Why Depalletization Robots Struggle With the Last Layer

Written by Christian Aschehoug | 2026-08-10

If you've spent time on the floor of a depalletizing cell, you've probably seen this: the robot handles the top of the pallet at full rate, then performance slows down as it works toward the bottom. Layer 1 looks great; the last layers are where the trouble starts.

In this blog, we look at why the last layer is the hardest case in vision-guided depalletizing, and what camera properties make the difference.

Working distance increases, so what?

Three important factors come into play when we extend the working distance of any camera.

Point cloud noise scales with distance

A stationary 3D camera mounted above a 2.3 m warehouse pallet sees the top layer at around 1.5 m and the bottom at 3 m or more. Depth noise and dimensional trueness error grow with distance, and spatial resolution degrades as every pixel covers more area. Edges that were sharp on the top layer become fuzzy on the last one.

Tip: When comparing 3D cameras for depalletizing, ask the vendor for accuracy and resolution at the longest working distance you'll actually use, not the optimal one. Check our datasheets here.

Temperature changes make the camera drift

A depal cell normally has compliance built in, typically in the vacuum gripper. That compliance hides small calibration errors, so nobody notices them at the top of the pallet. The effect of thermal drift is more noticeable with distance. By the last layer, the error is larger than the compliance can absorb, and your robot starts missing objects, or worse, damaging them.

Ambient light eats your signal strength

Structured light and active stereo cameras depend on the light they project themselves, and that light has to be stronger than the ambient light landing on the same scene. Simplified, intensity (I) is inversely proportional to the square of the distance (d):

I ∝ 1/d²

In practice, double the distance and the projected light drops to a quarter of its strength. So the signal-to-noise ratio is far better at the top of the pallet than at the bottom.

What to look for in the camera

Point cloud of a pallet with mixed items and multiple layers captured with Zivid 3 XL250.

When specifying a camera for depalletization, evaluate it for the last layer. Three properties matter most:

Consistent point cloud quality across the full pallet height. Evaluate the point clouds yourself at the top of the pallet, but most importantly at the bottom. Data that looks good at 1.5 m can be a different story at 4 m.
Resolution that holds at distance. You want high enough resolution to see details at long distance. 8 MP point clouds preserve millimeter-scale features at 3+ m: slip-sheet edges, gaps between boxes, the seam between two sacks.
Reliable performance over time. Depalletization cells often sit close to bay doors, where ambient light and temperature vary through the day and through the year. You want a camera that is factory calibrated across its full temperature range, so the numbers you validated during commissioning are the numbers you get twelve months in.

The takeaway

Last-layer failures may look like mechanical or planning problems, and in part, they are. But the robot can only avoid pallet-edges, choose the right approach vector, and confirm an empty deck if the point cloud on the last layer is as trustworthy as the one on the first. That's the quiet way camera choice determines whether your depalletizing robot's throughput holds through the entire pallet.

If you're looking for a depalletization camera, our depalletization application page walks through the full set of requirements, and the Zivid 3 XL250 is the camera we purposely built for these applications.

View full post