基于3维点云深度信息和质心距相结合的机器人抓取控制方法

Aiming at the problem that it is difficult to identify the position when robots grab irregular objects, this work puts forward a comprehensive robot grabbing control method based on 3D point cloud the normal vector angle and the centroid distance. The establishment of this method based on a comprehensive grab evaluation method of double target avoids the problem of insufficient judgment of the grabbing factors by the single target grabbing evaluation method. Firstly, the scene point cloud of Kinect sensor picking up is segmented, denoised, downsampled, and the target boundary point cloud is obtained. Then, through combining the evaluation function of the angle between boundary points cloud and the normal vector and the function of the distance between the grabbing line and the mass-center of the object grabbed, a synthetical evaluation function is proposed to obtain the optimal grabbing point by evaluating different grabbing positions. Finally, the simulation experiment is carried out on the comprehensive evaluation method and the feasibility and effectiveness of the method are verified.