Research On Robust Disturbance Rejection Control And Visual Servo System Of Aerial Manipulator

With the rapid development of artificial intelligence technology and the continuous improvement of advanced control theory,intelligent robots have played an important role in the fields of power inspection,traffic monitoring,military investigation,medical rescue,and so on.The new air operation robot represented by the rotor UAV with a manipulator provides favorable help for bridge flaw detection,photovoltaic cleaning,and dangerous goods handling because of its unique flexibility,high mobility,and interactive operation ability.At present,the research on aerial manipulators in China is still in its infancy.As a complex strong coupling and underactuated system,there are still many difficult problems to be solved in system design,modeling,control,and simulation.This paper focuses on the problem that the center of mass of the aerial manipulator system is offset by the motion of the manipulator,which leads to the reduction of the positioning accuracy and the instability of position and attitude.Firstly,the modeling method of the aerial manipulator is studied.The purpose is to accurately describe the dynamics model of the aerial manipulator with centroid deviation through an intuitive and visual modeling method,and to provide a theoretical basis for the design of the controller.Visual servo control is an important control means for operating robots to achieve high-precision operation.This paper studies the target pose estimation algorithm of the visual module and the robust disturbance rejection control method of the control module in visual servo control,focusing on the running speed of the visual algorithm,the accuracy of target detection and the response speed,disturbance rejection ability,and robustness of controller in visual servo control.For the control module,this paper studies the advanced control theory.Based on linear active disturbance rejection control,sliding mode control,and fuzzy adaptive control,two control methods to improve the performance of the aerial manipulator controller are proposed.For the vision module,this paper uses the characteristics of fast speed and high precision of neural network algorithm,and combined with the principle of camera three-dimensional reconstruction,designs a high-performance target detection and tracking algorithm to ensure the quality of the input signal of the controller.The main contributions of this paper include the following aspects:(1)An integrated modeling method of the aerial manipulator is proposed.In this method,the motion of the manipulator’s joint is mapped to the kinematics of the manipulator system centroid,and then the relationship between the manipulator joint motion and the aerial manipulator system centroid motion is obtained.Then,the dynamics of the aerial manipulator system centroid is decomposed and coupled to the attitude dynamics model of four-rotor UAV Based on the Newton-Euler method through centroid theorem and momentum theorem.(2)Two advanced control schemes of the aerial manipulator are proposed.Combined with the integrated dynamic model of the aerial manipulator and the mission characteristics of the aerial manipulator,to solve the problems of low control accuracy,weak anti-interference ability,and insufficient robustness of traditional control algorithms,a fuzzy linear active disturbance rejection control method for rotor UAV is proposed.Firstly,the key parameters of the linear active disturbance rejection control are deeply analyzed and tested,and the function and significance of each parameter are obtained.Then,the adaptive ability of fuzzy control is used to adaptively adjust the key parameters,to make the linear active disturbance rejection control have a faster response speed and stronger disturbance rejection ability.Experiments show that the response speed,disturbance rejection ability,and robustness of the fuzzy linear active disturbance rejection control method are improved.Then,to realize the robust control of the aerial manipulator when the model accuracy is insufficient and the parameter perturbation is large,the linear extended state observer is deeply studied,and the sliding mode control based on the linear extended state observer is designed.Aiming at the shake problem of sliding mode control law,a fuzzy adaptive controller is designed to adjust the synovial approach law factor to realize the smooth output of control quantity.Finally,a fuzzy sliding mode control method based on linear extended state observer is proposed.Experiments show that the anti-interference ability,response speed,and robustness of this method meet the performance requirements of the aerial manipulator.(3)YOSM target detection and tracking algorithm is proposed.Firstly,based on the neural network with the Darknet-53 structure,this paper increases the possibility of detecting small-size objects at a long distance by adjusting the convolution size and step size.At the same time,to improve the network reasoning speed,some affine operations are replaced by linear operations of convolution.Experiments show that the improved neural network target detection algorithm improves the running speed by three times on the premise of ensuring the accuracy of the original network.Finally,to meet the requirements of the visual servo system to obtain the target relative position information,the binocular stereo matching algorithm is used to calculate the depth information of the target in the neural network detection result box,and the position information of the target in three-dimensional space is estimated based on the internal parameter imaging principle of the binocular camera.Experiments show that YOSM can detect the target at the speed of 33 frames per second and output its spatial position three-dimensional information,which shows the effectiveness of the algorithm.(4)An advanced and fully functional visual servo open-source simulation system for aerial manipulator was developed and designed for the characteristics of the research object.The system is built based on ROS system and 3D physical engine,which solves the problems that the current aerial manipulator simulation only verifies the controller through MATLAB,and can not restore the real scene,image processing,physical attributes,and so on,and restores the motion control of the real aerial manipulator to the greatest extent.The visual servo simulation system designed in this paper realizes the target detection,tracking,and grasping tasks of the aerial manipulator,which provides great convenience for the research of visual servo of the aerial manipulator.