Research On Machine Vision Servoing Algorithm For Unmannned Aerial Manipulation

The practical applications such as load multi degree of freedom manipulators on ground mobile robot systems has been widely recognized,which combines the dexterous mobility of mobile platforms with the powerful operational capabilities o f traditional robotic arms,has inspired and produced The birth of the operational aircraft.meanwhile.Conbined with high-definition airborne image sensing,it ’s could be widely used in military and civilian fields,such as manipulative work in abusive environment,high-altitude complex sample collection,warehousing and logistics.The rotorcraft robotic arm system has underactuation,strong coupling,and nonlinear characteristics,and faced the challenges complex environmental disturbances in unstructured three-dimensional space and system dynamic uncertainty.In recent years,researchers have achieved stage results and major breakthroughs.In this paper,a multi degree-of-freedom flying robot with a multi-joint manipulator is designed,and the binocular camera is used to acquire the 3D point cloud for the target object’s recognition tracking and pose estimation,and finally realize the system’s semi-autonomous in the field of view,visual servo guida nce and target capture,as follows:1)The kinematics and dynamics model of the rotorcraft manipulator are established.considering about the influence of the manipulator motion on the change of the center of gravity,we derive the Newton-Eulerian model based on the force balance.2)Because of the underactuated structure,it is easy to deviate from the field of view of the airborne camera,a new hybrid visual servo control method is proposed.The image space controls the translation and the cartesian space controls the rotation,which reduces the influence of the translation and rotation,and achieves the decouping,improves global stability and immunity of the system to structural factors,Validated by MATLAB simulation environment.3)For the real-time tracking and pose estimation of the target in the visual servoing process,based on the Camshifts tracking algorithm,a three-frame differential detection and recognition method is proposed to locate the target,and then the point cloud acquired by the binocular camera is used to implement feature extraction matching and achieve real-time pose estimation of the target,Finally,the feasibility of the actual target experiment on the ROS platform is verified..4)Meanwhile,a multi-task decomposition control algor ithm is proposed for the multiple degrees of freedom and underdrive characteristics of the RAM,which weakens the coupling effect of the whole servo process and the influence of obstacles and changes of the center of gravity.The simulation results show the feasibility of the method,The simulation results verify the improvement of the control mechanism based on the servo simulation..In this paper,the Matlab/Simulink simulation platform and the ROS platform based on Ubuntu system is used to verify the hybrid visual servo and hierarchical task method,real vision servo verification,pose estimation and binocular tracking.It provides a sufficient theoretical basis for the practical research of operable aircraft.