Trajectory Design in UAV-Aided Mobile Crowdsensing: A Deep Reinforcement Learning Approach | Synapse