Pose Measurement Of A Space Tumbling Target And Trajectory Planning Of Dual-arm Capture Robot

With the continuous deepening of space exploration,the number of on-orbit spacecraft is increasing.Apart from a few them are still working normally,most of them have been abandoned at the end of their lives,and some of them are unable to work due to malfunction.As a result,orbital debris continues to accumulate.The use of space robots for satellite maintenance and orbital garbage cleanup has become a hot research topic for aerospace majors in recent years.However,as non-cooperative targets for on-orbit maintenance,most of them are in a tumbling state,and their dynamic characteristics are extremely complicated.This paper focuses on the need for on-orbit recognition and capture of tumbling targets,and develops a dual-arm space robot cooperative measurement and trajectory planning method,including: dynamic coupling characteristics of a dual-arm space robot and the target,a image fusion enhancement and pose measurement method based on binocular stereo vision,a non-cooperative target collaborative measurement method based on virtual stereo vision,a dual-arm collaborative capturing method based on parameter estimation,and so on.The experimental verification is carried out by the ground-based experimental system.Focusing on the problem that dynamic coupling between the base and the end,the two arms,the arms and the target,a modeling method is introudced to solve the dynamic coupling between the dual-arm space robot and the target.According to the linear momentum equation integrable,while it is a holonomic constraint,and the angular momentum equation is an irreguable constraint,it is possible to divide the kinematics equations into the position and velocity levels,and then solve the corresponding equations independently.This method extends the dynamic coupling characteristics of base-to-end and joint-to-end of a conventional single-arm space robot to the dual-arm space robot system.The dynamic coupling characteristics between the two arms and between the two arms and the base were evaluated,and the corresponding coupling maps were drawn.Furthermore,the kinematic coup ling relationship between the two arms and the tumbling target is established.The resolved motion rate control strategy is used to obtain the motion constraint relationship between the actual velocity and the desired velocity of the end-effector.The pose deviation adjustment gain is solved by the optimal quadratic programming method.According to the dynamic gain provided by the optimization algorithm,the dual-arm space robot can quickly approach the capture point.Focusing on the problem that the non-cooperative target does not have any cooperation marker,and always in a tumbling state,an image fusion enhancement and pose measurement method based on binocular stereo-vision is proposed,which solve the problem of blurred edges and large measurement erro r in moving target recognition.The engine nozzle is used as a recognition object,a binocular stereo vision system is designed according to the measurement requirements,and the measurement performance parameters are analyzed.In order to deal with the phenomenon of image smearing that occurs during the movement of the tumbling target's,the smeared image is restored through blind deconvolution algorithm.Further,based on the essence of space circle and near-circle feature imaging,the pose calculation equations suitable for efficient numerical solution are deduced.Independent five parameters of ellipse perform complete space conic coordinate transformation,avoiding complex coordinate system transformation of traditional methods,reducing process errors,and improving the accuracy and speed of pose calculation.The image pre-processing algorithm and feature recognition algorithm are designed into a parallel architecture of FPGA,which further speeds up the entire visual measurement.Focusing on the problem that a single-arm space robot carries only one camera and cannot independently form stereoscopic vision,a non-cooperative target collaborative measurement method based on virtual stereo-vision is proposed.According to the results of the imaging of the hand-eye camera carried by each robot arm,the respective identified features are converted into equivalent spatial circle features,and the relative relationship between the two manipulators and the isotropic characteristics of the circular features are combined.Build a "virtual stereo vision-virtual common feature" measurement framework to achieve pose measurement of non-cooperative targets.The two robot arms respectively identify the engine nozzle and the triangular bracket of solar windsurfing.The fo rmer directly obtains the circular feature,while the latter obtains the characteristics of the inscribed circle based on the triangle bracket recognition.Combined with the geometric relative pose relationship between the satellite nozzle and the triangul ar support through other previous means(based on other measurement methods,such as the base measurement system firstly obtains the 3D point cloud information,and obtains the key component geometric information of the target after 3D reconstruction,which is used as prior knowledge here).The equivalent stereo vision measurement equation was deduced and solved by least-squares method to obtain accurate pose information of the target.This method provides a new measurement approach for a non-cooperative target,reducing the number and complexity of visual sensors and reducing the cost of the measurement system.Further,aiming at the motion estimate problems that time delay of visual measurement and unknown inertial parameters of the target,a hybrid Kalman parameter estimation method that can simultaneously estimate the target motion state and inertial parameters is proposed,as well as a corresponding dual-arm coordination trajectory planning method is presented.Since the target is in a moving state,there is a time delay between the vision measurement system and the robot control system,and the mass characteristics of the target are unknown,which poses a great challenge for a dual-arm coordinated capturing.Considering that the Extended Kalman Filter(EKF)is very fast for system iteration and the Unscented Kalman Filter(UKF)can obtain better parameter estimation accuracy,an efficient Hybrid Kalman Filter(HKF)combining EKF with UKF is presented in this paper.It can simultaneously estimate the attitude,angular velocity and all inertial parameters.It applies to parameter estimation problems in two typical cases,i.e.rotation around the principal axes of inertia and rotation around arbitrary axes.Based on velocity decomposition control strategy,the estimated motion and inertia parameters are used to plan the trajectories of a dual-arm space robot to capture the tumbling target.At last,based on kinematic equivalence principle,the actual target capturing experimental system for a dual-arm space robot is apllied,the visual measurement and dual-arm coordination trajectory planning method for a non-cooperative tumbling target was verified through experiments.