Research On Intelligent Task Scheduling And Resource Allocation Methods In Mobile Edge Networks

The development of computation-intensive applications such as high-speed mobile access,augmented reality(AR),virtual reality(VR),and super-scale Internet of Things(IoT)has put forward higher demands on network delay,access bandwidth,and com-puting power.Traditional Mobile Cloud Computing(MCC)offloads computing tasks to remote data centers,which makes it difficult to guarantee the processing delay of the entire task.,Therefore Mobile Edge Computing(MEC)has emerged.With the MEC technology,servers are deployed at the edge of mobile networks closer to users to meet the demand for high computing power and low delay of tasks.However,the bandwidth of wireless access and computing resources of servers are limited,and unreasonable scheduling allo-cation will lead to waste of system resources,degradation of service quality,and increase of energy consumption.Therefore,this paper focuses on intelligent control algorithms in mobile edge network to achieve intelligent scheduling allocation of tasks.The offloading decision of tasks and the reasonable allocation of resources in the MEC scenario have an important impact on the energy consumption and task computation processing delay of the whole system.The offloading decision is designed to determine whether the computing tasks are processed locally or transferred to the MEC server for processing,and the allocation of resources includes the allocation of bandwidth resources and the allocation of computing power resources.There are usually two difficulties in the the problem of task offloading decision and resource allocation:(a)the problem can usually be modeled as a nonconvex optimization problem,and it is difficult to compute the optimal decision scheme in a short time using traditional methods;(b)most of the traditional algorithms assume that the offloaded tasks are transferred in parallel,and the computational resources may be in idle state during the transfer,resulting in a waste of resources.In this paper,a batch offloading model was studied,where tasks that need to be offloaded can be transferred to the server in a certain batch.Then the NNGA algorithm was proposed,which assumes that the server follows a single-service-desk queuing model,and obtains the offloading decision for each task through neural networks and genetic algorithms,followed by bandwidth allocation using a heuristic algorithm for the tasks transferred within each batch.The simulation results show that the algorithm can obtain better solution results in a shorter time compared to other comparative algorithms.The DASH(Dynamic Adaptive Streaming over HTTP)video streaming is an adap-tive streaming technology that can divide the content of a video file into small segments and encode the segmented videos with different bit rates,so that users can request the corresponding bit rate segments for their own network conditions.With the explosive growth of video services,the problem of intelligent transmission of DASH-based video streams has become another research hotspot in mobile edge network.In the Chapter 4 of this paper,the intelligent transmission problem of video streaming in a multi-base station scenario was studied,including two sub-problems of users'base station access decision and adaptive bitrate selection.However,it is difficult to predict the channel conditions in actual scenarios,and it is more difficult for the base station to collect the cache length in-formation of all terminals in real time.In this paper,considering the distance information between users and base stations,the distance path loss and the bandwidth resource limita-tion of each server,a multi-user bit rate allocation scenario with multiple base stations was modeled,and the MBRA algorithm that can implement self-learning was designed to max-imize the weighted sum of the user's overall Quality Of Experience(QoE)utility and the system's remaining bandwidth resources.The original problem was divided into two pro-cesses:(a)the better base station access decision scheme was obtained through the DNN network output decision and the KNN nearest neighbor algorithm better solution search;(b)the bitrate selection of users under a single base station was implemented based on the greedy algorithm.The simulation results show that the MBRA algorithm can achieve better solution results in a smaller time complexity range compared with other algorithms.