Research On Spatiotemporal Crowdsourcing Task Allocation Methods

In recent years,spatiotemporal crowdsourcing has developed rapidly,providing people with diversified social services.In particular,selling products online and distributing them to communities has become a rapidly popular consumption mode.However,the widely used modes of community crowdsourcing,such as point of consignment collection and intelligent express delivery box,have some problems in distribution: 1)In first-tier cities with high store rental pressure and high consumption level,limited by high equipment maintenance costs and store rental costs,the capacity of collection points and the number of delivery boxes are limited,and the number of items exceeds the capacity of collection points in peak consumption or daily situations;2)Although the cold box or incubator reformed by cold chain technology is set in the collection site,the freshness and security of the articles cannot be well guaranteed,especially the articles with strong timeliness.In order to solve the above problems,this paper mainly carried out the following research work:(1)The task allocation problem of community service-oriented spatiotemporal crowdsourcing(CSSC)is defined,and the problem is formalized by the role collaboration model E-CARGO.In the problem,the platform completes task allocation and determines the delivery time of workers according to the user’s order,and arranges workers to put items into the collection point in real time before the earliest time of receiving goods,so as to solve problems such as limited capacity of collection point and reduced freshness of items.The goal of this problem is to realize the delivery of high-value order sets by highly qualified workers,while taking into account the conflicting situations.(2)Effective data processing.On the one hand,the role perception method based on kernel density clustering is used to divide tasks into different roles to realize the conversion from oneto-many assignment to one-to-one assignment and simplify the assignment process.On the other hand,as the current crowdsourcing platform has increasingly higher requirements on workers’ road familiarity and strain capacity,a multi-stage quantitative model of agent location qualification value is designed based on learning forgetting curve to realize online learning and adaptive updating of agent location qualification value.(3)Two baseline allocation algorithms are proposed: greedy allocation based algorithm(PQGR),which preferentially matches agents with high qualification value for roles with high value(order set);Based on the conflict-considering group multi-role assignment algorithm(PQGM),the task assignment problem is transformed into a linear programming problem,which is completed by CPLEX toolkit.In order to optimize the PQGM algorithm,two improved algorithms are proposed: Task allocation method based on transition interval(PQGMA),which can shorten the running time by reducing the running times of linear programming algorithm;Classified-based task allocation method(PQGMC)uses classified-based role perception method to optimize the effect of task set partitioning and task allocation.In this paper,the effectiveness and efficiency of the algorithm are verified by experiments on g Mission dataset and synthetic dataset.