explainer
How robotic palletizing works
A palletizing cell picks each case off a conveyor and builds a stable pallet. Learn its parts, the motion and sensing behind each placement, where cells fail, and who makes the hardware.
- robotics
- automation
Overview
A robotic palletizer turns a stream of single cases into a stable, square unit load ready to ship. Done by hand, the job means lifting a 10 to 25 kg box, turning it and setting it in a precise spot, over and over for an entire shift.
Every cell, simple or complex, runs the same loop:
- Infeed: cases arrive on a conveyor, are spaced or accumulated, and are turned to a known orientation. A photo-eye confirms each one has reached the pick point.
- Identify: a barcode reader or camera confirms which product it is, so the cell knows the case's size, weight and pattern.
- Pick: the robot comes straight down, grips the case (vacuum, clamp or fork), and lifts straight up until it clears the conveyor guides.
- Sense and decide: the controller looks up the next slot in the pallet pattern. Cells with 3D vision also measure the stack's height, position and gaps and adjust before placing.
- Place: the robot swings over the pallet, finishes the swing above the slot, lowers straight down, releases, and lifts straight up.
- Pallet change: when the pattern is complete, the full pallet leaves the cell and an empty one arrives, often from a dispenser. A slip sheet may go down between layers.
- Secure: downstream, a stretch wrapper or strapper binds the load and a printer applies the shipping label.
The animation in Fig 1, the palletizing cell, shows steps 1 to 6 in one cell: a roller conveyor, a four-axis palletizing arm, a vacuum gripper and a 3D camera on a mast.
Companies automate palletizing mainly because it is heavy, repetitive work that is hard to staff across shifts. A robot also builds every pallet to the same pattern, which helps with truck loading and racking. One robot handles roughly 10 to 20 picks a minute, and gripping several cases or a whole layer at once multiplies that.
The cell, part by part
A palletizing cell is a small production line in its own right. Most of its engineering sits in the conveyors, grippers and pallet handling that surround the robot.
Infeed and case handling
Cases come off the packing line on a roller or belt conveyor. Accumulation zones let cases queue without crushing each other, so the robot is never starved and the upstream line never stops. Turners and pushers rotate cases to the orientation the pattern needs, and a hard stop or a metering belt presents each case at a fixed pick point. Some cells build a row or a whole layer on the conveyor first so the robot can pick several cases in one move.
The robot
Four main arm types are in use:
| Arm type | How it moves | Typical fit |
|---|---|---|
| 4-axis palletizing robot | Base rotation, shoulder, elbow and wrist rotation. A parallel-link arm keeps the tool level without a wrist pitch axis. | The workhorse. High payloads (roughly 80 to 800 kg across the main makers), long reach, fast cycles. |
| 6-axis articulated robot | Full wrist, so the tool can tilt and approach from any angle. | Mixed duty cells, awkward reaches, or when the same robot also does depalletizing or other tasks. |
| Gantry (Cartesian) | Moves on overhead rails in X, Y and Z, with a rotary tool. | Very large or multiple pallet positions, heavy layers, and lines with low ceilings over the robot base. |
| Collaborative robot (cobot) | A lighter arm, often on a column lift, with force limiting and area scanners instead of a full fence. | Lower speeds and loads, small footprints, frequent changeovers, and sites new to automation. |
Conventional (non-robotic) palletizers are still common at very high speeds. A layer former builds each layer on a table from a fast case stream, and a hoist drops the finished layer onto the pallet.
