Research On Load Balancing And Resource Allocation In Heterogeneous Cellular Networks

The explosive growth of smart terminals which combined to the mobile internet have sparked the rapid increase in mobile wireless data demand. Moreover, the traffic in a cellular network is typically unbalanced, such as more than50%of voice calls and more than70%of data traffic will originate from indoor users. Hence, providing good quality hotspot/indoor voice and data services is of great importance for the development of communication technique. Adopting a heterogeneous network (HetNet) architecture, composed of traditional macro base stations (MBS) overlaid with low power nodes (LPN), is a viable way to meet surging traffic demands in hotspot/indoor zones. However, the deployments of HetNet are facing two significant technical challenges, such as the load unbalancing due to the large transmitting power difference among various types BSs, the severe co-channel cross-tier interference, and the potential high energy consumption (EC). To meet the above challenges, this thesis studies the load balancing (LB) and resource mangement in HetNet. Through designing the proper optimization objective and applying the optimization theroy, this thesis proposes solutions for the following aspects:how to further improve the throughput performance, how to decrease the EC of BSs and improve the network energy efficiency (NEE), and how to decrease the EC of mobile terminals.Firstly, as the scarcity of spectrum resource, it is suitable that MBSs and LPNs are deployed with the same frequency bands. To improve the cell splitting gain, we need to design the new user association algorithms which actively "push" the mobile users onto the LPNs. Howerer, the severe co-channel cross-tier interference due to frequency reuse also needs to be carefully managed in HetNet. Otherwise, the gains from balancing load may be negated due to the bad link qualities that certain users suffer. Note that the interferce coordination has a strong interplay with user association, unlike the traditional studies that design the interferce coordination algorithm and the LB algorithm separately, this thesis develops a novel approach for jointly optimizing user association and almost blank subframe (ABSF) interferce coordination algorithm. This thesis formulates a network wide utility maximization problem corresponding to global proportional fairness (PF). Nevertheless, it is observed that the coupled problem is non-convex and NP-hard. Fortunately, this thesis reveals that the optimal ABSF density is the ratio of the number of vulnerable users and total users. Based on this relationship, the original combinatorial problem can be reformulated to a pure optimal user association problem. Then, by solving the problem with a greedy approach, this thesis proposes the marginal utility based user association algorithm, and obtain the cooresponding ABSF density. Moreover, this thesis extends the above results to the practical wireless system, through the simplification and approximation, this thesis proposes a cell range expansion (RE) algorithm, i.e., adding a positive bias to users measured received power of LPNs during user association base on the load condition of BSs, and then design the ABSF pattern.Secondly, based on the traffic volatility in the time domain and the spatial domain, this thesis proposes a LB and resource allocation to reduce the EC of BSs so as to improve NEE, by selectively turning off BSs and redistributing users originally associated with the turned off BSs. A NEE maximization problem subject to the rate fairness constraints among users is formulated, where NEE is defined as ratio of the network throughput and the total EC. This problem is NP-hard and thus is difficult to solve. To overcome this issue, this thesis proposes a new optimization approach, in which the LB scheme is first optimized under the fixed SCs on/off pattern, and the SCs on/off pattern is then derived iteratively. It is observed that the NEE maximization problem under the fixed SCs on/off pattern is a purely LB problem, which is a convex linear program by allowing multi-homing access, and the optimal LB algorithm can be obtained by interior point method. Due to the difficulty of applying multi-homing access in practice, by transforming the original maximization proplem to max-min problem, this thesis also derives a suboptimal but delicate user access rule for single access criterions. Combined with the optimal LB design, this thesis then proposes an optimal SCs on/off algorithm, by using the observation that the NEE decreases or first increases and then decreases as a function of the number of off SCs. By exploiting the optimal solution structure and the KKT condition, a practical suboptimal SCs on/off algorithm with linear computational complexity is also developed.Finally, due to the large transmitting power difference among various types BSs, the uplink coverage and downlink coverage will be very different. Hence, users can associate difficult BSs for uplink and downlink according to its quality of service requirements. For uplink transmission, how to reduce the EC of mobile terminal is of great importance. Considering that the real-life applications, e.g., video streaming and VoIP, are bursty traffic and delay-sensitive, different from the traditional studies that only focused on the physical layer performance without consideration of delay requirement, this thesis designs the LB and power control algorithms to minimize the EC of mobile terminal while ensure the delay requirements of each user, by taking t the dynamic traffic into account. Furthermore, the impact of non-ideal circuit power on the power control is also analysed. An EC minimization problem subject to the delay constraints is formulated, which is a no-convex problem and thus is NP-hard. To combat this issues, this thesis first studies the power control problem under fixed user association algorithm. Assuming that the full packet arrival information is known a priori at the transmitter (offline scenario), the thesis state that the optimal power control algorithm likes water-filling algorithm, where the water level will drease in the next frame when the packets arriving at D frame before are ending transmit at current frame, and the water level will drease in the next frame when the buffer is empty at current frame. According to this theorem, an uplink power control algorithm in HetNet is derived. Inspired by the optimal offline solution, this thesis also proposes a new online scheme, where only the information of the current queue backlog is available. After that, an iterative user association algorithm is proposed, where users always associate to the strongest BSs and adjust the transmit power according to target SINR in each iteration.