| In recent years,multi-UAV(Unmanned aerial vehicle)fleet has gradually become a hot topic in the UAV research area due to its advantages of high mission efficiency and high fault-tolerance rate.The UAV formation involves many fields,and the localization is the first consideration for formation cooperative flight.The complex and alternate mission environment requires high reliability and accuracy of localization of the UAV formation.In addition to the global localization information,the relative localization accuracy of UAVs within the formation is a key factor to the success of the mission.At the same time,except for the problem of localization,formation assembly,formation maintenance and obstacle avoidance need to be considered.The selection of appropriate formation control method plays an important role in the completion of the task.And adopting effective obstacle avoidance strategy can provide sufficient guarantee for the formation flight safety.In order to improve the flight control safety of UAV formations,this thesis has performed research on the visual relative positioning algorithm based on the dynamic properties of UAVs and formation control for multiple UAVs with dynamic obstacle avoidance based on the flocking behavior.The main contributions of this thesis can be listed as follows:(1)A novel multi-UAV intelligent detection and relative localization algorithm based on monocular vision is designed,which is aiming at the localization problem of the nano UAV formation in the GPS-denied environment.Firstly,the UAV dataset is au-tomatically annotated and constructed based on the background subtraction.Secondly,a convolutional neural network based on the Darknet is designed to enable a UAV to detect and recognize UAVs in the view online,and the detection accuracy is improved by a suitable selection of the loss funciton.Thirdly,the detection results based on the deep learning are transformed into the relative position of the follower relative to the leader.And the control algorithm is designed for the tracking control of the UAV.Fi-nally,the visual relative localization algorithm is successfully applied to the formation system composed of nano quadrotors with a mass less than 100 grams,and an indoor flight experiment of UAV formation based on the leader-follower method validates the effectiveness of the algorithm.(2)This thesis presents a novel distributed formation control strategy based on the flock-ing behavior for formation holding and dynamic obstacle avoiding control of UAVs.Firstly,considering the disturbance caused by the airflow between the UAVs,a dis-tributed multiple UAVs formation holding controller is developed based on the attrac-tion/repulsion potential field and consensus method,which ensures the formation main-tainance between the UAVs.Secondly,considering the effects of external moving ob-stacles,the repulsive potential field is introduced to generate obstacle avoidance strategy for UAVs,and control the UAVs to avoid moving obstacles.Thirdly,by using the Lya-punov based stability analysis,the ultimate boundedness of system state errors and the stability of closed-loop system are proved.Finally,the indoor realtime experimental verification are performed on the multiple quadrotor UAVs testbed.Experimental re-sults show that the proposed distributed flocking formation obstacle avoidance control strategy can effectively avoid external moving obstacles and is able to reconstruct the UAV formation after obstacle avoidance. |
PDF Full Text Request