Research On Cell Selection And Enhanced Inter-Cell Interference Coodination In Hetnets

As the demand for higher data rate is increasing rapidly, some LTE-Advanced techniques launched by Third Generation Partnership Project (3GPP) have been investigated intensely, such as Coordinated Multipoint Transmission and/or Reception (CoMP)、Relaying and Heterogeneous Network (HetNet). Among all these techniques, HetNet could improve system capacity as well as coverage efficiency by deploying additional Low Power Nodes (LPNs), i.e. Pico base stations (PBSs), femto BSs and relays under the coverage area of traditional Macro BSs (MBSs). HetNet enables flexible and low-cost deployments and provides a uniform broadband experiment. But the change of network deployment will bring new problems such as load and interference, and they play a vital role in cell capacity and system performance.In this paper, on the issue of load and interference, the technology of cell association and time domain enhanced Inter-Cell Interference Coordination (elCIC) for Macro-Pico HetNets has been researched deeply. The concept and key technologies of HetNets have also been introduced in this paper. And the previous cell association and ABS configuration algorithm have been summarized and analyzed. We investigated a novel user association scheme and proposed asynchronous dynamic ABS configuration approach in Macro-Pico HetNets scenario. Then the two algorithms are verified by system level simulation.In the matter of cell association, a new cell association algorithm is proposed in this paper, by setting up the problem as multi-objective optimization problem, and then Greedy OF algorithm is proposed to solve the problem. So that the scheme could make a performance compromise between load balancing and network throughput in HetNet. The numerical results demonstrate that our scheme could shift the load from over-loaded MBSs to under-loaded PBSs and improve network throughput, especially for cell edge users.For another, ABS configuration problem of elCIC introduced Q-learning theory considering the factor of user throughput of picocell and macrocell. The Q-learning algorithm developed in previous works is extended so as to dynamically determine ABS configuration which is asynchronous in HetNets. MBSs optimally learn their ABS densities and reduce power levels in the proposed approach. System level simulations demonstrate that the proposed Q-learning algorithm based on eICIC works properly, and outperforms conventional approach with fixed ABS configuration.