Research On Task Offloading Strategy Based On Edge Computing In Industrial Internet Of Things

Rapidly advancing in recent times,the industrial Internet of Things has seen its equipment widely employed in industry.In the industrial internet of things environment,due to the rapid growth of computing power demand for various tasks,Internet of Things devices have encountered a bottleneck in computing power,and edge computing has become an important means to solve this problem.Although edge computing has important application value in the field of industrial internet of things,it still has some limitations,especially in data transmission,storage and resource allocation.Thus,a sensible computing offloading approach is essential to capitalize on the benefits of these two technologies.Aiming at the computing offloading problem in the Industrial Internet of Things,we propose a computing task offloading algorithm considering system latency and energy consumption.To begin,a model of an edge computing system is created to facilitate the dynamic distribution of server resources.Further,the computing tasks offloaded from the device end to the edge server are modeled as queues.Modeling the edge to end offloading problem based on the established model,we break it down into two sub-issues: resource allocation and task offloading.To solve the task offloading problem,we suggest a deep meta learning algorithm that combines meta learning and deep learning algorithms.Simulations revealed that the meta learning-based offloading algorithm was more effective in enhancing system utility than the deep Q-learning algorithm.Then,for scenarios where some devices in the Industrial Internet of Things have mobility,we propose a service migration algorithm that considers latency and energy consumption.Firstly,in view of the characteristics of some devices in the industrial Internet of Things that continuously change location,we establishe a service migration system model and model the migration decisions.On the basis of the decision model,the service migration problem is constructed into a long-term utility maximization problem.Then,we describe the service migration problem as a Markov process,which optimizes the migration cost in terms of latency and energy consumption.We present a hierarchical reinforcement learning-based service migration algorithm to address the service migration issue.Simulation results demonstrate that this algorithm can ensure the success rate of service migration while meeting the constraints of computation and storage resources.Compared with other algorithms,the algorithm proposed in this thesis has superior performance in optimizing system utility.