Research On Spatial Crowdsourcing Dynamic Assignment Algorithm Based On Fog Computing

With the booming development of the commercial Internet and the rapid upgrading of mobile communication technology,people can complete a series of crowdsourcing tasks anytime,anywhere,such as reporting current road traffic jam information and current weather information in the area.These tasks have obvious time and space attributes,resulting in the computational paradigm of spatial crowdsourcing.Spatial crowdsourcing not only facilitates people's daily life,but also promotes the development of space-time crowdsourcing platforms such as Meituan and Uber,which will play an increasingly important role in the construction of smart cities in the future.In spatial crowdsourcing,task assignment is not only the basis,but also the key.A reasonable task allocation is an important guarantee for the success of spatial crowdsourcing.Therefore,more and more experts and scholars have begun to study spatial crowdsourcing tasks assignment issues.However,most task data transmission under traditional research is directly from the terminal to the cloud,there will be a problem of excessive transmission delay;in addition,the traditional crowdsourced task assignment research is mainly carried out in a fixed time interval,there is no considering the dynamic emergence of workers and tasks,which leads to low efficiency of crowdsourcing task assignment,which also makes people's enthusiasm for participating in crowdsourcing tasks greatly reduced.Therefore,in order to effectively solve the above problems,this paper proposes a spatial crowdsourcing dynamic allocation algorithm based on fog computing.The research content of this article is as follows:(1)A novel cloud-fog-Io T three-layer fog computing task assignment architecture is proposed to take advantage of the advantages of neighboring user fog nodes and reduce the delay of task processing;(2)Using the deep learning GRU model,the number of tasks in the future can be predicted more accurately;(3)Use the deep Q network(DQN)and dual deep Q network(DDQN)in reinforcement learning to design an adaptive time batch processing algorithm to dynamically adjust the size of each time batch;(4)Apply the Kuhn-Munkres(KM)algorithm to match tasks and workers in each time batch,so as to maximize the platform's overall benefits.