| In recent years,ubiquitous drones are increasingly becoming an important sensing platform for the smart city.Due to the advantages of mobility,autonomy,rich sensors,excellent communication/processing capabilities and broad "bird’s eye view",drones can offer more flexible and powerful urban sensing capabilities than ground sensor networks.Therefore,it has important academic significance and application value to investigate how to leverage a drone swarm to cooperatively conduct urban sensing tasks.Energy and cost are two important issues to consider when leveraging drones for sustainable urban-scale sensing.With the advances of wireless charging technologies and the inspiration from the sparse sensing paradigm,this thesis proposes a novel drone-based collaborative sensing framework DroneSense.It selects a minimum number of Points of Interest(POI)to schedule drones for physical data sensing and then infers the sensing data of the remaining POI to meet the overall sensing quality requirement.However,drone-based sensing is very different from human-centric crowdsensing,resulting in a series of new problems,including which POI are visited first,when and where to charge drones,which drones to charge first,how much to charge,and when to stop the scheduling.To this end,this thesis designs a holistic solution,including:1)context-aware matrix factorization for data inference,which introduces urban context data to make the inference results more accurate;2)progressive determination of task quantity,which first uses a deep learning model to preliminarily predict the number of tasks,and then uses a multi-stage determination scheme to gradually determine whether to stop task allocation,thus effectively reducing unnecessary quality assessments and avoiding redundant sensing tasks;3)deep reinforcement learning(DRL)based task selection,which selects fewer POI for sensing by optimizing long-term utility;4)energy-aware DRL-based task scheduling,which comprehensively models complex factors and dependencies under strict routing and energy constraints;5)adaptive charger scheduling,which determines which drones to charge first and how much to charge.Finally,this thesis focuses on a typical use case,namely leveraging DroneSense to sense the occupancy status of urban open parking spaces and conducts extensive experiments with a real-world on-street parking dataset from Shenzhen,which proves the obvious advantages of DroneSense and demonstrats its feasibility for sustainable urban-scale sensing. |
