Research On Cell Selection And Load Balancing In Heterogeneous Cellular Networks

By offering higher spatial reuse, eliminating coverage holes and creating hot-spots, the heterogeneous cellular networks (HCNs) can achieve significant performance gains. However, due to the different transmit power among various base stations (BSs), the interference management and user association (cell selection) in HCNs are challenging. Considering that some load balancing schemes may not achieve the desired goal because of limited backhaul constraints, the Device-to-Device (D2D) communications are introduced into HCNs for improving system performance. The thesis studies the subject of "Research on user association and load balancing in heterogeneous cellular networks", the content contains the design of joint uplink and downlink user association, the design of QoS (Quality of Service)-aware user association, the design of joint user association and resource partitioning for load balancing, the design of joint user association and power control for load balancing, and the design of energy-efficient user association in downlink HCNs. The specific contributions of this thesis are listed as follows.1. Since various BSs in HCNs have disparate transmit power, there exists asymmetric uplink and down-link. To achieve the system-wide performance gains, an effective user association should integrate uplink as-sociation and downlink association. So far, how to design joint uplink and downlink user association schemes is still an open topic. The thesis tries to design an association scheme that jointly maximizes downlink through-put and minimizes uplink power consumption, and formulates it as a sum-utility maximization problem.It is easy to find that this problem is in a nonlinear and mixed-integer form, and hard to tackle.To solve it, the thesis designs a centralized association algorithm using gradient descent method, and develops a distributed association algorithm using dual decomposition method. As for the proposed algorithms, the thesis gives the corresponding convergence and complexity analyses. Simulation results show that, compared with the common methods, the proposed scheme has some significant advantages in the mass.2. Considering that the conventional association schemes may not be appropriate due to the special characteristics of HCNs, some novel schemes should be designed for HCNs. To fully exploit the network resources, an effective association scheme should have offloading capability, which can relatively balance the loads among various BSs. In addition, considering that HCNs provide a multi-radio access environment, a good association scheme should be designed for heterogeneous users with different quality of service (QoS) requirements. To this end, the thesis designs an offloading scheme for these heterogeneous users in HCNs, which is formulated as a sum utility maximization problem. To solve it, the thesis utilizes a gradient descent method and a dual decomposition method to develop centralized and distributed algorithms respectively. To highlight the effectiveness of designed algorithms, the thesis designs an unbalanced association with different QoS requirements and develops centralized and distributed algorithms for it. As for the proposed algorithms, the thesis gives the corresponding convergence and complexity analyses. Numerical results show that the proposed scheme significantly outperforms the introduced scheme in load balancing gain, rate fairness and QoS guarantee.3. As for HCNs and D2D-enabled HCNs, the thesis designs two association schemes with resource partitioning. In the balanced association, the offloaded users who are offloaded to underloaded small cells from overloaded macrocells often suffer the strong interference from high-power BSs. That leads to a bad result that the offloaded users achieve lower signal-to-interference-plus-noise-ratio (SINR) than these users in macrocells. In order to partially alleviate the SINR degradation, a two-subband partitioning model is designed in the association problem of HCNs. Considering that a mere association for load balancing may not fully balance loads among different BSs due to the limited backhaul capacity of HCNs, the thesis introduces D2D communication technique into HCNs. To mitigate the interference received by offloaded users and D2D receivers (RXs) from BSs, a three-subband partitioning model is proposed in the association of D2D-enabled HCNs. At last, the association schemes with resource partitioning in HCNs and D2D-enabled HCNs are formulated into two general utility maximization problems. According the forms of these problems, the thesis tries to design distributed algorithms correspondingly. As for the proposed algorithms, the thesis gives the corresponding convergence and complexity analyses. Simulation results show that load balancing, by itself, is insufficient, and additional resource partitioning is required to improve system performance. Meanwhile, simulation results show that the D2D communication is a good option for offloading traffic.4. Instead of achievable rate in the conventional association, the thesis utilizes effective rate to design two association schemes for load balancing in HCNs, which are both formulated as such problems with max-imizing the sum of effective rates. In these two schemes, the one just considers user association, but the other introduces power control to mitigate interference and reduce energy consumption while performing user as-sociation. Since the effective rate is the ratio of the achievable rate to the load (number of associated users) of some BS, these schemes can balance loads among different BSs. To solve the association problem without power control, the thesis designs an one-layer iterative algorithm. By combining this algorithm with power control algorithm, the thesis proposes a two-layer iterative algorithm for the association problem with power control. Specially, the outer layer performs user association using the algorithm of problem without power control. In the inner layer, the thesis updates the transmit power of each BS using a power update function (PUF). At last, the thesis gives the corresponding convergence analyses for user association procedure and power control procedure, and also gives the complexity analyses for the whole algorithms. As shown in sim-ulation results, the proposed schemes have superior performance than existing ones, i.e., achieve higher load balancing gains.5. In downlink HCNs, the thesis designs two kinds of association schemes, which include energy-efficient association and non-energy-efficient association. In fact, the latter can be regarded as a part (as-sociation phase) of the former. Finally, they are formulated as a problem with maximizing overall energy efficiency and a problem with maximizing the network utility that is a logarithmic function in terms of ef-fective rates respectively. Considering that the energy-efficient optimization problem is in a fractional and mixed-integer form, it is challenging for designers to achieve its optimal solutions. To this end, the thesis designs an effective three-layer iterative algorithm. In the first layer, the energy efficiency parameter is found via bisection method. In the second layer, association index and transmit power are optimized alternately. In the third layer, the user association problem is solved via dual decomposition method and the transmit power is updated by employing a power update function. As for maximal utility problem, it can adopt the association section of energy-efficient algorithm. In addition, the thesis gives the corresponding convergence analyses for user association procedure and power control procedure, and also gives the complexity analyses for the whole algorithms. Numerical results show that, compared with non-energy-efficient association, the energy-efficient association can achieve higher performance gains including energy efficiency and throughput.