Research On Multi-Access Edge Computing Based Task Scheduling Strategy In IIoT Networks

The Industrial Internet of Things(IIoT)is a central enabler that pushes the industries from traditional mode to intelligent mode.As 5G deeply empowers vertical industries,IIoT enters the stage of rapid development.At the same time,the number of intelligent industrial devices explodes exponentially.And computation-intensive and time-sensitive applications increase rapidly.However,the limited computing power and battery capacity of IIoT devices make it difficult to meet the service requirements of those computation-intensive and time-sensitive tasks.In IIoT enabled with Cloud computing(CC),the tasks are offloaded to cloud servers with large computing power,which solves the above problems effectively.In the traditional CC mode,a large amount of data is offloaded to the cloud server,which leads to network congestion.In addition,data transmission in long distance increases the communication delay of tasks processing,which cannot meet the requirements of delay-sensitive tasks.In this context,Multi-access edge computing(MEC)is proposed and widely used in IIoT systems.In the MEC-based IIoT system,the devices can offload tasks to small servers at the edge of the network to obtain nearby computing services,where the task processing delay and device energy consumption are reduced,and the task execution efficiency is improved.However,the dynamic channel state and random arrival of tasks in the MEC-assisted IIoT system would greatly bring challenges to the design of task offloading and resource allocation strategy.The uneven workload of multiple MEC servers and the low utilization of resources impact the strategy design as well as.To overcome the above challenges,this thesis first proposes an environment awareness based real-time task scheduling strategy in single-edge computing networks.This strategy minimizes the long-term task processing delay of the system through jointly optimizing the task offloading and computation resource allocation.On this basis,an edge cooperative based task scheduling strategy is proposed in multi-edge computing networks.This strategy reduces task execution costs of the system through jointly optimizing the task segmentation,computation offloading and communication resource allocation,where the resource utilization rate is effectively improved.The main contributions and innovations of this thesis are summarized as follows:(1)An environment awareness based real-time task scheduling strategy in single-edge computing networks is proposed.Considering the dynamic channel state and random task generation in IIoT scenarios,a long-term task scheduling strategy of the system is studied.A joint optimization problem of task offloading and MEC server’ computation resource allocation is constructed to minimize the long-term task processing delay of the system,where the task scheduling decision is made in the unit of time slot.In order to solve this problem,which has huge state space,and the decision space contains continuous and discrete variables,this thesis models it as Markov decision process,and adopts deep reinforcement learning method for the strategy designing.Then,a deep deterministic strategy gradient based real-time task scheduling(DDPG-RTS)algorithm is proposed,which can make the optimal decision with real-time system state information.Simulation results confirm convergence characteristics of the proposed algorithm,which can significantly reduce the task processing delay.(2)An edge cooperative task scheduling strategy in multi-edge computing networks is proposed.In the MEC-assisted IIoT scenarios with unevenly distributed MEC network resources,an edge cooperative task scheduling strategy is designed for partial offloading tasks.Through the edge collaboration method,the tasks generated by the devices can be offloaded to other lightly loaded edge servers instead of the local edge.Then,the cost function of the system(weighted by delay and energy consumption)is defined,and a joint optimization problem of task segmentation,target edge server selection,wireless bandwidth resource allocation and wired communication link capacity allocation was comprehensively analyzed to minimize the system cost.To this end,a double delay depth deterministic policy gradient based edge cooperative task scheduling(TD3-ECoTS)algorithm is proposed.Simulation results show that the proposed algorithm can effectively reduce the cost of task processing and achieves efficient use of network resources.