Research On Clustering Scheme For Base Station Cooperation Under Nonideal Condition

In order to utilize cooperation processing techniques in realistic cellular syseme, a crucialstep is to divide the whole system into several small clusters. The existing dynamic clusteringschemes attracted more attention because of the good performance. But they were performedunder ideal condition. Channel state information CSI distortion and fast channel vibration werenot considered that reflected in the following points. First, the existing clustering model did notguide clustering very well with CSI distortion that would lead to bad cluster structure and lowcooperation performance. Second, with fast channel vibration, the existing clustering schemewould update cluster structure frequently that lead to increasing overall cost. It is difficult tocluster and cooperate in time. Third, obtained the whole CSI matrix (the CSI of each base station)is hardly achieved actually. So, this dissertation would focus the research on these problems ofclustering scheme under nonideal condition.Existing clustering model didn’t guide base stations clustering effectively. Aiming at theissue, this paper established a clustering model based on the average signal to interference ratio(SIR) gain. Based on the clustering mode, a clustering scheme based on maximizing the averagesystem performance is proposed that was aiming at the issue of channel vibration. The wholeCSI matrix is difficult to obtain. Aiming at the problem, a fast clustering scheme based on partialCSI and uniform cluster size was proposed. The chief content of this article is as follows.1A clustering model based on the average SIR is proposed. The average SIR gain candecrease the influence of CSI distortion. So the base stations cooperation degree is measured bythe average SIR gain in the model. And clustering is performed aiming at maximizing the sum ofaverage SIR ratio gain. And then the clustering objective of maximizing the system performancewas approximately to maximizing the sum of average SIR ratio gain in base stations. Simulationresults showed that, with the CSI distortion, the benefit-tree clustering scheme under theproposed clustering model provided a0.3bit/s/Hz increase of the average system performancecompared with under the clustering model based on degree of wiliness to cooperate.2A clustering scheme based on maximizing the average system performance was proposed.Base stations merged into a cluster structure with the principle of maximizing averageperformance in a period of time. Then, the cluster structure is used in the period of time. So theoverall cost is reduced by way of decreased the cluster structure updating frequency. Simulationresults showed that, when the factors of CSI distortion and channel vibration both are0.3, theproposed scheme provided a0.5bit/s/Hz increase of the average system performance comparedwith the benefit-tree clustering scheme.3A fast clustering scheme based on partial CSI and uniform cluster size was proposed. Thecooperation possibility of base stations away from earch other is poor. It only needed the partialCSI of neighboring base stations. And the sparse similarity matrix was formed by the averageSIR gain of neighboring base stations cooperation. Then, among neighboring base stations,collaborative information was interactive update to quickly generate multiple clusters. Finally, the smaller clusters were combined to achieve the purpose of uniform cluster size and fastclustering. Simulation results show that not only the cluster structure is uniform but also theconvergence speed is not slowdown with increasing the system size, while the schemeperformance is better than existing schemes.