Robotic grasping method with 6D pose estimation and point cloud fusion

With the advancement of robotics technology, industrial automation has become a trend in many scenarios. In fields such as warehousing logistics and service robotics, a reliable method for estimating the grasp pose of objects has become a necessary task. At present, many robotic grasping systems have achieved good results. However, due to the irregularity of the object surface and variation of illumination, the point cloud obtained by a single view has large holes and errors at the edge of the object. These errors easily lead to wrong estimations of grasp poses. To address these issues, this paper proposes a practical robot grasping method. The method is based on the idea of 6D pose estimation and point cloud fusion and complements the input point cloud with the model through 6D pose estimation. Then, the fused object point cloud is used to estimate the grasp pose through the Angle-View Net and fast search strategy. In order to ensure the accuracy of point cloud fusion, we use iterative closest point to correct the 6D pose. Finally, we conduct experiments in simulated and real environments, and the performance evaluation shows the feasibility of the method.