Research On Task Scheduling Algorithm Of Mobile Edge Computing

With the continuous development of mobile communication technology,there are more and more wireless devices connected to the Internet,such as mobile phones,sensors,vehicles,robots,etc.,which have produced many computationally intensive and real-time applications,such as autonomous driving,online games,pattern recognition,etc.However,due to the limitation of the physical size of the wireless device,it only has limited computing resources and battery capacity,which cannot meet the requirements of real-time of applications and low energy consumption of the wireless devices at the same time.In addition,the cloud server has powerful computing capabilities.However,the physical distance between wireless devices and cloud servers is long,and there are large delays and environmental interference problems in the transmission of application data,which cannot meet the real-time requirements of these applications.In response to the above problems,mobile edge computing(MEC)migrates the computing power of cloud computing to the edge of the network,and is closer to the wireless device in physical location.The wireless device transmits some tasks to the edge server for execution,achieving low latency,high bandwidth,low energy consumption and obtaining excellent service quality.How to efficiently coordinate wireless devices and edge servers to perform computing in mobile edge computing is a vital research direction.Among them,task scheduling is the core issue that determine system performance and user experience quality.In-depth research on the task scheduling problem in mobile edge computing,starting from the two focuses of user task computing offloading and MEC server load balancing,the following two aspects of research work are carried out:First of all,aiming at the problem of task computing offloading in mobile edge computing,a mobile edge computing scenario with multiple users,multiple tasks and multiple MEC servers constructed.Based on this scenario,the delay and energy consumption of computing offloading are comprehensively considered to minimize the system overhead.To solve this problem,an online computing offloading method based on multi-agent reinforcement learning(MAOO),is proposed.Each wireless device is treated as an agent,which can effectively reduce the complexity of state space and action space.At the same time,a collaborative optimization mechanism between wireless devices is introduced to ensure the stability of the environment and improve the algorithm convergence speed.The simulation results show that MAOO has an advantage over other benchmark algorithms in reducing system overhead.Secondly,aiming at the problem of MEC server load balancing in mobile edge computing,considering running tasks on edge servers in the form of edge containers to achieve dependency separation,resource isolation.It's easy to migrate and deploy task according to the system load status.At the same time,multi-dimensional resource(CPU,memory,and network bandwidth)conditions are used as a measure of system load.To avoid hot issues in the system,the optimization goal is minimizing the average system load.Based on the above problems,a method based on deep reinforcement learning(Edge Container Load Balancing)is proposed,to achieve optimization goals of system load balancing.In addition,in order to improve the algorithm's exploration of the action space,an action generation mechanism based on the k-nearest neighbor algorithm introduced.The experimental simulation results show that the ECLB algorithm can reduce the average load of the system according to the usage of multi-dimensional resources.