Research On Radio Resource Allocation In Non-orthogonal Multiple Access Networks And Heterogeneous Networks

The rapid development of mobile Internet and Internet of Things has put forward higher requirements for transmission rate,communication capacity and energy efficiency of future wireless communication networks.Non-orthogonal multiple access(NOMA)can increase the number of active users and system capacity by encouraging multiple users to share the same radio resource.Device-to-device(D2D)can improve spectrum efficiency and energy efficiency by allowing users to communicate directly without passing through a base station.Heterogeneous network(HetNet)can enhance coverage and improve system capacity by deploying multiple smallcells in the traditional macrocell.Therefore,NOMA,D2D and HetNet have become key technologies in future wireless communication networks.However,in the large-scale NOMA network,NOMA-D2D network and multi-layer HetNet,due to different user requirements and diverse technological scenarios,traditional radio resource allocation methods are challenging to achieve efficient radio resource allocation as well as meeting system requirements in different technological scenarios.Thus,the research on radio resource allocation for NOMA networks and HetNet has become a hot topic in wireless communication.Addressing the requirements of massive connectivity,diverse services access,large capacity and high energy efficiency,the dissertation studies radio resource allocation in NOMA networks and HetNet.This dissertation focuses on a sub-channel and power allocation algorithm based on user fairness to solve the problem of maximizing the system capacity undergoing guaranteeing user fairness in the NOMA network.This dissertation focuses on joint power allocation and spectrum sharing resource allocation algorithms to solve the problem of high system energy efficiency in the NOMA-D2D network.This dissertation focuses on a distributed power allocation algorithm based on Q-learning to improve overall performance under severe interference in the dense two-tier HetNet.The main work and innovations of this dissertation are as follows:1.Sub-channel and power allocation algorithm based on user fairnessAiming at the problem of maximizing the system capacity while guaranteeing user fairness in the NOMA cellular network with massive connectivity,large capacity and diverse services access,this dissertation proposes a sub-channel and power allocation algorithm based on user fairness.The algorithm adopts the idea of step-by-step optimization to decompose the original non-convex problem into user pairing,sub-channel allocation and power allocation sub-problems and solve them one by one.At first,a low-complexity user pairing scheme is proposed to obtain NOMA user pairs.A channel priority-based subchannel selection algorithm is proposed to realize the bidirectional selection between NOMA user pairs and sub-channels.Then,a power-allocation-factor-based construction method is proposed to realize the intra-subchannel power allocation.Finally,the successive convex approximation method is used to optimize the intra-subchannel power.The research results show that,compared with the suboptimal matching scheme for sub-channel assignment,the proposed channel priority-based subchannel selection algorithm can improve the system capacity by 3.73%.Compared with the equal power allocation method and the difference of the convex method,the proposed successive convex approximation method can improve the system capacity by 11.97%and 11.92%,respectively.Compared with the fractional transmit power allocation method and the power-allocation-factor-based selection method,the proposed power-allocation-factor-based construction method can improve the user fairness by about 23%and 6%,respectively.2.Joint power allocation and spectrum sharing resource allocation algorithmsApplying D2D technology to the NOMA cellular network can further improve the network access capability and system capacity.However,massive connectivity will lead to high energy consumption.Aiming to improve the system energy efficiency in the NOMA-D2D network with massive connectivity,this dissertation proposes a joint power-allocation and iterative-3-dimensional-matching-based spectrum sharing(JPA3DMSS)resource allocation algorithm and a joint power-allocation and greedy-3-dimensional-matching-based spectrum sharing(JPAG-3DMSS)resource allocation algorithm.The proposed JPA-3DMSS algorithm firstly adopts an iterative power allocation algorithm that combines the Dinkelbach method and the Lagrangian dual decomposition method to obtain the power of any D2D user pair and then adopts an iterative 3dimensional matching algorithm to obtain the final matching.The proposed JPAG-3DMSS algorithm adopts the iterative power allocation algorithm.It then adopts a greedy 3-dimensional matching algorithm for spectrum sharing,converting the optimal local solution of the iterative 3dimensional matching algorithm into the overall energy efficiency optimization to solve the maximum matching problem of the iterative 3dimensional matching algorithm.The research results show that,compared with the random resource allocation algorithm and the energy-efficient joint resource block and power allocation algorithm,the JPA-3DMSS algorithm can improve the system energy efficiency by 39%and 5%,respectively,and the JPAG-3DMSS algorithm can improve the system energy efficiency by 44%and 9%,respectively.3.Distributed power allocation algorithm based on Q-learningAiming at the problem of system performance improvement under severe interference conditions in the dense two-tier HetNet,this dissertation proposes a distributed power allocation algorithm based on Qlearning.The algorithm leverages a distributed learning framework based on multi-agent Markov decision process,adopts the ε-greedy strategy to select actions,and adopts a new product reward function to update Q tables.When a new femtocell base station joins the network,the new femtocell base station adopts an independent learning strategy or cooperative learning strategy to learn.As shown in the results,under the cooperative learning strategy,the proposed algorithm can improve the total capacity of femtocell users by 16.4%,27.6%and 22.4%,compared with the Qlearning algorithms based on quadratic reward function,exponential reward function and proximity reward function,respectively.It can also reduce the total power consumption of femtocell base stations by 56.2%,36.6%and 14.5%,respectively while ensuring the quality of service and fairness of all users.