Mobile crowd sensing (MCS) has been widely used as a cost-efcient way to collect data for smart cities, which typically starts with participant recruitment and task allocation. Previous work mainly focused on selecting a proper subset of humans for contributing sensing data. However, there often exist situations where humans are not able to reach the target areas, such as traffic jams or accidents. One solution is to complement manual data collection with autonomous data collection using unmanned aerial vehicles (UAVs) equipped with various sensors. In this paper, we focus on the scenarios of UAV-assisted MCS and propose a highly efficient task allocation method, called UMA (UAV-assisted Multi-task Allocation method) to jointly optimize the sensing coverage and data quality. The method incentivizes and guides human participants to contribute high-quality sensing data. Meanwhile, the UAVs are employed to sense data from rarely sensed points of interest, and calibrate data contributed by human participants. The method leverages emerging deep reinforcement learning techniques for directing UAVs sensing and movement actions based on the human participants locations and tasks achievement. The results well justify the efficiency of UMA in terms of coverage completed ratio, calibrating ratio, task fairness and energy efficiency, compared with the state-of-the-art.