Research On Resource Allocation Strategy In Edge Computing Of Industrial Internet Of Thing

The Industrial Internet of Things(IIo T)encompasses ground and emergency scenarios,utilizing advanced sensors,real-time data processing,and machine interconnectivity to achieve functions such as real-time monitoring,remote control,and intelligent optimization.IIo T is driving transformation in the manufacturing and automation industries.However,due to limited device resources and the constraints of traditional cloud computing architectures,handling massive IIo T data remains a challenge.To address this issue,this paper introduces Mobile Edge Computing(MEC)and proposes a resource scheduling strategy based on deep reinforcement learning to optimize computation and storage resource allocation,effectively handling multi-user computing tasks under large-scale data.Experimental results demonstrate that the proposed strategy can efficiently improve system efficiency,achieve cost reduction in resources,and optimize performance.The specific work is as follows:(1)A dynamic resource allocation strategy based on double deep Q-network is designed to address data privacy,long battery life requirements of terminals,and limited terminal computing resources in IIo T scenarios.This strategy aims to reduce system latency and energy consumption by optimizing the offloading decisions of multiple terminals and the allocation of server computing resources.Experimental results validate the advantages of this strategy under random dynamic conditions,effectively jointly optimizing the latency and energy consumption of computing tasks,improving task execution efficiency,and enhancing the energy utilization efficiency of terminal devices.(2)A dynamic server cache update strategy is designed to address evolving user demands in real-world scenarios.This strategy utilizes the storage resources of MEC servers to dynamically update Virtual Network Functions(VNF)in the cache,reducing the data volume of terminal offloading computation transmission,thus reducing system latency and energy consumption.Additionally,to tackle the challenges posed by the diversity of computing tasks,an improved deep neural network is studied to solve the problem of slow convergence caused by the large action space of neural network outputs.Experimental results validate that the improved deep neural network achieves faster convergence,and the proposed dynamic server cache update strategy accurately caches the required VNFs for requested tasks,reducing the data volume of terminal offloading and optimizing system latency and energy utilization.(3)To address the impact of the dynamic failure probability of MEC servers on IIo T task computation latency,a resource scheduling strategy based on dueling deep Q-network is proposed.A real-time updating MEC module is designed to accurately read the real-time status of servers and calculate their dynamic failure probabilities.Experimental results demonstrate that the algorithm in this paper achieves sufficient utilization of resources within the system and effective allocation of task offloading locations through the analysis of computing tasks and server statuses.This further enhances the reliability of MEC servers and reduces task processing latency.