Robotic Depalletizing: Stretch Film, Interlayers, Incomplete Layers, and Exceptions

A pallet arrives at the depalletizing cell wrapped in stretch film, topped with a slip sheet, and stacked in layers that were packed somewhere else by someone else. The robot is expected to start picking within seconds. A successful carton pick proves only one part of the process. Film, sheets, damaged cartons and the transition to the next layer need equally clear handling.

The question a buyer should ask is not "can a robot pick a carton." It is "what does this incoming pallet actually look like, and who handles the parts a robot should not touch." This article works through that question in the order a project engineer should: the real state of the incoming load, why film and interlayers are a separate problem from picking, how incomplete and mixed layers break the plan, what vision can and cannot do, how to record exceptions, and how to assemble a depalletizing evidence chain.

Define the Incoming Pallet State

Depalletizing specs are usually written against an ideal pallet: uniform cartons, square corners, a clean wrap, full layers. The pallets that actually arrive are rarely ideal. Film is torn or double-wrapped. A slip sheet sits between layers. The top layer is half empty because the last order was short. One carton is crushed. The stretch film has been re-applied by hand at the dock.

Before anyone quotes a cell, the buyer should define the incoming condition as a set of measurable states rather than a single "normal" case. A practical starting list:

• Pallet condition: intact, leaning, damaged, non-standard footprint
• Wrap: single film, double film, film plus strapping, film plus top board
• Interlayers: none, slip sheet, cardboard sheet, foam, mixed
• Layer completeness: full layer, partial layer, mixed SKU layer, hand-stacked top
• Carton condition: uniform, varied sizes, damaged, overhanging

Each of these states changes what the cell has to do. A robot that handles full uniform layers well may have no defined behavior for a half layer. If the spec only describes the ideal pallet, the gaps show up on the floor, not in the proposal.

For the distinct upstream task of building stable outbound loads, see our palletizing robot system selection guide; inbound depalletizing still needs its own exception matrix.

Separate Unwrapping from Picking

This is the most common structural mistake in depalletizing projects: treating "remove the wrap" and "pick the boxes" as one capability. They are two different tasks, often done by two different machines, and a robot arm with a gripper is usually only good at one of them.

Stretch film has to be cut, gripped, pulled, and disposed of. That is a film-handling problem. Picking cartons off a layer is a second problem. Dedicated film-cutting or automatic film-removal equipment may be used before a depalletizing robot starts picking. Specify its handling and recovery separately. Assuming a single arm will do both is how projects end up with a robot that picks beautifully but cannot get to the first layer.

The honest question for a buyer is who cuts the film. If the answer is "an operator walks up and does it," that is a valid answer, but it must be written into the plan as a defined manual step with a defined trigger, not left as an assumption. If the answer is "the robot does it," the buyer needs the film handling to be specified as its own subsystem, with its own failure modes.

Handle Interlayers and Empty Packaging

Below the film sits the layer of material that is even easier to overlook: slip sheets, cardboard sheets, foam, and the empty packaging left behind as the pallet is unloaded.

Slip sheets and interlayers have to be lifted and removed without dragging the layer below. That is a sheet-handling problem, separate from both film removal and carton picking, and it has its own machines and its own failure modes. A sheet that is dragged instead of lifted takes the carton beneath it out of position, and the pick pattern that follows is built on a layer that has already moved.

Empty packaging is the other half of this problem. As cartons come off, the film, the sheets, and the empty trays have to go somewhere, and where they go is a design decision. If the cell has no defined path for empty packaging, it accumulates on the pallet and blocks the next layer. The buyer should specify, for each layer, what is removed, by what, and where it goes, so that unwrapping, sheet removal, and packaging disposal are three named steps rather than one vague "the robot handles it."

