Research On Mobile Edge Computing Task Offloading And Resource Allocation In Multi-scenarios

With the rapid development of artificial intelligence and Internet of things,intelligent devices and many new applications have been popularized on a large scale.However,intelligent terminal devices are limited by their own battery capacity and computing resources,so it is difficult to carry the needs of computing-intensive tasks.Mobile Edge Computing(MEC)makes up for the above shortcomings by deploying MEC servers on the edge side to provide computing resources to devices.However,there are still performance bottlenecks in large-scale MEC systems,so it is necessary to make reasonable decisions on task offloading and resource allocation.Currently,the research on task offloading and resource allocation of mobile edge computing is limited,ignoring the characteristics of the real scenario,such as the dynamics of the task itself in the time dimension,the change of the location of the terminal device and the time-varying fading of the channel in the scenario.Based on the shortcomings of the current research scenarios,this paper enriches the research scenarios of mobile edge computing task offloading and resource allocation,mainly in the following aspects.(1)Based on the consideration of task dynamics,a model of task offloading and resource allocation in dynamic task scenario is proposed,and a Bidirectional Neighborhood Search Algorithm based on Adaptive Selection(BNSAS)is designed to solve the problem.Meanwhile,the generating state of task is simulated based on Poisson distribution.The simulation results show that BNSAS is effective for dynamic tasks and single batch tasks.(2)Based on the consideration of the dynamic characteristics of terminal devices,a model of task offloading and resource allocation in device mobility scenario is proposed,and a BNSAS Integrating Location Information(BNSAS-ILI)is proposed,which integrates location information.Based on the location information of devices,different neighborhood search operators are designed to improve the search ability of the algorithm under this model.The simulation results show the effectiveness of BNSAS-ILI and its neighborhood search operator based on location information.(3)Based on the consideration of channel fading in MEC networks,a model of task offloading and resource allocation in time-varying fading channels is proposed.In this model,the wireless charging technology is used to power the terminal device and support the calculation and offloading task of the device.Meanwhile,a Deep Reinforcement Learning Algorithm based on LSTM(DR-LSTM)is designed to solve the model.The simulation results show that the DR-LSTM can effectively solve the problem of task offloading and resource allocation in time-varying fading channel scenario.