A Coverage-Aware Task Allocation Method for UAV-Assisted Mobile Crowd Sensing

Abstract

<p>Mobile Crowd Sensing (MCS) is an emerging paradigm that engages participants collaboratively in completing sensing tasks. The mobility and intelligence of mobile devices offer an efficient solution for large-scale sensing applications, such as in smart cities. Unmanned aerial vehicles (UAVs), considered as mobile devices, can be integrated into MCS to collaborate with human participants in order to meet the task sensing coverage requirement. In this paper, we investigate a UAV-assisted task allocation method (U-TAM) that allocates tasks to human participants and UAVs concurrently. Distinct from existing methods, U-TAM prioritizes minimizing the privacy leakage of human participants while maximizing sensed coverage. To achieve this, it initially predicts their trajectories using a deep reinforcement learning approach, relying solely on the information provided by their start and destination locations. In addition to the predicted trajectories, the proposed U-TAM allocates tasks to human participants based on their tolerance levels and limited budget. This approach aligns with the Pareto optimal theory, seeking to balance the trade-off among participants' tolerance level, limited budget, and the requirement for task sensing coverage. In the meantime, the UAVs sense data efficiently from areas that are not sensed by human participants or other UAVs. To this end, we propose a multi-agent deep reinforcement learning framework for multi-UAV trajectory planning, which integrates the greedy method into deep Q-learning. We evaluate the proposed method using simulation and a small-scale practical experiment. Extensive experiments are used to verify the method's efficiency.</p>