基于认知层的认知网络结构及其认知方法
Architecture of Cognitive Network Based on Cognitive Layer and Corresponding Cognitive Method
-
摘要: 设计了一种认知网络系统结构即独立于传统OSI七层结构的新的认知层,方便认知网络实现中的模块化,并降低未来网络系统实现的复杂度.给出了认知层的功能实现模块,阐述了2种认知能力具体的实现方法:利用多目标遗传算法对认知网络进行“离线内省”学习,解决“有经验指导”的学习问题,提高认知网络的离线学习能力;利用案例推理对认知网络进行“在线判决”,解决模型的快速匹配问题,提高认知网络的实时处理能力.以异构网络中流量控制为例对这2种认知能力进行了仿真.仿真结果表明,加入认知以后网络的各项Qos性能有明显提高.Abstract: The architecture of cognitive network based on congnitive layer is proposed.Characteristics of the architecture can be found as follows.Firstly the cognitive layer independent of traditional OSI seven layers can realize the modularity and reduce the complexity of the network implementation.Secondly,implementation modules of cognition are given in cognitive layer,especially demonstrating certain methods to realize two kinds of cognition,which implement the off-line learning based on genetic algorithms,and implement the real-time decision based on case-based reasoning.Finally,an example of using the cognitive layer to control the data flow in heterogeneous network through reconfiguring network parameters is given.Simulation result shows that the performance is greatly improved when the mechanism of cognition is adopted.