Research On Joint Multidimensional Resource Management In 5G Networks

With the development of smart devices and computer technology,the number of mobile devices and the traffic in cellular networks have increased at an unprecedented speed.At the same time,novel applications emerge,such as ondemand high-definition videos,pattern recognition,virtual reality/augmented reality(VR/AR)games,and so on,which are data-intensive,computation-intensive,and delay-sensitive.These applications pose more rigorous requirements to the fifth-generation mobile communication(5G)networks,such as higher data rate,lower latency,more functions,and so on.To provide massive connections and meet the demands of emerging applications,heterogeneous small cell networks,ultra-dense networks,and mobile edge computing(MEC)are introduced into 5G,which deploys small cell base stations(SBSs)in an ultra-dense method and enables cache and computation functions at the network edge to improve the network capacity,spectrum utilization,data rate and so on.With the introduction of MEC,multiple resources are needed to be efficiently managed.Since each type of resource plays an equally important role in ensuring the system performance and affects each other,it makes joint allocation of multiple resources necessary.Thus,we focus on the joint allocation of multiple resources in 5G.Considering different applications and their characteristics,we propose corresponding efficient joint multidimensional resource management schemes,which could improve the resource utilization and the users' QoS and quality of experience(QoE).The main innovations and contributions of this paper are as follows:1.In the small cell networks with energy harvesting(EH)and caching,a joint caching and user association optimization scheme for high-definition videos is proposed.First,energy harvesting(EH)and caching are introduced into small cell networks to power themselves and caching popular contents,achieving green communications.Second,we formulate the joint caching and user association problem as a joint optimization problem to maximize the number of requests that are handled by SBSs.In this way,the traditional grid power is saved.To decrease the complexity of the problem,we divide it into two subproblems,user association and content caching.Then we construct two potential games with respect to these two subproblems and a two-dimensional iteration algorithm is proposed to solve the problem.At last,simulation results show that it can efficiently decrease the energy consumption of SBSs by introducing EH and caching.Also,simulation results demonstrate that the proposed algorithm converges and reduces energy consumption.2.In the small cell networks with MEC,an efficient computation offloading scheme for computation-intensive applications is proposed.First,MEC is introduced into small cell networks to provide distributed computation offloading services for mobile users,which relieves the burden on the core network,improves users'QoS,and prolongs the battery life of mobile devices.Second,to improve the resource utilization and avoid load unbalances,we formulate the joint offloading decision,channel allocation,and computation allocation problem as a mixed-integer non-linear programming problem to minimize the energy consumption of all users.To solve this problem,a hierarchical offloading algorithm based on heuristic algorithms is proposed,where a genetic algorithm is in charge of global searching and a particle swarm optimization algorithm is for local searching.At last,simulation results show that the proposed algorithm can significantly decrease the users' energy consumption and increase the number of offloading users.3.In the small cell networks with mmWave and MEC,a joint multidimensional resource optimization scheme for wireless VR games is proposed.First,we introduce millimeter wave(mmWave)and MEC into 5G and propose an adaptive wireless VR architecture,which provides adaptive rendering and caching services at the network edge to improve the quality of experience of VR users.Second,considering the limited resources and mobility of users,the joint user association,caching,and offloading mode selection problem is formulated as an optimization problem to maximize the QoE of VR users.Considering the high complexity of the proposed problem,we propose a distributed algorithm based on machine learning and potential games,which consists of two phases,an offline training phase based on deep reinforcement learning and an online running phase based on potential game.At last,simulations show that the proposed distributed algorithm maintains scalability and adaptation capability and could increase QoE utility values,and reduces latency.