Resource Allocation In Intelligent Reflecting Surface Aided Mobile Edge Computing

The era of 5G communication is the era of the Internet of things(IoT).Hundreds of millions of IoT devices are connected to the Internet and continue to work as functional nodes.The common IoT devices include road monitoring,bike-sharing control modules,traffic flow sensors and wearable devices.Due to the small size of IoT devices,the energy capacity and computing resources are often scarce.However,with the functional requirements of IoT devices,there are often computation-intensive or time-delay sensitive computing tasks to be processed on the IoT devices.As a new computing paradigm,mobile edge computing(MEC)technique allows users to offload task information to the MEC servers,and then get the processed data return from the MEC servers after calculation.Intelligent reflecting surface(IRS)is a novel and revolutionising technology that is able to significantly improve the performance of wireless communication systems.In practical,the IRS is consisted of three parts,IRS controller,surface pad,and several reflecting elements.The IRS can dynamically adjust the wireless propagation channel by numerous passive scattering elements integrated on the surface.Each scattering element is able to reflect the incident electromagnetic wave with amplitude and/or phase shift through a controller inside the circuit.Therefore,the IRS-assisted wireless network is able to greatly increase the channel capacity,with extremely low power consumption and a high degree of flexibility for large-scale deployment.For this reason,the IRSassisted wireless network becomes a hot research topic in the next generation wireless communication systems.In this dissertation,we consider a wireless powered MEC network that is equipped with an IRS,with the aim of optimizing the users' system utilization by dynamically adjusting wireless network resource allocation.In the system model,it is assumed that the channels between users and the base station are non-line of sight channel with poor communication quality,while the reflecting channel aided by the IRS can improve the communication rate of the overall system.By jointly considering the energy efficiency,time latency,and channel state information,an optimization problem including the phase state of the IRS,offloading and time allocation is formulated.By exploiting the characteristic of the optimization problem,two efficient algorithms are designed,which are the discrete policy optimization algorithm based on the deep reinforcement learning with discrete action space,and the contineous polity optimization algorithm based on the joint optimization theorey and deep reinforcement learning with contineous action space,respectively.Numerical results are provided to verify the performance of the proposed algorithms compared with several benchmark algorithms.