Research On Autonomous Grasping Method For Non-Cooperative Target Based On Multi-Rotor UAV

This paper focuses on the autonomous aerial manipulation problem of the multi-rotor UAV based on visual servo control.A configuration of hexa-rotor UAV system equipped with a 4R robotic manipulator is designed.Research on the system dynamics modeling,adaptive control and trajectory planning method of the UAV and manipulator,visual servo control of the system and other technical issues are carried out as follows:The dynamics model of UAV-manipulator complex system is established.According to the UAV and manipulator configurations selected,corresponding coordinate systems are set based on open-source flight control standard and modified Denavit-Hartenberg(D-H)parameters respectively.Based on kinematics analysis,the dynamics model of the composite system is established according to Lagrange equation.The subitems’ characteristics of Inertia,Coriolis and Gravity of the dynamic model are analyzed.In order to facilitate the control system design,a simplified system dynamics model is deduced.An adaptive control method based on inertial compensation UDE-GS method is designed.Based on the analysis of the general underlying cascade PID control logic of hexa-rotor UAV,uncertainty and disturbance estimator(UDE)as well as gain scheduling(GS)logic are introduced into the control system,and the feedforward compensation was designed according to the variation of inertial parameters of the real-time configuration of the manipulator.The inverse kinematics of the manipulator was solved based on the quasi-Newton iteration method,and the trajectory of the manipulator in joint space was planned using the cubic polynomial method of "receding horizon".The feasibility and robustness of the proposed method are verified by comparison simulation of the basic configuration of the system and the extreme configuration of the manipulator.A visual servo control scheme combining object detection and EPnP pose estimation method is proposed.An object detection and recognition model based on YOLOv5 s is trained for ground micro-explosive target.EPnP method is further used to estimate the pose of the detected object feature points.A combined visual servo system contained an airborne fixed camera(Eye-to-Hand)and an end follow-up camera(Eye-in-Hand)is adopted.The position-based visual servo(PBVS)and image-based visual servo(IBVS)control methods are used in the approaching and grasping phases respectively to realize the accurate grasping of the target.The co-simulation scheme based on Simulink-ROS platforms is carried out.The simulation model of the integrated system was built based on Simscape multibody dynamics module.The UAV-manipulator visual servo control scheme designed in this paper is combined with Simulink and ROS for closed-loop simulation.Fixed stereo-target and moving target were respectively set and appropriate disturbance is added.The feasibility of the whole system visual servo control scheme under different working conditions is verified.The above object detection and visual servo control scheme were also verified by partial physical experiment on M600 Pro UAV platform.