Research On Objects Recognition And Grasping Position Planning For Robot Automatic Assembly

As indispensable and essential parts in the process of automatic assembly,objects recognition and grasping position planning have a vital influence on assembly quality.The methods of mechanical parts recognition and grasping planning based on vision technology can obviously improve the automation flexibility of products assembly,therefore,time and cost can be reduced greatly and manufacturing efficiency can be improved.The study on objects grasp planning at home and abroad mostly focus on stable grasping position for simple grasp-and-place purpose,which can't be applied to the automatic assembly task.The paper is aimed at automatic assembly unit and accomplish objects recognition based on 3d point cloud and parts model database.The paper first defines the grasping parameters for two-finger manipulator,and finishes the interference detection for approaching points,approaching direction and contact position step by step.Based on force-closure,the stability of grasping position is judged qualitatively and calculated quantitatively to find an optimal grasping position which is suitable for automatic assembly.Details are described as follows:First,the topological structure of point cloud is established based on octree,which improves the search and calculation efficiency of point cloud data.The surface point cloud of CAD models is extracted by subdividing triangular mesh.The dense and scattered scanning point cloud is preprocessed,and planar background of objects is segmented through region growth with plane-feature points as the original seed,which makes the segmentation perfect.Targeted different types of noise,the paper applies different methods to eliminate them.The simplification of point cloud is accomplished by down-sampled based on octree,which increased the processing efficiency.Based on the improved PPF feature and RANSAC,feature matching is accomplished with better accuracy and relatively higher speed.Then the factors which influence the grasp of objects are analyzed thoroughly.Interference detection for feasible region of the approach direction is carried out,during which the paper first get the interference region between the mechanical arm and thesurrounding box of parts qualitatively,and then calculates quantitatively the interference region with triangular plane.The interference detection of contact position is carried out through hierarchical box volume tree(HBVT).According to the assembly path and by means of simulating assembly,interference detection of approaching point,approaching direction and contact point is carried out sequentially.Based on force-closure,the stability of grasping is determined qualitatively,and the grasping quality is calculated accurately with the defined index,through which an optimal grasping pose suitable for assembly is obtained finally.Contrast experiments between the classical ICP and PPF feature matching algorithm are carried out which prove that the algorithm in the paper can improve matching accuracy and speed.The results of grasping position planning is verified through experiment.