Dynamic allocation for complex mobile crowdsourcing task with internal dependencies

Recently, mobile crowdsourcing utilizes the mobility of crowds and the computing capability of smart computing devices, which has received considerable attention from the research community. Recent pieces of literature focus on the multi-task allocation for mobile crowdsourcing, considering improving the accuracy of the collected data by inferring and utilizing the reliability of users. However, they neglect the fact that the internal dependency relationship exists on sub-tasks, which is decomposed via a divide-and-conquer method from the source tasks. Neglecting this fact may cause a series of dilemmas including high latency, low quality, low execution rate, and even more allocation errors. To address these problems, we should renormalize the task allocation problem of mobile crowdsourcing, and the task is divided into logically dependent sub-tasks, and then sub-tasks are sorted by the Depth-first search (DFS) algorithm. Different from previous approaches, an allocation method SGTA considering task internal dependencies is introduced to ensure that the workload does not exceed the executive ability of each worker and try to avoid misallocation. Based on the submission report during task execution, we devise a SimilAr Task Dynamic (SATD) allocation method to dynamically re-allocate these failure tasks. Finally, experimental results based on both synthetic and real data sets demonstrate that the proposed method is effective.