3D Vision Guided Automatic Battery Cell Loading

The new energy automation production line is continuously upgrading towards high flexibility, high precision, and fast cycle times. The loading and unloading of workpieces such as battery cells and cylindrical shells impose strict requirements on positioning accuracy, recognition stability, and equipment compatibility. Traditional robot grasping methods relying on manual teaching and fixed tooling struggle to adapt to scenarios such as AGV feeding deviations, small workpiece gaps, and mixed-line production of multiple specifications. 3D industrial vision has become the core technical solution for high-precision depalletizing and palletizing.

This project was implemented at a domestic new energy equipment integrator site, where a 3D vision system guides the robot to grasp and place battery cells and cylindrical shells.

Project Challenges

Extremely small gaps, almost zero tolerance for error The gap between the battery cell and the protection board groove is only 1 mm. Slight deviation in visual positioning can cause jamming or workpiece damage during placement. Multi-layer stacking, alignment must be consistent The protection board may shift, so the vision system must precisely locate the center of the protection board when stacking the first layer and ensure that subsequent layers follow the same reference. Once the reference deviates, the entire stack will become increasingly misaligned.

High cycle time pressure The production line is a continuous flow, so the vision cycle time must keep pace with the robot; otherwise, it becomes a bottleneck for the entire line. Complex working conditions On-site lighting, tray/protection board deformation can easily affect imaging and recognition stability.

Solution

Adopting an "eye-in-hand" configuration, the Epic Eye Pixel Pro 3D smart camera is mounted on the robot end-effector, taking vertical downward shots to cover workpiece and tray pose detection, achieving full-process closed-loop guidance.

Workflow: Step 1: Protection board positioning. The robot carries the camera above the palletizing position, capturing two images of the bottom protection board diagonally to determine if the protection board has rotated 90°. If rotated, an alarm is triggered; if normal, the center pose of the protection board and all battery cell placement positions are calculated. Step 2: Battery cell imaging and recognition. The robot moves to the depalletizing position, and the camera captures part of the battery cells to calculate the pose information and grasping coordinates of a full row of 10 battery cells. Step 3: Layer-by-layer grasping and palletizing. The robot grasps one full row at a time and places it into the corresponding groove of the protection board. After each layer is completed, an intermediate cover is placed, and the next layer proceeds until the entire stack is finished.

Value

This project achieves stable vision guidance for battery cell and cylindrical shell loading/unloading scenarios. In actual production, the direct contributions are reflected in three aspects: Significant improvement in precision and stability: Achieves ±0.5 mm repeatability for battery cells and ±1 mm for cylindrical shells, with a grasping success rate ≥99.9%, eliminating faults such as empty grasping, misalignment, and jamming. Production efficiency and cycle time met: Single capture and recognition time ≤3 s, with one-time positioning of tray/protection board reused throughout the entire process, meeting high-speed production requirements of automated lines. Enhanced line flexibility: One system is compatible with both battery cells and cylindrical shells, supports quick switching between multiple stack types and trays, making line reuse and expansion more flexible.