On QoS Autonomous Control Of Cognitive Networks Based On Bayesian Networks

The Internet has developed into a huge and complex system with the characteristics ofnonlinear and dynamic. Meanwhile, QoS management and control in the traditional networkis undergoing many challengeswith the new network access technologies and continuousenrichment of network bearing service. Due to the lack of autonomy and intelligence, thenetwork does not have a comprehensive knowledge of its own conditions and behaviors.And the increasingly complex network contributes to the results that the traditional QoSguarantee method of isolated and static performs inefficient and makes decisions passively.Then the network is often congested and its QoS performance becomes deteriorated. Whenservice transmission and QoS guarantee cannot be dynamically adjusted according tochanges in the environment, both network resource utilization and User Satisfaction Degreewill be degraded. Cognitive network is proposed as an active network with the characteristicof cognitive with the inspiration of the technology of cognitive radio. A cognitive network isa network with cognitive process that can perceive current network conditions. It can plan,decide and act on those conditions. It is considered to be the inevitable trend of futurecommunication network with fundamental features of wireless, mobile, broadband andall-IP.Autonomous control technology of QoS for cognitive network studied in this thesismainly refers to the autonomous, intelligent and self-adaptive methods during theimplementation of Qos decision and control in the cognitive environment. It can ensureend-to-end QoS of network and promote the resource utility. According to the problemssuch as multi-service, demand differences, dynamic change, resource scarcity in currentnetwork, several issues are taken into consideration in the paper, such as QoS health degreeassessment, QoS degradation location and QoS autonomous control, etc. Bayesian networktheory is used to realize global monitoring, analyzability, control ability in the dynamic andcomplex cognitive network environment.1. A service-oriented cognitive network QoS control framework is proposed in thispaper. We first research end-to-end QoS requirements of cognitive network service aimingat cognitive network QoS autonomous control. A service-oriented cognitive network QoScontrol framework is provided combining with OODA process based on the analysis of thegeneral QoS framework.2. A QoS health degree assessment method based on Fuzzy Dynamic BayesianNetwork is proposed. From the macroscopic point, with the consideration of differentnetwork elements, services and links of the network, re-define the three QoS parameters-delay, jitter, packet loss rate, and build the dynamic Bayesian network model ofcognitive network QoS health degree assessment. Using fuzzy classification method, thecontinuous variable fuzzy classification DBN evidence can be applied to the learning andreasoning, reasoning by directly reasoning algorithm, getting the cognitive network QoShealthy degree of probability and its development trend in continuous time slices.Theseassessment results provide a reference value for the network decision and control.3. A deterioration location method based on Bayesian Network is established forcognitive network, which allows for the high cost in end-to-end probe method andlimitation of imprecise positioning. Path status information is obtained through a smallnumber of end-to-end probing in LM1packet loss rate model. A link degradation prioriprobability learning for link failure is carried out in accordance with the Bayesianestimation method. On this base, part of paths that have nothing to do with degradation andthe relative links these paths cover are deleted from model, which can degrade thecomplexity of the model. A precise positioning of QoS degraded links is achieved throughreasoning of local joint tree algorithm, which achieves accurate foundation for QoSindependent control of traffic level.4. An autonomous control method of QoS for cognitive network based on influencediagramis proposed. Combing with related knowledge of traditional network and cognitivenetwork, it analyses and determines the influencing factors of service-oriented autonomouscontrol of QoS for cognitive network, constructs the corresponding influence diagrammodel. Ultimately the autonomous control methodof QoS for cognitive network based on akind of influence diagram can be realized. This method achieves self-learning andself-reasoning with Bayesian network cluster tree reasoning algorithm and optimal decisionmethod of nodes instantiation. It realize independent decision and control of the QoS forcognitive network through influence diagram and maximizing the utility equation.