Research On Distributed Spectrum Access Mechanism In Cognitive Radio Networks

The future wireless network will face two major challenges.(1) Center controllednetwork cannot meet the complexity and flexibility requirements of large-scaleheterogeneous network;(2) The scarcity of spectrum resource has become a bottleneckfor the sustainable development of wireless communication. Considering that distributednetwork has the advantage of rapid deployment, high stability, flexible structure, it willgradually become the main way in future wireless communications. Users can enhancethe spectrum utilization by the dynamic spectrum access technology in cognitive radio.Therefore, dynamic spectrum access in distributed network has always been a researchhotspot all over the world.In this thesis, the dynamic spectrum access technology for large-scale distributedcognitive networks has been studied. The main works can be summarized as follows:First, the optimization problem of distributed system is abstracted to the problem ofcognitive process of distributed groups. On this basis, according to the characteristics oflarge-scale distributed network, the true relationship of users is modeled as a graphicalgame. The basic structure of cognitive processing is established with the perspective ofartificial intelligence. To solve the dimension disaster problem caused by the arbitrarinessand complexity of user-relationship topology in the large-scale distributed wirelessnetwork, graph theory is applied to establish the real game model for solving Nashequilibrium.Second, this thesis studied the dynamic spectrum access in distributed cognitionnetworks based on cooperation graphical game. A dynamic spectrum access algorithm ispresented to achieve pure-strategy Nash equilibrium without conflict based on no-regretlearning. This algorithm minimizes the personal regret value instead of that of system.Compared with the existing literatures, the presented algorithm in this article can reducethe computing complexity effectively and meet the real-time communicationrequirements. Simulation results show that the pure-strategy Nash equilibrium achievesbetter system performance in terms of capacity and power utilization, specifically in thesystem with scare resources.Finally, for the problem of dynamic spectrum access in unknown distributedenvironment without a priori knowledge and information exchange, Multi-Q learningalgorithm is proposed based on non-cooperation graphical game. Q-value table is updated by return value which is obtained through p-CSMA protocol. The algorithm aims toachieve long term optimal and the pure-strategy Nash equilibrium. In addition, the thesisproves that the proposed game is an exact potential game which has at least one purestrategy Nash equilibrium; more importantly, the pure strategy Nash equilibrium is theoptimal strategy. Simulation results show that Multi-Q learning achieves high systemcapacity and utility of users in the graphical game are mainly determined by the degree ofnodes, without any direct relationship with the number of users.