Research On Robot Measurement Technology Of Space Complex Slender Tube

Tube is often used as the key parts of braking system in automobile,aircraft and other products,and its spatial shape machining accuracy directly affects the life and safety of products.The measurement of its spatial shape is the basis of generating qualified tube products.However,it is difficult to realize the automatic measurement of the tube due to its complex surface,long and thin structure and many kinds of characteristics.To realize automatic measurement of slender tube in industrial production site,a tube robot measurement system based on robot,linear stage and various sensors is designed in this paper.It has the advantages of robot flexibility and laser profilometer precision and high stability,and it is a good solution to realize the automatic measurement of slender tube spatial shape.Research on the measurement accuracy and measurement process automation of the tube robot measurement system is as follows.To maintain the measuring accuracy,the tube robot measurement system needs to be calibrated regularly when it is used for a long time in the industrial production site.Therefore,a rigid-flexible autonomous calibration method for robot measurement system is proposed in this paper.Firstly,by introducing the mass center coordinate system of robot link and its Jacobian matrix,the torques on each joint of the robot generated by the gravity of robot link and load are calculated systematically.Based on the theory of linear torsion spring,a rigid-flexible measurement model of the robot measurement system is established.Secondly,the optimal objective function is established by the standard sphere center constraint.The robot kinematics parameters,joint stiffness coefficients and hand-eye parameters can be identified simultaneously by using iterative singular decomposition method.Then,a simulation method of calibration data generation is designed to analyze the influence of standard ball center measurement error on parameter identification accuracy,which can be used to eliminate the parameters with poor identification and guide the design of 3D camera accuracy.Finally,by using a single standard sphere,the spherical coordinates are collected by the 3D camera in the robot measurement system,and the autonomous calibration of the robot measurement system is realized.Although the calibration method improves the measurement accuracy of the robot measurement system,there is still dislocation in the overlap area of point clouds from different views.In order to reduce the position error caused by point cloud dislocation,a point cloud dislocation correction method based on multi-view point cloud registration is proposed.Some auxiliary views are added near key view,and the relationship of point cloud pairs is obtained according to pair-wise point cloud registration.The optimization objective function of multi-view point cloud registration is established,and camera position and pose error of the key view is solved by trust region method.The camera position and pose error is substituted into the compensation model to reduce the point cloud dislocation error and improve the point position accuracy.The tube position and pose in the robot base coordinate system is needed before scan path planning.Due to the complexity of tube characteristics,common visual localization system cannot locate the slender tube stably and quickly.Therefore,a position and pose estimation method for slender tube based on sensor data fusion is proposed.Firstly,a localization unit based on array camera and laser displacement sensor is designed.Secondly,a coordinate calculation model of key location points on cylindrical axis is established by integrating the image of the array camera,the distance of the laser displacement sensor and the position and pose of the robot.Then,the key location points are registered with tube 3d model to achieve position and pose estimation.Finally,a sensor relationship calibration method is proposed to improve the calculation accuracy of key location point coordinates.According to the laser point images obtained by the array camera and the distances obtained by the laser displacement sensor,the constraint equation is established,and the identification of the sensor relationship parameters is realized by singular value decomposition method.The complex spatial trend of slender tube and the small measurement range of3 D camera make manual teaching scan path time-consuming and laborious.Therefore,an autonomous approximate optimal scan path planning method for slender tube is proposed.Tube scan path planning is divided into key scan view planning and continuous path segment planning between key scan views.The latter is achieved through bilateral RRT approach.Aiming at the key scan view planning,the number of key scan views is taken as the optimization objective,and the position and pose sampling space containing pose constraints is established.By using random sampling and genetic algorithm,a set of key scan views with as few scan times as possible and complete coverage of tube could be obtained.In addition,a forward coverage factor is introduced in the iterative process of genetic algorithm,which increases the probability of sampling points falling in the optimal individual small neighborhood and improves the planning efficiency.Aiming at the automatic measurement of tube spatial shape,the experimental platform of the tube robot measurement system is established,and tube measurement software is developed,which provided the foundation for experimental research and subsequent engineering application.Compared with three coordinate measuring machine,experiments verify that the absolute accuracy of the robot measurement system can reach 0.377 mm with the rigid-flexible autonomous calibration method and point cloud dislocation correction method.According to the proposed method of tube position and pose estimation and scan path planning,the actual scan coverage of slender tubes is more than 99%,and the scan path planning time is less than 2 s.For the actual automobile pipeline measurement,the maximum coordinate error of tube nodes is 0.504 mm.These experiments verify that the tube robot measurement system and related technologies can meet the needs of the vehicle tube spatial shape measurement.