End-of-arm tooling
The gripper, or end-of-arm tooling (EOAT), is where most of the application knowledge lives:
| Gripper | Holds by | Good for | Watch out for |
|---|---|---|---|
| Vacuum (cups or foam pad) | Suction on the top face | Sealed corrugated cases, trays, shrink-wrapped packs | Porous or damaged board, open flaps, tape seams under a cup |
| Clamp | Squeezing two sides | Heavy or bulky cases, bundles, cases without a good top face | Crushing soft packs, needs space between cases |
| Fork or tine | Sliding under the load | Bags, unstable products, heavy layers | Needs a gap under the product; the fingers must slide out after placing |
| Bag gripper | Fingers that cradle a sack | Feed, cement, chemical and resin bags | Bag shape varies; patterns must suit soft loads |
| Layer gripper | Clamps or vacuums an entire layer | Full-layer handling, depalletizing | Large, heavy tool; layer must be uniform |
Many tools combine methods, such as vacuum plus a side clamp, or add a slip-sheet cup or pallet fork so one robot can also place sheets and empty pallets.
Pallet handling
A pallet dispenser feeds empty pallets into the cell, and a slip-sheet magazine holds tier sheets. Many cells use two pallet stations so the robot keeps working on one while the other is being changed. Full pallets leave on a roller conveyor to a stretch wrapper, or are taken by forklift or AGV.
Safety guarding
Industrial robot cells sit behind fixed fencing with interlocked gates. Light curtains or a tunnel with a muting system let pallets and cases pass through the guard while keeping people out. Area scanners slow or stop the robot when someone enters a zone. All of this runs through a safety-rated controller, separate from the logic that runs production.
Key technology
Kinematics and motion
A dedicated palletizing robot has four axes: base rotation (J1), shoulder (J2), elbow (J3) and tool rotation (J4). A parallel link, a rod that runs beside the upper arm, keeps the wrist flange horizontal at every pose. The robot never needs to tilt a case, so dropping the two wrist-pitch axes buys payload, speed and stiffness.
Three motion details separate a good cell from a jerky one:
- Solve for the tool centre point. The gripper often hangs to one side of the arm's centre plane, so inverse kinematics must put the tool centre point (TCP) on the target. Skip this and every pick and place lands off to one side.
- Approach and retreat vertically. The last few hundred millimetres into a pick or a slot are straight down, and the first move out is straight up. Horizontal motion starts only once the case clears the guides and its neighbours.
- Blend the path and limit jerk. Between those vertical legs the controller blends the lift, swing and descent into one smooth arc. Acceleration is capped at what the grip can hold, which depends on case weight and how well the cups seal.
Payload and reach are a trade-off. The payload rating must cover the gripper plus the heaviest pick (several cases at once if the tool does multi-picks), and the reach must cover the far corner of the tallest pallet.
Patterns and load stability
A pallet pattern says where each case goes, layer by layer. Two families dominate:
- Column stacking: every case sits directly on the one below. Corrugated boxes carry load on their corners and walls, so columns give the most compression strength, though nothing ties one column to the next.
- Interlocked (brick) stacking: alternate layers are rotated or mirrored so each case bridges a seam below it. The load holds together much better in transit, at some cost in compression strength.
Pattern software also checks overhang (cases past the pallet edge get crushed or snagged), underhang (wasted cube), weight distribution and centre of gravity, and the stack height limit for the truck or rack.
Mixed-case palletizing is the hardest version: building a stable pallet from many different SKUs for one store or route. The software picks the build order on the fly, much like packing a car trunk, while following rules such as heavy items low and crushable items high.
Sensing and vision
Sensors close the loop between the plan and the real world:
| Sensor | What it checks | Where |
|---|---|---|
| Photo-eyes and proximity sensors | Case present, spacing, jams | Infeed, pick point, pallet stations |
| Barcode readers (fixed-mount) | Which SKU is arriving | Infeed, before the pick |
| 2D cameras | Label presence and orientation, case position on the conveyor | Pick point |
| 3D cameras (structured light, time of flight, laser line) | Case pose, stack height map, gaps and leaning, pallet presence | Above the pick and the pallet |
| Vacuum switches and pressure sensors | Whether the gripper actually holds the case | On the tool |
| Load cells | Case weight, dropped or double picks | Infeed or tool |
In a fixed-pattern cell, the stack is placed blind and sensing is mostly confirmation. In mixed-case and depalletizing cells, 3D vision is essential: the robot must find each box in an unstructured pile and decide how to grab it.
