Theories And Methods Of Communication And Computing-Integrated Resource Allocation In Wireless Networks

In order to meet the rapid development of the mobile Internet and the Internet of Things,as well as the high-quality service requirements of the emerging computation-intensive and resource-intensive applications,the wireless networks need to evolve towards the direction of communication and computing convergence.However,the traditional wireless networks encounter problems such as the difficulties in unifying the architecture and coordinating the resource and the low networking efficiency when integrating and collaborating.Therefore,how to use the integration of communication and computing to achieve the flexible adaptation and resource collaboration for various scenarios and services remains a hot and difficult issue in the current research.For this reason,this dissertation focuses on the edge computing network,the fog wireless access network(F-RAN)and the unmanned aerial vehicle(UAV)-enabled communication network,respectively,and gradually launches the research on resource allocation theories and methods from the integration of communication and edge computing to the cloud edge computing.Firstly,to deal with the challenge of low resource utilization caused by the coupling and restriction between communication and computing,a joint communication and computing resource allocation scheme for the non orthogonal multiple access(NOMA)-enabled edge computing networks is proposed,which uses the interaction between the communication and computing resources to improve the offloading efficiency;Secondly,to deal with the challenges of complex networking modes and difficult performance analysis and optimization in F-RANs,the cloud and fog accessing modes are considered,and an optimal outer block approximation-based resource allocation scheme and a suboptimal resource allocation scheme that balances the performance and complexity are proposed to effectively improve the spectral efficiency;Finally,to deal with the challenge that the deployment and resource allocation in the UAV-enabled communication networks cannot autonomously adapt to the diverse service requirements,as well as the challenge of the synchronous motions and dynamic resource scheduling difficulties caused by the dynamically changing network topology,a joint optimization scheme for location deployment and resource allocation in the UAV-enabled communication networks that combines the advantages of convex optimization and deep reinforcement learning(DRL)is proposed to balance the consumption of latency and energy,and a dual timescale-based joint optimization scheme for trajectory planning and resource allocation in the UAV-enabled communication networks is proposed to improve the system throughput.The main research content and contributions of this dissertation are summarized as follows:1.Communication and computing-converged resource allocation for the edge computing networksTo deal with the challenge of low resource utilization caused by the coupling and restriction between communication and computing,this dissertation proposes an efficient joint communication and computing resource allocation method,which utilizes the interaction between resources to improve the offloading efficiency.Firstly,an offloading scenario in the NOMA-based massive IoT is modeled,and a joint optimization problem of subcarrier,power,and computing resource allocation is formulated to maximize the energy efficiency(EE),while simultaneously ensuring the delay constraints of the devices.Then,considering the non-convex nature of the optimization problem,it is further decoupled into three independent subproblems,in which the closed-form solution of power is derived,and the matching,Lagrange dual and Knapsack algorithms are utilized to solve these subproblems iteratively until convergence.Finally,simulation results show that the proposed algorithm can effectively improve the EE by 61%compared to the traditional orthogonal multiple access(OMA)scheme.In addition,the simulation results also show that the system performance is limited by the restricted resource,showing how the maximum EE can be achieved by judiciously matching the communication resources with computation resources.2.Communication and computing-converged resource allocation for the F-RANsTo deal with the challenges of complex networking modes and difficult performance analysis and optimization in F-RANs.this dissertation considers the cloud and fog accessing modes.and proposes an optimal and a suboptimal communication and computing-converged resource allocation schemes to effectively control the interference while improving the spectral efficiency.Firstly,the downlink transmission scenario in F-RAN is modeled,and a joint optimization problem of subcarrier and power allocation is formulated to maximize the weighted sum rate of users in the fog accessing mode,while simultaneously ensuring the quality of service for users in the cloud accessing mode.Then,to solve the optimization problem,an outer block approximation algorithm based on monotone optimization is proposed to obtain the global optimal solution,and a suboptimal algorithm based on swapenabled matching and sequential convex programming(SCP)methods is proposed to balance the performance and computational complexity.Finally,simulation results verify the performance of the proposed algorithms,in which the suboptimal scheme achieves nearly 92%of the performance of the optimal scheme,while reducing the computational complexity from the exponential level to the polynomial level.3.Communication and computing-converged resource allocation for the UAV-enabled communication networksTo deal with the challenge that the deployment and resource allocation in the UAVenabled communication networks cannot autonomously adapt to the diverse service requirements,this dissertation proposes a joint optimization scheme for location deployment and resource allocation in the UAV-enabled communication networks that combines the advantages of convex optimization and deep reinforcement learning(DRL)to balance the consumption of latency and energy.Firstly,the uplink and downlink transmission scenarios in the UAV-enabled communication network are modeled,in which the edge UAVs cooperate with the cloud to provide users with timely and all-around cache transmission and computation offloading services.Then,a joint optimization problem for the caching and offloading decision-making,UAV deployment,and communication and computing resource allocation is formulated to minimize the weighted system cost of the delay and energy consumption,while simultaneously satisfying the UAVs’ cache and computing capacity constraints as well as users’ delay and energy consumption constraints.To solve the nondeterministic polynomial(NP)-hard problem,a SCP and sequential quadratic programming methods-based deep Q network(DQN)algorithm is proposed,which allows the system to adaptively adjust the caching and offloading decisions based on the UAV deployment and resource allocation schemes obtained by the convex optimization algorithms.Simulation results validate the advantages of the proposed algorithm,and demonstrate that the combination of the convex optimization and deep reinforcement learning can significantly improve the system performance and training efficiency,with a 38%reduction in the cost and a 36%improvement in the convergence speed.To deal with the challenge of the synchronous motions and dynamic resource scheduling difficulties caused by the dynamically changing network topology,this dissertation considers the timescale differences in resource optimization and proposes a dual timescale-based joint optimization scheme for traj ectory planning and resource allocation in the UAV-enabled communication networks to improve the system throughput.Firstly,the downlink transmission scenario of the UAV-enabled wireless caching network is modeled,and a joint optimization problem of the user association,UAV cache placement,and trajectory planning is formulated to maximize the system throughput,while simultaneously meeting the UAVs’ cache capacity and kinematic constraints.Then,considering the timescale differences among variables,a hierarchical two-timescale resource management architecture is proposed.Specifically,the outer layer is based on the double DQN algorithm to optimize the cache content placement in the large timescale,in which the double Qlearning is introduced to eliminate the overestimation,and the cache placement can be dynamically adjusted based on the historical user requests.The inner layer is based on the improved multi-agent deep deterministic policy gradient algorithm to optimize the user association and UAVs’ trajectory planning in the short timescale,in which the attention mechanism is introduced to extract the key information as inputs to achieve more efficient performance in the complex multi-agent environment.Simulation results validate the superiority of the proposed algorithm,and demonstrate that a well-designed cache placement strategy can effectively improve system throughput by 18.6%.This dissertation focuses on the edge computing network,the F-RAN and the UAVenabled communication network,respectively,and gradually launches the research on resource allocation theories and methods from the integration of communication and edge computing to the cloud edge computing.The relevant research results can provide theoretical guidance for the resource management and performance optimization of the communication and computing-integrated wireless networks.