Research On Computation Task Offloading Methods Of Mobile Edge Computing For Industrial Internet Of Things

In recent years,the industrial internet with low latency,high reliability and wide coverage has been the guarantee for industrial manufacturing,and the catalyst of its unprecedented progress.However,as the number of industrial terminal devices grows and becomes more intelligent,terminal information explodes,putting enormous pressure on the central communication link.Edge computing technology in the Industrial Internet is an effective tool of alleviating this situation.In mobile edge computing for the Industrial Internet,offloading terminal computing tasks to edge servers in a reasonable way could accelerate terminal tasks while relieving the pressure on the central communication link,therefore meeting the demand for computing tasks in the large-scale integrated environment of the Industrial Internet and paving the way for the leap from industrial manufacturing to smart manufacturing.However,it is an imminent to find a way to reduce task offloading overhead(latency and energy consumption)in the edge network environment to achieve reasonable task offloading.To address this problem,this paper works on different edge task offloading scenarios for the Industrial Internet,so as to minimize task offloading overheads in different offloading environments.The main work of the paper is as follows.First,for the associated task offloading scenario,a task chain-based multi-centre associated task offloading scheme is proposed and implemented considering factors such as the associated task network link transmission overhead and task instantiation overhead offloading in multiple edge servers in the industrial Internet environment.The scheme models the offloading of sequential tasks in a multi-centre MEC system and transforms the original computational task offloading problem into a 0-1 integer linear programming problem with multiple constraints.In the multivariate coupled case,the minimisation overhead problem is transformed into two sub-problems by Lagrangian relaxation and solved using a heuristic algorithm and an SCA algorithm,respectively.The solution effectively reduces the solution size under minimisation overhead and obtains a suboptimal solution in polynomial time,meeting the needs of practical applications of the Industrial Internet.Second,the distributed task offloading problem with unknown global information of the system is studied for the unassociated task offloading scenario.By modelling the task offloading in the case of multiple intelligent terminals with unknown server information and unknown information between intelligent terminals,and transforming the task offloading overhead minimisation problem into a centralised optimal offloading and distributed optimal offloading benefit regret minimisation problem.A distributed unassociated task offloading scheme based on a multi-armed bandit is also proposed and implemented using the explore-utilise strategy in the multi-armed bandit algorithm to solve the minimum regret problem.A regret evaluation function is also introduced to evaluate the solution results intuitively.Finally,through the simulation of experimental scenarios and relevant comparison experiments,it is shows that the multi-centre associative task offloading method and the distributed unassociative task offloading method proposed in this paper have lower offloading overhead,thus ensuring better offloading benefits for task offloading.These results validate the performance and advantages of the offloading method in this paper.