded as distributed automatic systems,and can provide various effective technologies and methods for science and engineering field.In biological world,the large-scale systems,such as human society,bee swarm,and human body systems have developed many important theories and mechanisms that can be satisfied with the future requirements of the Internet.In the thesis,we apply the correlated theories and mechanisms of cytokine network to the Internet,and bring forward a novel bio-network computing model and some intelligent algorithms.Firstly,we investigate the development of artificial intelligent technologies,including artificial neural network,artificial immune system,artificial endocrine system,and the others intelligent technologies.And their difficulties and further developments are summarized.Some relative physiological theories and modulation mechanisms or models of neural system,endocrine system,and immune system,are briefly introduced.They provide the biological bases for the intelligent network models and algorithms studied in this thesis.We introduce characteristic of intelligent mobile agents,which is useful for designing bio-network models.Secondly,from theories and studies of the cytokine netowork,we abstract a computing framework integrated by neuro-endocrine-immune networks and design bio-network architecture and its middleware.The bio-network middleware includes bio-network platform and bio-entities.The bio-network platform consists of general basic components,bio-network services modules and context modules.A bio-entity is the smallest unit in the network applications.Then we design communication mechanisms of the bio-network. Again,a two-step,two-signal(TSTS) model for the primary activation of precursor helper T cells is proposed to incorporate a mechanism of peripheral self-nonself discrimination.Based on the TSTS model,we propose a novel artificial immune algorithm to generate different TSTS bio-entities.We design a distributed intrusion detection system (DIDS) based on the TSTS bio-entities to protect the security of our bio-network architecture.We also discuss its running mechanism.Next,inspired by schedule model of hematopoiesis cytokine network,extending the design of bio-entity,a co-schedule model is proposed for Web services.In the model of hematopoiesis cytokine network,the synergy and inhibitor mechanisms adjusting the capability of hematopoiesis cells are studied.On the basis of cytokine network,the request and schedule of the Web composite services are modeled by incorporating the model of hematopoiesis cytokine network to design an approach of computing resources. Considering two strategies of services selection,the capability of response and service quality are studied in the extended bio-network platform.The experimental study shows that the real-time schedule method based on hematopoiesis cytokine network is valid and has promising performance.After that,inspired by co-evolutionary mechanism in neuroendocrine-immune system, a cytokine network platform is proposed for Web services composition.In the control of cytokine network,bio-entity delegates Web service to construct a Melay evolutionary unit with conditions.Web composition is then transferred to Melay evolutionary process. Bio-entities construct cytokine network through message matching and conditional constraints to support Web services composition.During the co-evolution process of Web services,the composed services are dynamically adjusted to finish the dynamical composition and management of Web services.The simulation results show that the approach is adaptive in dynamic environments.And,inspired by co-evolutionary mechanism in cytokine network,the features and computing capability of bio-entity are designed for a clustering approach in Web services composition.In the control of cytokine network,bio-entity delegates Web service with basic functions of cytokine network entity.Moreover,bio-entity is extended to support distributed awareness.Considering the connect affinity,multiple affinity-based service quality model,and trust affinity,the concrete methods for service awareness are supported. The simulation results show that the approach is of high performance and extendable capability.Finally,a summary of the thesis is made,and the deficiency in the project and the further development are narrated respectively.更多还原