UAV Network Deployment And Topology Control Method Based On Relative Positioning

In scenarios such as emergency rescue,battlefield environment,and polar exploration,it is usually necessary to build temporary communication network.Mobile ad-hoc network(MANET)provides a feasible solution to this problem.However,in some special scenarios,the complex terrain and various unknown obstacles will seriously interfere with the communication between the relay nodes on the ground.Therefore,it is optional to use unmanned aerial vehicle(UAV)to build the communication network.As the relay nodes,UAVs have more advantages in terms of speed and moving space compared with the nodes on the ground,and can quickly reach the designated locations due to the high mobility and flexibility.The communication between the UAV and the nodes within its coverage can be hardly interfered by terrain and obstacles,which guarantees the higher communication quality and efficiency.In most cases,the controller can determine the position of UAVs in the network before the deployment with the absolute position information provided by the satellite positioning system.However,it will be a great challenge for the deployment of the UAV relay network when the satellite positioning system is unavailable or the accuracy cannot be guaranteed.Thus,this thesis proposes a deployment and topology control scheme based on the relative position information.The main work of this thesis can be summarized as follow:(1)A deployment and topology control scheme of the UAV network,ETTGA,is proposed,which is based on the relative position information.As a dynamic scheme,ETTGA contains a series of algorithms and strategies such as the deployment,recovery,topology control and flight control,which completes the methodology system in this field.Compared with the two static schemes,greedy-based scheme(Greedy)and genetic algorithm-based scheme(GA),which are all based on the absolute location information,the network based on ETTGA achieves 90% and 70% of the coverage performance of them respectively and is significantly better than those two static schemes in robustness and load balancing.The simulation experiments on the NS-3 show that the network based on ETTGA has the higher throughput and lower packet drop rate than GA and Greedy under the heavy load.(2)The deep reinforcement learning framework is introduced in the process of the flight control of UAVs.The flight control decision of the UAVs can be obtained in real time by taking the relative position information acquired periodically as the input of the reinforcement learning model,so as to realize the autonomous flight of the UAVs.The simulation results show that the UAVs can fly from the starting position to the destination with 100% success rate under the guidance of the reinforcement learning model.(3)The problem of improving the measurement accuracy of the relative position information in the actual environment is researched.At first,in the scenario of this thesis,the theoretical control model is constructed for the flight process of the UAV,and then the extended kalman filter(EKF)algorithm is used to fuse the data predicted from the control model and the data measured from the sensors.The experiments show that the EKF algorithm can improve the accuracy of distance measured effectively,and makes a contribution to migrate theoretical methods to the actual environment.