Research On Task Allocation And Data Collection Strategy With User Characteristics In Mobile Crowdsensing

Mobile Crowdsensing(MCS)collects data relying on a large number of users and their mobile smart devices,and complete some relatively complex sensing tasks.Providing reliable sensing results is the primary goal of MCS,and effective task allocation and data collection strategies are important guarantees for this goal.In the current MCS related research,there are still some challenges in the following two aspects.In the task allocation of MCS,due to the dynamic change of users' location and the spatiotemporal sensitivity of tasks,it is difficult to match the task and the user effectively.What is more,the sparsity of the users' historical visit data may influence the task acceptance rate in the task allocation.In the data collection of MCS,different users have different reliability,so the quality of the sensing data may not be guaranteed.In addition,it is also difficult to infer the truth of the data from the data submitted by many users.Therefore,in order to solve the above problems,it is very important to design an effective task allocation and data collection strategy.This thesis introduces the research background of MCS,summarizes the related research methods in task allocation and data collection in MCS,and proposes the solutions to the existing challenges and shortcomings.The main research results of this thesis are as follows.A spatiotemporal characteristic aware task allocation strategy towards sparse user data is proposed.In the strategy,the tasks are first divided into scheduled tasks and unscheduled tasks according to the task release time and execution time.For the user data sparseness in scheduled tasks,based on the user's direct spatiotemporal preference,a matrix factorization method is designed to evaluate the user's potential spatiotemporal preferences.For the problem of user data sparseness in unscheduled tasks,a similar-user-enhanced semi-Markov model is used to predict probability of users' task completion.Then,the scheduled tasks are allocated according to the user's spatiotemporal preferences,and the unscheduled tasks are allocated based on the user's spatiotemporal preferences and the task completion probability.The simulation results show that this strategy can effectively allocate the above two types of tasks under the condition of sparse user historical visiting data,and improve the task acceptance rate.A users' reliability-driven data collection strategy with truth inference is proposed.In this strategy,firstly,a dynamic model of user reliability is constructed by evaluating the task proficiency,task complexity,and task contribution of each task performed in the user's historical task record,so the changes of the user's long-term reliability can be quantified,and the task participant's current reliability can be obtained.Next,a user reliability-driven truth inference model is designed.Using an unsupervised method to analyze the data submitted by the task users set,the truth of the task data can be inferred and the user's task contribution can be quantified.Based on the contribution,the user's reward distributed for the task can be confirmed.The simulation results show that this strategy can effectively infer the true of the data,complete the data screening work,and encourage users to improve data quality.