Vision-Guided Battery Assembly

01

Assembly Process

Small discs traveled on a conveyor where a backlit camera detected each part and returned pickup coordinates. A robot picked the disc, presented it to a second positional camera, received information about the disc relative to the gripper, and then moved to the battery assembly.

Two controlled dots of adhesive were dispensed on the target surface before the robot placed the disc into one of 198 positions.

02

Why a Standard Pallet Was Not Enough

The 198 positions followed a semicircular or otherwise irregular layout. A rectangular row-and-column pallet model could not represent the physical target pattern or the way error changed across the working area.

03

Precision and Plane Alignment

The process was sensitive to the relationship between fixture level, robot plane, battery orientation, and UVW angles. A small mismatch near one point could become a larger placement difference farther across the assembly.

Calibration therefore combined mathematical alignment with repeated observation on the real machine.

04

Dynamic Offset Strategy

My original contribution was the moving, per-cell offset strategy. The second camera measured how each disc sat relative to the gripper, and the system used that information to adjust the target placement for the current cell.

That approach addressed variation at the point where it mattered instead of relying on one fixed correction for every position.

05

Calibration and Tuning

I worked through robot frames, plane alignment, UVW calibration, camera feedback, adhesive placement, and iterative commissioning as one connected process. Each adjustment was evaluated on hardware so the model and physical behavior stayed aligned.