Research On Task Assignment In Spatiotemporal Crowdsourcing

With the continuous development of crowdsourcing applications,the content of research becomes more diversified,and the constraints of time and space begin to appear in the crowdsourcing tasks and Traditional crowdsourcing transformed into spatio-temporal crowdsourcing.In recent years,most of the research on spatio-temporal crowdsourcing has focused on three core issues: task allocation,quality control and privacy protection.The task assignment problem in traditional crowdsourcing research is often described by a bipartite graph matching model.Most of the task assignment are carried out under the premise of knowing the specific information of the crowdsourced task and the crowdsourced worker In practical applications,tasks in a crowdsourcing environment have certain real-time constraints.In the context of crowdsourcing workers completing multiple tasks,it is necessary to consider path planning issues.With the development of space-time crowdsourcing,the research on task allocation based on two types of objects has gradually failed to meet the actual needs.More scholars have begun to focus on the task allocation problem for three types of object and more research on dynamic online scenes instead of static offline.The online task assignment problem of three types of objects refers to that crowdsourced workers need to go to the task location to pick up the task,and then go to the task execution location to complete tasks.We study online task assignment problem of three types of objects that minimizes the average task cost and maximizes the number of matches under this scenario.1.A tabu search algorithm based on adaptive threshold is proposed for the problem of minimizing the average task-consuming time and maximizing the number of matches for three types of objects.In this part,spatiotemporal crowdsourcing and path planning are combined to reasonably plan the sequence of task the crowdsourcing workers carry out.Through online learning,the reasonable estimated waiting time of each task was calculated to match crowdsourcing tasks and complete tasks in the shortest time.The comparison with similar algorithms proves the efficiency of this method.2.For the continuous task allocation problem of three types of object in dynamic scenario,it is proposed to realize the task allocation model in dynamic scenarios by predicting the objects that may appear in the next time window.We predict the number of tasks that may appear in the current area through a decision tree in the fist place,Then use the agglomerative hierarchical clustering algorithm to obtain the location of each task according to the number of tasks.Finally predict the location of the task execution location to allocate the task and plan the route for three types of objects by Tabu-Search algorithm.