With the booming mobile Internet and sharing economy, online spatial crowdsourcing services, such as Uber and Didi, are becoming important infrastructures of our daily life. Existing studies about spatial crowdsourcing usually focus on the platform interests and ignore experiences of individual requesters and workers. Because of the dynamically changed task demand and worker supply, the major experience issue, for both workers and requesters, is caused by insufficient valid assignments. In this thesis we discuss the problem of how to allocate limited assignment resources fairly to both worker and requester sides.

For a requester, because his/her task needs to be assigned as soon as possible, the resource is nearby available workers. Thus, we study the minimizing maximum delay...[ Read more ]

With the booming mobile Internet and sharing economy, online spatial crowdsourcing services, such as Uber and Didi, are becoming important infrastructures of our daily life. Existing studies about spatial crowdsourcing usually focus on the platform interests and ignore experiences of individual requesters and workers. Because of the dynamically changed task demand and worker supply, the major experience issue, for both workers and requesters, is caused by insufficient valid assignments. In this thesis we discuss the problem of how to allocate limited assignment resources fairly to both worker and requester sides.

For a requester, because his/her task needs to be assigned as soon as possible, the resource is nearby available workers. Thus, we study the minimizing maximum delay spatial crowdsourcing (MMD-SC) problem and propose solutions aiming at achieving a worst case controlled task assignment. The MMD-SC problem assumes that both workers and requesters come dynamically and considers not only the workers’ traveling time costs but also the buffering time of tasks, thus it is very challenging due to two-sided online setting. To address these challenges, we propose a space embedding based online random algorithm and two efficient heuristic algorithms.

For the worker side, the resource is new tasks which need to be distributed to workers as equally as possible. The first challenge is to formally define the worker fairness and combine it with existing platform level goals, and the second challenge is to conduct task assignment with consideration of worker fairness and platform interests. To address these challenges, we formally define an online bi-objective matching problem, namely the worker-fairness-aware assignment problem (WFAA), and some special cases/variants of it to fit in most spatial crowdsourcing scenarios. We give corresponding solutions for different cases of WFAA. Particularly, we show that the dynamic sequential case, which is a generalization of an existing fairness scheduling problem, can be solved with an O(n) fairness cost bound (n is the total worker number), and give an O(n/m) fairness cost bound for the m-sized general batch case (m is the minimum batch size).

In addition, we show the effectiveness and efficiency of our methods via extensive experiments on both synthetic and real datasets.