Spatiotemporal characteristic aware task allocation strategy using sparse user data in mobile crowdsensing

Mobile crowdsensing (MCS) requires users to move to a specific area to complete the sensing task within a specified period. In order to match the tasks and users when users' historical visit data is relatively sparse, and to allocate the tasks effectively, we propose a spatiotemporal characteristic aware task allocation strategy using sparse user data. Firstly, tasks are classified into scheduled tasks and unscheduled tasks. Next, for the sparsity problem in the allocation of scheduled tasks, a matrix decomposition method is designed to evaluate the potential spatiotemporal preference of the users based on their direct spatiotemporal preference; for the sparsity problem in the allocation of unscheduled tasks, we use a similar user enhanced semi-Markov model to predict the task completion probability. Then, the scheduled tasks are allocated based on the users' spatiotemporal preference, and the unscheduled tasks are allocated based on the spatiotemporal preference and the task completion probability of the users. Finally, we conduct simulations to verify the proposed strategy based on a real world dataset (geolife dataset). Experimental result shows that the proposed strategy can effectively allocate the tasks under the condition of sparse users' historical visit data and improve the acceptance rate of tasks.

Automated summary

This paper proposes a spatiotemporal characteristic aware task allocation strategy using sparse user data for mobile crowdsensing (MCS). It effectively matches and allocates tasks even when users' historical visit data is relatively sparse, improving the acceptance rate of tasks.