Manage Mixed and Partial Layers

A full, uniform layer is the easy case. The hard cases are the ones that break the layer model:

• A partial top layer, where the robot has to reach past empty space to the remaining cartons
• A mixed-SKU layer, where cartons are different sizes, weights, and pick points
• A hand-stacked top, where the arrangement does not follow the pattern the robot was taught
• A layer with one crushed or leaning carton that must be skipped

Vision-guided depalletizing systems address part of this. A suitably configured 3D-vision depalletizing system can estimate accessible carton poses and support picking from some non-uniform layers, within its trained detection and gripping envelope. Mech-Mind's published liquor-company case describes recognizing cartons with different sizes and patterns, selecting picking points, and using collision detection and path planning. Its version 1.8.3 single-case depalletizing guide separately identifies supported carton conditions and excludes surface-damage detection. That is a real capability, and it is worth having.

But note what it does not say. It does not say the cell needs no operator, and it does not say every exception is handled automatically. Vision expands the range of layers a robot can pick from. It does not remove the need for a defined behavior when the layer does not match any trained pattern. The buyer should ask what the cell does when vision cannot resolve a carton: does it skip, does it stop, does it call an operator, and how is that recorded.

Protect Products During Exceptions

It helps to separate what a vision system is actually doing on a depalletizing cell, because the exceptions are where products get damaged.

Ask the vision trial to demonstrate carton localization, pose estimation, gaps and overhangs, pattern checks and remaining-carton counts where those functions are specified. Confirm each required output against the actual camera view, carton condition and software configuration.

Vision alone does not physically cut film or lift an interlayer. Require separate evidence for damage detection and stack-stability decisions rather than assuming they follow from pose estimation. Assign unresolved cartons and unstable loads to a defined machine or operator recovery process.

A vendor performance figure is not a substitute for this. ABB's robotic depalletizing product page lists product-specific figures such as pallet heights up to 2.8 m, peak rates up to 750 cycles per hour, and payloads up to 30 kg. Those are the vendor's own numbers for its own product, and they are a useful reference point for what an industrial depalletizer can do. They are not a baseline for every cell, and they say nothing about how the cell handles film, sheets, or a partial layer. When a supplier quotes a rate, the buyer should ask what incoming condition that rate assumes.

The value of a depalletizing cell is not only in the cartons it picks. It is in how it handles the cartons it does not pick. A cell that quietly skips a crushed carton and moves on is hiding a problem. A stop-and-wait response needs a defined safe state and controlled recovery; stopping alone does not establish that the load is stable. Choose the response for each exception and measure its effect on throughput.

The exception plan should name, for each abnormal state, what the cell does and what it records:

• Torn or missing wrap: whether to quarantine, stabilize or remove the remaining film, and who is notified
• Interlayer present when none was expected: does the cell stop
• Partial layer below a full layer: how the pick pattern adapts
• Damaged or non-conforming carton: skip, stop, or route to a reject lane
• Pallet not matching the taught pattern: escalation path

Every one of these should produce a record. "Operator handled it" is fine as long as the record says which exception, when, and what was done. Without that record, the cell's real performance is invisible, and no one can tell whether the exception rate is improving.

Run an Incoming-Pallet Trial Matrix

The output of this work is a single table that ties the incoming pallet to the cell's behavior. One row per incoming state, with columns for: the condition (wrap, interlayer, layer completeness, carton condition), the expected handling (robot, film cutter, operator), the detection method, the exception behavior, and the record produced.

Build that table and the depalletizing question answers itself. The robot takes its place as the picking subsystem, film and sheets get their own defined handling, incomplete layers get a defined behavior, and every exception leaves a trace. The buyer who defines the incoming states before specifying the cell avoids the most expensive depalletizing mistake: buying a robot that picks perfectly and has no plan for the pallet that shows up on a bad day.

Turn pallet exceptions into a usable cell requirement. Before choosing an arm or gripper, send the incoming-pallet formats that actually reach your dock: photos of full and partial layers, wrap and strapping types, interlayer samples, carton dimensions, damage frequency, and the expected cartons per hour. Use our depalletizing application assessment to organize the intended picking task, the separate film- and sheet-removal steps, and the unresolved exceptions. We can help structure the information gaps and questions for prospective manufacturers or integrators.