Research On Wireless Resource Management Of Mobile Edge Computing In B5G Networks

With the full deployment of the 5th Generation Mobile Communication System(5G),the discussion and research on Beyond 5G(B5G)and the 6th Generation Mobile Communication System(6G)are gradually deepening.With the access of massive devices,the future challenges of B5 G and 6G networks come from high data rates,high system capacity and high communication costs.Mobile edge computing(MEC)technology delivers computation capabilities to mobile devices,which lack computing and energy resources.Thus,MEC technology will greatly reduce the computing energy and processing delay,while relieve the pressure of the core network caused by massive data transmission.Therefore,MEC is considered to be a promising paradigm in the B5 G and 6G and studied widely.In addition,B5 G and 6G put forward more rigorous requirements for mobile devices.New types of applications and complex tasks running on the terminal devices makes the energy consumption of mobile devices increasing gradually.By the energy harvesting(EH)technology,devices can collect energy from the environment,effectively prolonging batteries’ lives and reducing energy limitations of user equipment.In summary,to solve the problem caused by massive mobile device access in B5 G and 6G scenario,EH are applied into the MEC system.Focus on the EH based MEC system,we mainly study the wireless and computing resource allocation in computing offloading and data analysis scenarios.The main contributions and innovations of this paper are shown as follows.Firstly,we consider a multi-user offloading MEC scenario with EH devices.To prolong the batteries’ life time with low energy consumption,we minimize the long term energy consumption constrained by the processing delay,communication resources and computing resources.Due to the randomness of channel gains,task arrival and energy arrival,we allocate the limited resources by the deep reinforcement learning algorithm.Secondly,in the computing offloading scenario,to obtain a resource allocation scheme with continues transmission power,we use the deep deterministic policy gradient(DDPG)framework based on deep reinforcement learning,which overcomes the issue of dimensional disasters and slow convergence.Under the settings of simulation parameter,the result shows the resource allocation algorithm based on DDPG can effectively reduce the system energy consumption.Thirdly,we propose a data analysis scenario with an MEC system,where wireless devices collect data from the environment and an MEC server computes and analyzes the data centrally.To upload the collected data,wireless devices harvest energy by the EH technology.Consider the requirement of AoI(Age of Information),we maximize the long term average utility throughput under the constraints of required AoI.Finally,in the data analysis scenario,to solve the long term average problem,we proposed a Lyapunov optimization based algorithm.By transforming the AoI constraint into a virtual queue,we rewrite the original problem into a new form which can be solved by Lyapunov optimization.Based on our proposed algorithm,we decouple the original problem into four subproblems,which can optimize the data discarded and the collection decision,wireless resource allocation and computing resource allocation independently.The simulation result shows the effectiveness of the proposed algorithm in improving the system throughput under AoI constraints.