Pallet docking

​

How it works and what it returns

Two recognition modes are configured and validated separately. The 2D image model uses depth to recover a 3D docking pose and classifies the lower entry-face beam as Closed or Open. Pure 3D recognition uses point clouds, templates and calibration; it can distinguish empty and loaded pallets when configured, but does not measure weight. A failed 2D detection does not trigger an automatic switch to pure 3D. Outputs include entry-reference X/Y/Z in millimetres, roll and yaw in degrees, detection status and count. Pitch is fixed at zero and is not a valid measurement. Supported two-, three- and four-leg pallets depend on the selected model or template; pure 3D near/far calibration uses a three-leg pallet.

​

Operating and installation conditions

The white paper lists maximum detection distances of 1.8 m for S11 with an external RK3588 and 3.0 m for M4 Mega with onboard computing. Minimum distance depends on pallet geometry and field of view. Reflective surfaces are unsuitable; black pallets require camera-specific sample testing. Validate both recognition modes with the actual load and background. Install only the AW3 frontend, without its backend. Deploy SmartDocking 3.0.2 on the compatible RKU20 / Linux ARM64 device. Earlier PalletPro software and its interface are maintained as a separate archive, not as an AW3 frontend or SmartDocking package.

​

Deployment and integration

Watch the English video on YouTube

​

Latest downloads

​

Software downloads

​

Manual and white paper downloads

Software and document versions are managed separately. Select archived material for the application you have installed. GitHub