Research On Industrial Multi Terminal Computing Offload Based On Deep Reinforcement Learning

In the process of traditional industrial workshop equipment monitoring and operation,it is difficult for edge equipment to process the data of workshop equipment efficiently and in real time.The complete Internet of Things(Io T)system provides a more practical service for the industrial scenario,where the tasks generated by large heterogeneous terminals are becoming more and more diverse,and the limitations of terminal resources result in unmet computing needs.To handle industrial tasks more efficiently,edge computing provides technologies such as computational offloading and resource allocation that interact with the environment and enhance learning to make good decisions in uncertain environmental conditions.Deep reinforcement learning combines the advantages of deep learning and reinforcement learning that can be used in decision making under complex environmental conditions.This thesis takes the computational-intensive industrial scenario as the research object,constructs the environment of computing and unloading,and proposes a multi-terminal task unloading technique based on deep reinforcement learning.The work of this thesis is as follows:In order to solve the problem of poor task service quality caused by the unreasonable allocation of task resources in the time-constrained scenario,a multi-task resource allocation scheme based on Deep QNetwork(DQN)is proposed.In this thesis,the influence of the resource allocation scheme of edge server on the result of task processing is studied.At the same time,the allocation rules of computing resources need to be related to the task and as many tasks as possible need to be completed within the time limit.Experimental results show that the processing time association allocation scheme can improve the efficiency of computing resources,and the proposed algorithm can complete most tasks accurately in the given time.In order to solve the problem of low efficiency of unloading strategy in industrial multi-terminal task unloading scenario,a multi-terminal computing unloading strategy based on Deep Deterministic Policy Gradient(DDPG)algorithm is proposed.This thesis analyzes and constructs a model of terminal and edge server in complex Internet of Things(Io T)industrial scenario,and relieves the urgent relationship between industrial equipment resources and computationally intensive scenario by offloading computing services to edge server.In this thesis,the effect of multi-terminal continuous unloading on the total task processing cost of the system is studied.Experimental results show that DDPG algorithm can achieve the lowest system task processing cost in multiterminal scenario and is feasible in many industrial scenarios.