Efficient Data Sensing and Collection for Multi-UAV Assisted Mobile Crowdsensing with Energy Harvesting Devices

Mobile crowdsensing (MCS) has been recognized as a potential paradigm, due to its mobility and expandability. However, existing works have ignored or simplified the modeling of the users' energy supply in MCS systems. In this paper, we formulate the user energy harvesting model in MCS systems, while Unmanned Aerial Vehicles (UAVs) can effectively serve the MCS system. We aim at minimizing the service latency and maximizing the data utility by jointly optimizing variables such as bandwidth allocation, UAV trajectories, and other related factors. To this end, we transform the energy consumption problem into a Lyapunov drift minimization problem, and utilize a Transformer-enhanced Deep Reinforcement Learning approach to solve this problem. Experiments demonstrate the convergence of our proposed algorithm, and the proposed method can significantly outperform several benchmark approaches.