Study On User Fairness Guarantees In Cellular Networks

With the growth of demand for wireless communications, cellular communication systems have undergone the evolution from the first generation to the fourth generation. History demonstrates that cellular networks always play the most important role in providing wireless communication services in anytime and anywhere. By now, the investigation on the fifth generation mobile communication systems has already begun. The goal of 5W in wireless communications requires that everyone can get its needed communications. Therefore, user fairness must be guaranteed to make sure that everyone receives the same treatment, such as throughput, delay, call blocking probability.User fairness in the whole network has two layers of meaning, i.e., intra-cell user fairness and inter-cell user fairness. The nonuniform distribution of traffic results in high call blocking probability and low user throughput in hot cells, which is referred to as inter-cell unfairness problem. On the other hand, intra-cell unfairness problem is often caused by inappropriate slot and channel allocations. Moreover, as system efficiency and user fairness is reversely related as one waxes the other wanes, user fairness guarantees should also take into account system efficiency. As a result, there are three key issues in guaranteeing user fairness in cellular networks: how to utilize efficient load balancing schemes to improve inter-cell user fairness, how to design intra-cell resource allocation strategies to guarantee intra-cell user fairness, and how to combine the former two aspects to efficiently trade off system efficiency and user fairness. This dissertation mainly focuses on guaranteeing intra-cell and inter-cell user fairness in cellular networks. We firstly investigate the load balancing problem to provide inter-cell fairness. Then, intra-cell user fairness is guaranteed based onα-fairness and Jain’s index. Finally, we consider inter-cell fairness in conjunction with intra-cell fairness in heterogeneous cellular networks. The main achievements and results of this dissertation are listed as follows:1. Traditional cell selection based load balancing schemes only implemented twocell cooperation, which leads to high call blocking probability in hot cells. To address this inter-cell user unfairness problem, this dissertation proposes a cell-cooperationchain based load balancing strategy by multi-cell cooperation. The proposed strategy helps the users utilize the network resource as efficiently as possible, improves the resource utilization efficiency, and thus decreases call blocking probability of hot cells and the system. In order to decrease the cooperation cost, we consider the minimumlevel cell-cooperation-chain selection problem, which is formulated as a shortest path selection problem in graph theory. Therefore, this dissertation designs an efficient load balancing algorithm based on BF algorithm. In addition, we prove that the proposed algorithm can achieve the lowest system call blocking probability and present the lower bound of the system call blocking probability by using multi-dimensional Markov chains. Simulation results show that the cell-cooperation-chain based load balancing method can decrease hot cell and system call blocking probability and enhance user fairness of the network.2. In order to efficiently trade off efficiency and fairness while guaranteeing a certain fairness degree, we discusses in detail the α-fairness about its tradeoff between efficiency and fairness from the perspective of Jain’s index in single cell system over slow fading channels. By solving the corresponding α-utility maximization problem,we get the expressions of system efficiency and Jain’s index. Further, we conclude that Jain’s index increases with α while system efficiency decreases with α. According to these results, a fast converging α selection algorithm is designed to make sure that α-fairness can provide the required Jain’s index or system efficiency. In addition,in soft-frequency-reuse based macro-only cellular networks, we also optimize user association and intra-cell scheduling jointly to achieve network α-fairness.3. The efficiency-fairness tradeoff problem is investigated, where there are certain requirements in both short-term fairness and long-term fairness. In particular,we consider slot and subchannel allocation problem in single cell downlink OFDMA systems, whose constraints are composed of short-term and long-term Jain’s index requirements and objective is to maximizing system throughput. We solve this discrete NP-hard problem by relaxation, optimization and rounding. The analysis also indicates that the relaxation and rounding process does not bring about large errors.Moreover, the relation between short-term and long-term fairness is discussed in detail. We point out that short-term fairness does not always imply long-term fairness.Specifically, it depends on the fairness definition and the averaging method, where the averaging method indicates the way we average the short-term performance metric to obtain the long-term one.4. In heterogeneous cellular networks, different tiers of base stations have different coverage area, which results in load imbalance problem. Meanwhile, the macro base stations produce great interference to the users in small cells. These two issues result in the user unfairness situation in the network. Therefore, this dissertation jointly optimizes user association, enhanced inter-cell interference coordination(eICIC), power control and intra-cell resource allocation to achieve global user proportional fairness, where eICIC is achieved by intermittently muting macro BSs, known also as resource partitioning. As the formulated problem is a multi-dimensional combinational non-convex optimization issue, we propose a univariate search based iterative algorithm. Particularly, in each iteration, we first optimize resource partitioning, user association and intra-cell resource allocation together, and then optimize power.Simulation results show that the proposed policy greatly improves user fairness and system efficiency.