DaoAI 3D Robot Vision Enables Random Bin Picking of Auto Parts

With the development of the embodied intelligence trend, devices such as lidar lawn mowers have put forward higher requirements for 3D grasping and guiding functions. In the automotive parts production field, the 6D pose problem of random bin picking also needs to be solved urgently, and DaoAI 3D robot vision of WeLinkirt emerges as the times require.

User Scenario: On the production line of a leading automotive parts supplier, it is mainly responsible for producing various parts of automobile engines. In the feeding process of parts, the parts are randomly placed in the bins, and they need to be accurately grabbed and placed at the specified processing positions. The detection objects are these engine parts with different shapes and random arrangements.

Pain Points: The traditional grabbing methods have many problems. On the one hand, due to the random arrangement of parts, the manual grabbing efficiency is very low, only about 60 grabs can be completed per hour, and the labor cost is high. On the other hand, manual operations are prone to missed grabs and mis-grabs. The missed-grab rate reaches about 5%, and the mis-grab rate is about 3%. In the trend of embodied intelligence, devices such as lidar lawn mowers are all pursuing more efficient 3D grasping and guiding functions, and the automotive parts production also urgently needs to improve the efficiency and accuracy of random bin picking.

Technical Principle

DaoAI 3D robot vision uses a self-developed 3D camera for imaging. The camera uses the principle of structured light. By projecting a specific structured light pattern onto the object surface, and then the camera captures the deformation of the reflected pattern. Since different shapes and positions of objects reflect and deform the structured light pattern differently, according to the triangulation principle, the three-dimensional coordinates of each point on the object surface can be calculated, thus realizing the three-dimensional shape reconstruction of the parts.

• In terms of 6D pose estimation, a deep-learning algorithm is used to analyze the reconstructed 3D point-cloud data. Through a large number of labeled samples for training, the model can learn the feature and spatial pose information of different parts, so as to accurately predict the 6D pose of the parts.
• For random bin picking, the system will plan the best grabbing path according to the 6D pose information. Considering the mutual occlusion and spatial layout of parts in the bin, the algorithm will avoid obstacles to ensure that the robotic arm can accurately and stably grab the target parts.
• In the process of glue coating, assembly and loading and unloading guidance, the system will monitor the position and pose of the robotic arm in real-time and compare it with the target position. Through the brain-eye - body closed-loop control, the motion trajectory of the robotic arm is continuously adjusted to achieve sub-millimeter hand-eye coordination and ensure high-precision operation.

WeLinkirt's Solution and Product

Centered on DaoAI 3D robot vision, this product has high-precision 6D pose estimation ability and can accurately identify the position and pose of parts in the random bin. In the implementation, first, the self-developed 3D camera is installed in a suitable position to scan and image the parts in the bin in real-time. Then, the 6D pose estimation algorithm is used to process the imaging data to obtain the accurate pose information of the parts. Next, this information is transmitted to the robotic arm control system to plan the best grabbing path.

DaoAI 3D robot vision realizes the brain-eye - body closed-loop, making the operation of the robotic arm more accurate and efficient.

Quantitative Results: After using DaoAI 3D robot vision, the grabbing efficiency has been greatly improved. About 180 grabs can be completed per hour, which is 3 times the efficiency of manual grabbing. The missed-grab rate is reduced to <0.5%, and the mis-grab rate is reduced to <1%, greatly improving the accuracy and stability of production. At the same time, due to the realization of automated operation, the labor cost is reduced, and the production cycle is also effectively controlled.