Research On Data Credibility Driven MCS Multi-task Assignment Strategy

In Mobile Crowd Sensing(MCS)system,task assignment is one of the indispensable processes for the system to provide efficient and reliable sensing services.However,in a large number of concurrent task scenarios,it is difficult to accurately match the dynamic state of participants and the diversification of task characteristics;at the same time,task requesters have high requirements for data credibility.In addition,in the task assignment process,the number of users is huge,and there is a lack of prior knowledge to perform specific tasks,which will reduce the participants' search efficiency and matching accuracy,as a result,the data credibility is impacted.Therefore,it is very meaningful to design an efficient task allocation strategy to improve the task completion rate,data credibility and platform utility.This paper introduces the research background,significance,and current situation of MCS network.And we analyze and summary the existing challenges in task assignment,in particular,focusing on the data quality issues in task assignment process,and detailed the current research situation of task assignment.A Service Benefit Aware Multi-task Assignment Strategy(SBMAS)for MCS is proposed.Service benefits of participants are modeled based on their task difficulty,task history,sensing capacity,and sensing positivity to meet differentiated requirements of various task types.An improved fuzzy clustering method is used to cluster users according to their task preferences similarity to narrow the search range of participants.Through an iterative process based on the gradient descent algorithm,the participants with the highest benefit are quickly matched for each type of task.The results show that the proposed strategy not only effectively reduces the number of iterations of the algorithm but also improves the task completion rate.An Edge-assisted Data Quality-Aware Task Assignment Mechanism(E-DQATAM)is proposed.According to the characteristics of MCS data,an evaluation model of data validity and spatial and temporal correlation is constructed in the cloud to determine the set of pre-selected participants.It obtains the instantaneous status information of preselected participants at the edge server to evaluate their service capabilities.According to the results of the spatiotemporal correlation quality indicators of the sensing data,it is determined whether to recruit new users to participate in the sensing task in the cloud,so that it supplies the participant resources to the edge server and use the Thompson sampling algorithm of Multi-Armed Bandit(MAB)to select participants with strong service capabilities,which can improve the data qualification rate.The experiment results show that the proposed mechanism improves the sensing data quality and the platform utility.