Research On Access Selection And Resource Allocation In Heterogeneous Networks

With the rapid development of communication technology, the conflict between the growing traffic demand and limited resources becomes increasingly. How to match the demand for resources with the supply of resources has become a hot topic. The future mobile communication networks will become dense, organic, and irregular heterogeneous networks (HetNets) and the user terminals will be covered by a variety of wireless networks. Under this background, matching terminals’demand with resource allocation means which network terminals should be access to for a more efficient use of resources, which is named as the network access selection. How to make reasonable and effective access selection in such complex heterogeneous networks to improve system performance and meet users’requirements of service of quality (QoS) has important significance.The methods to access selection can be divided into two categories:network-centric methods and user-centric methods. In network-centric methods, network selection, also named as user association, is usually considered together with resource allocation. Most existing investigations assumed that frequency reuse schemes in these systems are static and unchanged, which is no longer suitable for the irregular, overlapping cells. In the user-centric methods, terminals make decision to select best networks with terminal capabilities, service requirements, user preferences, network coverage, network loads as well as price taken into consideration. Although a variety of mathematical models has been proposed, the feature of services, the weights of attributes and terminals’ characteristics are not considered comprehensively enough. This thesis focuses on problems mentioned above in these two methods and have a further study of the heterogeneous network access selection problem.A network-centric user association and resource allocation algorithm is proposed for heterogeneous hyper cellular networks. First, the user association and resource allocation model is formulated as an optimization question with the objective of maximizing the log sum users’ rate, which makes a tradeoff between system throughput and user fairness. The joint optimization on these two problems is considered as NP-hard and hence it is difficult to achieve the exact solution effectively. The optimization problem on the whole networks is divided in two subproblems and a two-stage resource allocation scheme is designed:(1) resource allocation for each base station (BS) and (2) user association. In stage one, based on the priori traffic spatial distribution, the system resource is pre-allocated to each cell BSs with the objective of improving the frequency reuse. As the subproblem one is a complex integer programming problem, a heuristic allocation algorithm is proposed to reduce the complexity. In stage two, each user is associated with a BS and assigned some physical resource blocks (PRBs) from cell BSs. The subproblem two, a mixed integer non-convex optimization problem, is converted into a relaxed convex optimization and solved by Lagrange dual function method. Numerical results show that the approach takes a tradeoff between network throughput and user fairness, and achieves load balancing across macro and small cell BSs at the same time.A user-centric vertical handoff decision algorithm is proposed for the heterogeneous wireless networks, which combines the Markov Decision Process (MDP) with analytic hierarchy process (AHP) method. First, the decision factors such as signal strength, quality of service (bandwidth, delay, jitter and packet loss rate), network costs, user’s velocity and power consumption, which determine the choices of the wireless networks, is discussed in detail. Then the reward function of each decision factor is defined according to their characteristics. The analytic hierarchy process is used to assign the weight to every attribute according to the requirements of diverse types of service. Finally, The value iteration algorithm (VIA) is used to determine the optimal policy and the corresponding expected total reward. Numerical simulations of two different scenes show that the proposed algorithm can achieve higher expected total reward over other algorithms and effectively reduce witching times.