Research On Task Scheduling Strategy Under 5G Ultra-dense Access Network

With the rapid development and official commercialization of the fifth-generation mobile communication technology,mobile users and mobile data traffic have shown explosive growth.The ultra-dense network architecture is considered to satisfy the requirements of massive access and explosive growth services,which is formed by intensively deploying low-power small base stations in hotspot areas.Facing the differentiated Qo S requirements for different tasks,such as video,voice,virtual reality/augmented reality,autonomous driving,in the mass access environment,it is a key of the reasonable access strategy of mass users and the heterogeneous task scheduling strategy in the network time and frequency domain to guarantee the performance of the network system.Firstly,we consider the network selection problem of users under multi-signal overlapping coverage in the 5G ultra-dense access network.We formulate the network selection problem as an MDP by analyzing the different requirements of Qo S indexes such as rate,delay,and packet loss rate for video and voice tasks which will be transmitted by users,and considering the various network parameters such as available resources,security,service costs in ultra-dense networks.And then,we focus on the explosion of MDP state-space caused by multi-type tasks and multi-attribute decisions,and we propose an NS-MDP algorithm to optimize the state space and return function of the MDP model.In order to obtain the optimal scheduling strategy,we propose the dynamic programming value iteration method.The optimal scheduling policy can not only satisfy the Qo S requirements of each user's task but also reduce the switching rate and blocking rate of the user.Secondly,we study the heterogeneous task scheduling problem in e MBB and URLLC coexisting scenario.We analyze the low latency and high-reliability characteristics of URLLC tasks,and introduce the flexible numerology structure which defines a set of flexible transmission time intervals(TTI)to satisfy different Qo S requirements of heterogeneous services.Then,we formulate a Markov decision process(MDP)-based throughput optimization problem with the flexibilities of time and frequency domains.And,we prove this optimization problem to be NP-hard and propose an innovative joint scheduling strategy HTSA based on flexible numerology and deep reinforcement learning method to schedule the heterogeneous task whichcan fastly obtain the approximate solution of the np-hard problem.The algorithm proposed can guarantee Qo S requirements for the URLLC task and improve the resource utilization of time and frequency domains.Finally,we conduct the simulation experiments to evaluate the performance of the NS-MDP network selection algorithm and HTSA heterogeneous task scheduling algorithm.Through experiments,the NS-MDP algorithm can effectively reduce the user's switching rate and blocking rate,and improve the throughput of the ultra-dense access network system.And,the HTSA algorithm can effectively reduce the loss rate of URLLC tasks,ensure the low latency and high reliability of URLLC tasks,and improve the utilization rate of time and frequency domain.