Research On User Association Schemes In Heterogeneous Cellular Networks

With the proliferation of mobile devices and the explosive growth of data traffic, cellular networks are evolving towards increasing heterogeneity. By overlaying low power low cost small cells with macro cells, heterogeneous cellular networks (HCNs) bring more radio resources to users and promise large gains in network throughput. However, due to the geographical randomness of small cells and the massive power difference between macro cells and small cells, the traditional user association scheme that assigns users to the strongest base stations in terms of received signal power results in underutilized small cells and does not work well in HCN scenarios. Hence it is highly desirable to devise a user association scheme that can proactively offload users to small cells and achieve load balancing in HCNs. On the other hand, user association schemes have a significant impact on user QoS (Quality of Service) performance. The available resources at the associated base station dictate the data rate of a user. Due to the existence of the cross-tier interference between macro cells and small cells as well as the intra-tier interference between adjacent small cells, the choice of serving base station greatly affects the perceived SINR (Signal to Noise plus Interference Ratio) for a user. Moreover, as HCNs are becoming increasingly dense, there exist underutilized small cells with low loads or empty user sets. These underutilized small cells increase system energy consumption and bring undesired interference to nearby users.This thesis addresses the user association problem in the context of downlink HCNs. Towards this end, the main contributions of this thesis are summarized as follows:(1) This thesis proposes a novel framework to solve the user association problem subject to user minimum rate requirements. A joint optimization problem over user association and resource allocation that accommodates user minimum rate requirements is proposed. The joint optimization problem is then reformulated as a resource allocation problem with cardinality constraints. The reweighted l1 regularization method is used to obtain a sparse solution to the resource allocation problem. User association is then derived from the sparsity of the obtained solution. The obtained user association scheme guarantees that a user is served by a single base station. Simulation results verify the effectiveness of the proposed scheme.(2) This thesis proposes an outage-aware user association scheme to address the outage deterioration for offloaded users. A set of weighting parameters are devised to take into account the user outage performance and a combinatorial optimization problem with weighted base station loads is formulated. The combinatorial optimization problem is then reformulated as a monotone submodular optimization problem with matroid constraints. An accelerated greedy algorithm with lazy evalations is then used to obtain a near optimal solution to the reformulated problem. Simulation results show that the accelerated greedy algorithm has a low computational complexity and the proposed scheme achieves load balancing as well as reducing user outage probabilities.(3)This thesis proposes a user association scheme that strikes a tunable tradeoff between load balancing and user long term SINR. A bi-criteria optimization problem is formulated, which is then linearized as a utility maximization problem with a tunable parameter. In addition, we show the utility maximization problem can be reformulated as a maximum bi-partite matching problem and can be solved in polynomial time via the Hungarian method. Simulation results show that that the proposed scheme achieves load balancing and can strike tradeoffs between load balancing and user QoS by tuning the optimization parameter.(4)This thesis proposes a novel framework to jointly decide user association and base station sleep operation in HCNs. A heuristic algorithm is proposed that iteratively determines the sleep operation of base stations and solves the user association subproblem. By exploiting the load conditions of small cells and the interference relations between users and adjacent small cells, an interference relation matrix is constructed to help us identify the candidate base stations that can be put into sleep. User associations are then aggregated to selected small cells that remain active. Simulation results show that our proposed scheme achieves load balancing across macro and small cells and reduces the number of active base stations. Numerical results show that user SINR can be improved by base station sleep control.