Vision-based Localization And Nonlinear Fault Tolerant Control Of Small Multi-rotor Unmanned Aerial Vehicle

The application scenarios of the multi-rotor UAV(Unmanned Aerial Vehicle)have gradually expanded from the military area to the civilian area in the past few years,and it is widely employed in search and rescue,remote sensing,aerial photography,power maintenance,and agricultural irrigation.With the miniaturization of chips and sensors,the rapid development of artificial intelligence,and to meet the diversification and complexity of application scenarios,the multi-rotor UAV is gradually becoming smaller and more intelligent.Researchers began to explore non-traditional methods of autonomous sensing and control in the hope that the multi-rotor UAV could autonomously perform sensing,decision making,path planning,and control in complex,unstructured environments.For the multi-rotor UAV,two important directions to improve its intelligence are autonomous sensing and behavior control.Firstly,this thesis analyzes the feature point method and direct method in the V-SLAM(Visual Simultaneous Localization and Mapping)algorithm,and propose a sparse semi-direct method based on the combination of the two methods.Position and pose estimation can be performed without extracting feature points and calculating descriptors for each frame of the image,which saves calculation time and possesses good real-time performance.At the same time,the BA(Bundle Adjustment)optimization is used based on the sliding window,which can improve the accuracy of the localization and ensures the real-time performance of the SLAM algorithm.This thesis builds a small quadrotor UAV testbed to verify the accuracy and robustness of the proposed visual SLAM algorithm.Indoor experiments are implemented and the contrast experimental results with the position information measured by the motion capture system show that the proposed localization algorithm is accurate.During the experiment,the average algorithm frame updating rate can reach about 290 Hz,which indicates the good real-time performance of the proposed algorithm.Finally,outdoor experiments show that the proposed SLAM algorithm possesses good applicability and robustness for different textures and illumination variation.This paper focuses on the FTC(Fault Tolerant Control)design for a tilt tri-rotor UAV with the rear servo’s stuck fault,which is affected by unknown disturbance and model uncertainties.After analyzing the nonlinear dynamic model of the tilt tri-rotor UAV,the rear servo’s stuck fault is added to the torque solution equation.The adaptive backstepping methodology is combined with the non-singular terminal sliding mode control to develop the nonlinear robust FTC without the need of a fault diagnosis mechanism.The Lyapunov-based analysis method is employed to prove the stability of the closed-loop system and the asymptotic convergence of the attitude error.Finally,numerical simulation and real-time experiments on a self-build hardware-in-loopsimulation tri-rotor testbed are implemented,and the results show that the proposed algorithm has achieved good control performance for the tilt tri-rotor UAV against the rear servo’s stuck fault.