Pose Estimation And Gripping Of Poles With Biped Robots

The biped climbing robot moves in pole environment,and each movement of the robot involves the problem of pole gripping,which requires the robot to correctly identify the pole and obtain the specific position of the target pole.The robot hand claw and the target have rigid contact when the robot grips the target.In order to improve the gripping reliability,the posture accuracy of the pole is required to be as high as possible.The robot hand-eye calibration and the gripping strategy play an important role.As a kind of special mobile robot,the climbing robot has three different climbing gaits,which generates the motion trajectory through the gait planner.The specific work and innovation of this paper are as follows:1.Combining image processing and point cloud data,an algorithm is proposed to recognize the graspable-pole and estimate its posture.By analysing the features of the pole,the image filter,edge detection,line detection and other operators are used to identify the target object in the image and locate its position.The image information was used to segment the pole point cloud data,and the pole pose was obtained by fitting methods.2.The robot needs to know the specific position and orientation of the target pole to implement the pole clamp.The results of the detection pose need to be transferred to the robot coordinate system before gripping the poles.In this paper,camera calibration is carried out and the calibration accuracy is evaluated by HALCON software.3.After obtaining the pose of the pole by the camera,a method for evaluating the detection accuracy of the pose is proposed.The UR3 platform is used to obtain the true value of the axis direction of the pole using the method of the sharp point touching,and the real value is compared with the detection results of the camera.4.Due to the special configuration of the climbing robot,the gripping posture of the robot is limited,sometimes the robot cannot reach the initial pose of the pole and the robot has no corresponding inverse kinematics.Combining the graspable characteristics of the round pole and the square pole,the methods are proposed to adjust the position and pose of the target poles.5.The climbing robot has three gait motions,and the single-step motion is generated by DMP algorithm.Given the teaching trajectory and changing the starting and ending positions,the robot can reproduce the movement of the gait trajectory.Combined with the global target point,the robot claw and the camera can reach the periphery of the pole,and then use the camera to detect the pose of the pole and implement accurate gripping.