Research On Radio Resource Management Technology In Hetnet System

Along with the development of wireless communication technology and the ever increasing requirement of indoor coverage, the deployment of Heterogeneous Network (HetNet) has become the trend of the next generation wireless communication networks. The introduction of HetNet system brings opportunities to the traditional cellular network, and at the same time it brings new challenges. This thesis focuses on the wireless resource management (RRM) technology in HetNet system. Furthermore,we abstract three primer contradictions from centralize RRM,semi-centralize RRM and distributed RRM algorithms separately, which is the performance and the overhead, the independent organization and coordination, the complexity and limitation. The main contributions of the dissertation include the following aspects. To tackle these three coordination, we spread our research on RRM algorithm in large-scale HetNet system.Firstly, the first contradiction confronted with centralized RRM in large-scale HetNet system is the urgent need for centralized control to achieve the optimal system performance and the ever-increasingly overheads along with the scale of the HetNet. To balance the performance of algorithm and the cost of the execution, we introduce clustering mechanism in this paper. The whole HetNet system is modeled as a multiple-input and multiple-out nonlinear system. In order to reduce the complexity of centralized algorithm, we adopt clustering to decouple the nonlinear system. Through clustering, the whole network becomes a loosely-coupled system. The centralized algorithm only needs to execute resources allocation in form of cluster, which greatly reduces the complexion and latency. Base on the above idea, a dynamic clustering algorithm is proposed. This clustering algorithm can simultaneously generate several clusters with dynamic size, which can fully adapt to smallcell deployment density and interference environment. Based on this clustering algorithm, a cluster based subchannel matching algorithm is designed. This matching algorithm can dynamic adjust sub-channel quote which is allocated to smallcell clusters according to the cluster size. The proposed matching algorithm can achieve the balance of the performance of fairmess and effective.Secondly, the second contradiction confronted with semi-centralized RRM in large-scale HetNet system is the individual autonomy in the micro level and the overall coordination in the macro level. On the one hand, the smallcell has initiative, which can independently carry out the decision to maximize its own interest. On the other hand, the smallcell has selfishness, which needs the coordination of the core network control entities to balance their conflicting interests. However, there exists hysteretic in the coordination. To solve the problem of hysteretic, in this paper, we decouple the coordination into del ay-sensitive power control process and delay-tolerant topology adjustment process. Furthermore, the power control process is operated on open-loop control and the topology adjustment process is operated on close-loop control. Based on this idea,we propose an evolutionary clustering algorithm. Through driving the smallcells to spontaneously switch to less interfered cluster to balance the interference suffered by the whole network, smallcells can gradually evolve to form an orderly consistent behavior, this evolutionary clustering algorithm can enable the HetNet system to reach a dynamic balance and keep up with the change of the external environment.Thirdly, the third contradiction confronted with distributed RRM in large-scale HetNet system is the complexity of the system environment and the limitation of individual smallcell's cognition. On the one hand,the characteristic of self-plugging and self-deployment causes the strong dynamic of the environment. On the other hand, the feature of large scale of network size leads to the limitation of the cognition of small cell. To solve the problem of incomplete information and incomplete rationality of the smallcell caused by the cognition limitation, we introduce reinforcement learning algorithm in this paper. The smallcells are viewed as independent learning agents, and furthermore, the whole HetNet system is modeled as a multi-agent system. Each smallcell agent dynamically adjusts its strategy according to the changes of the environment to maximize its own throughput under the constraint of guaranteeing the communication quality of the adjacent macrocell users.However, the smallcell agent faces a dilemma, that is to make the optimal decision it needs the global information while it lack this information when centralize entity is no longer exists. In order to overcome the cognitive limitations of small cell intelligence, we design conjecture mechanism. Based on its own accumulation of incomplete information,the smallcell agent can learn the optimal strategy.