Research On Autonomous Cruise Strategy Of UAV For Large-Scale Sensor Network Communications

The Fifth-Generation(5G)mobile communication brought larger communication scale and denser node deployment.Comparing with the civil area which has a complete communication system,building communication infrastructures at remote areas will face many challenges,such as inconvenient transportation and high cost,while these areas usually need a large-scale monitoring.Under this situation,using low-cost unmanned aerial vehicle(UAV)to provide communication services for remote sensor networks in cruise is an acceptable scheme.When beyond-5G(B5G)era comes,air-space-ground integrated information network will be one of the keys to achieve full coverage of communication scenarios,and the UAV will be an indispensable role in this network.The process of the aforementioned scheme can be summarized as follows.the UAV takes off from the charging station,and then cruises based on the planning trajectory to complete the communication task,and back to the charging station in the end to prepare for the next task.This process contains three parts,i.e.,planning trajectory,cruising based on the trajectory,and communicating with the sensor networks.Considering the consumer-grade UAV has limited endurance,guaranteeing the efficiency of completing task is the key problem.To improve the task-completion efficiency of the three parts,this dissertation deeply studies the UAV autonomous cruise stategy for large-scale sensor network communication from the perspectives of theory,simulation,and physical evaluation.The contributions are summarized as follows.1)An opportunistic cooperative relaying scheme performed in two-hop network.Considering the UAV has limited endurance,improving the efficiency of data aggregation can shorten the time required of completing task.In multi-hop sensor networks,a two-hop opportunistic cooperative relaying scheme is elaborated here.The scheme fully utilizes the communication resources of the idle nodes surrounding the default relaying node.Under a certain condition,the idle nodes will help the relaying node to forward the relaying packet sent by source node.The theoretical performance with different SNR,buffer capacity,and the number of node is verified in simulations which show that the scheme improves the efficiency of data aggregation.2)A parallel reinforcement learning-based cruising trajectory planning method.Considering the UAV has limited endurance,optimizing the cruising trajectory of UAV can improve the efficiency of the data aggregation in the large-scale sensor networks.A greedy-model-based parallel reinforcement learning method is proposed to plan cruising trajectory for UAV to deduce the time cost of aggregating the sensing data.Two greedy methods are proposed as baselines,and then their greedy ideas are used to design a reward scheme.Meanwhile,the method proposes a parallel structure with different parameters to explore environment to acquire low-correlation samples which accelerate the trainning.In simulations,four configurations with determined nodes or random nodes and cooperative mode or non-cooperative mode are set.Compared with two greedy methods,the proposed method has lower time cost and smoother generated trajectory,which mean it is power-saving and easier to implement.3)ω-free accelerometer pair method for trajectory reconstruction.Considering the UAV is controlled by navigation system,improving navigation method can guarantee the cruising.In a scenario of inertial trajectory reconstruction,an ω-free accelerometer pair method is proposed.The method uses a gyro-free low-energy hardware structure and abandons the calculations of angular velocity(ω)to alleviate the error accumulation.The proposed method introduces the minimum mean square error criterion to estimate the trajectory points,and its closed-form solution is derived.In simulations and physical experiments,compared with the other two typical inertial methods,the proposed method has lower power consumption,lower accumulated error,and relatively lower time complexity while it still guarantees the accuracy of trajectory reconstruction.That means the proposed method is more applicable in some low-energy inertial applications which support deploying sensor array.4)Bias error accumulation-oriented theoretical evaluation framework for inertial method.This framework puts the reading of static IMU into an inertial method,and treats its outputs as its bias-error-accumulation effect.Three typical inertial methods are evaluated by the framework and the theoretical results are verified in simulations,which show the framework is applicable to evaluate inertial method while the evaluation results are basises for designing navigation system.