Research On Resource Allocation In Heterogeneous Cellular Networks

With the rapid development of communication systems,the 5 G era is coming.Traditional cellular network architectures are no longer efficient under the tremendous pressure of increasing demands of data rates and large amounts of access devices.The heterogeneity of network keeps rising and there will be multiple wireless access technologies in the future networks.To address the aforementioned challenges in the future network architectures,this dissertation deals with the resource allocation problems in two scenarios under heterogeneous cellular networks.The first is the resource allocation and user association problem under multi-layer heterogeneous cellular networks.Most of the current literatures provide little coverage on dynamic frequency assignment between base stations.The second is the resource allocation problem under Device-to-Device(D2D)and cellular fusion heterogeneous networks.The main problem is that there are few studies on the applications of social relations in D2D.In addition,there is a lack of joint solution of mode selection,power control and channel allocation in D2D.The main contributions of this dissertation can be summarized as follows.1.In the multi-layer heterogeneous cellular networks,a resource allocation algorithm is proposed based on automatic differentiation.In the proposed algorithm,the mathematical model is established under the joint consideration of resource allocation and user association.The logarithm of the user rates is selected as the optimization target of the problem.This problem is a mixed integer optimization problem,which is proved to be an NP-hard problem and is hard to get a close-form solution.In this dissertation,automatic differentiation is used to solve the problem.First,penalty functions are designed to relax the constraints of the original problem.Penalty function method is taken to connect the objective function of the original problem with the constraint condition so the initial constraint problem is turned to be unconstrained.Specifically,the penalty functions transform the objective function and constraints into a loss function of the dynamic calculation graph,which can be solved by automatic differentiation.After that,the gradient descent algorithm is adopted to iteratively obtain the feasible solution.The result includes not only the resource allocation and the user association scheme but also the frequency reuse scheme.The simulation verifies the convergence and effectiveness of the proposed algorithm.Compared with other algorithms,it can adaptively obtain frequency reuse scheme and improved the system performance.2.Under the D2D and cellular fusion heterogeneous network,social relations are introduced.A resource allocation algorithm that jointly solves mode selection,channel allocation and power control is proposed.First,the model of resource allocation problem is established,where the throughput of the whole system is proposed as the objective function.This problem is also proved to be an NP-hard problem which can be divided into three sub problems.The first is mode selection.A mode selection algorithm based on the social relations is proposed.Second,the optimal power allocation on the fixed channel is solved based on the geometric programming method.Finally,the channel assignment problem is solved using the maximum weight matching of the bipartite graph.An improved IPPO algorithm is proposed.The simulation shows that the proposed algorithms can achieve a good balance between system performance and computational complexity compared with the KM algorithm.The results also show that social relations have significant influence on improving the throughput of the system.