Controls and integration
The robot controller runs the motion. A PLC usually orchestrates everything else: conveyors, stops, pallet stations, wrapper hand-off and recipes. Upstream, the cell talks to the plant's MES or the warehouse's WMS to learn which product and pattern come next and to report each finished pallet.
Safety runs on its own rated hardware: a safety PLC or safety relays read the gate switches, light curtains, area scanners and e-stops, and can remove power from the robot independently of the production logic. Modern controllers add safe-zone functions that fence the robot's motion in software, which can shrink the guarded footprint.
The plant MES or warehouse WMS sends the next SKU and pattern to the line PLC and receives "pallet complete" back. Sensors feed the line PLC, which sequences material handling. The line PLC and the robot controller trade handshakes on every case (case ready, pick done). The safety PLC (gates, curtains, scanners) sits outside that loop and can remove drive power whatever the production logic is doing.
Typical failure points
Most palletizer downtime traces back to the cases, pallets and conveyors around the robot. The arm itself rarely causes a stop.
| Failure | Usual causes | How cells prevent or recover |
|---|---|---|
| Dropped or shifted case | Vacuum leak on porous or recycled board, a cup landing on a tape seam or flap, too much acceleration for the grip, a heavier case than the recipe expects | Vacuum switches per zone, foam pads instead of cups, motion limits tied to case weight, a clamp or under-support for heavy items |
| Pick misses and double picks | Cases not at the stop, touching cases lifted together, wrong orientation on the infeed | Photo-eyes at the pick point, spacing belts, vision or a load cell to confirm one case |
| Leaning or collapsing stacks | Column patterns with no interlock, bulged or crushed cases, placing on a case that settled, overhang | Interlocked patterns, slip sheets, height checks with a 3D camera, compression-rated packaging |
| Pattern or data errors | Wrong SKU or recipe, mixed product arriving out of sequence | Barcode verification before the pick, recipes driven from MES or WMS, rejecting unknown cases to a lane |
| Bad pallets | Broken boards, missing blocks, wrong pallet type, pallets stacked crooked in the dispenser | Pallet inspection (vision or sensors), reject stations, dispenser alignment guides |
| Vision misreads | Glare from shrink-wrap or glossy print, changing ambient light, dust on the lens, dark or black cases | Controlled lighting, enclosures, regular lens cleaning, models trained on real product |
| Conveyor jams | Damaged or soft cases, accumulation pressure, guide rails set wrong for a new SKU | Low-pressure accumulation, jam detection eyes, quick-change guides |
| Throughput shortfall | Cycle time set by the worst case (far corner, top layer, single picks), slow pallet changes | Multi-pick tooling, dual pallet stations, simulation before purchase |
| Safety stops | People entering for minor jams, muting misconfigured on light curtains, nuisance trips | Zone-based slowdowns instead of full stops, good jam access, muting tuned for each load |
| Changeover pain | New SKUs need new patterns, tooling and guide settings | Pattern-building software, universal grippers, recipe-driven guides |
Two habits head off many of these. Test with real production cases, including ones that have sat in a humid warehouse, because fresh samples from the box maker grip better than anything the line will send. And design for the worst case, such as a damaged box arriving at full speed for the top layer.
Vendors and ecosystem
A palletizing cell is usually assembled by an integrator or system builder from parts made by four groups of vendors: robot makers, vision and sensing companies, gripper makers, and software firms. Specifications below come from vendor pages opened for this post; check current catalogs before buying.
Robot makers
The big four industrial robot makers all sell dedicated palletizing arms:
| Vendor | Palletizing line | Published range |
|---|---|---|
| FANUC | M-410 series | 110 to 800 kg payload, mostly 4-axis |
| ABB | IRB 460, IRB 660, IRB 760 | IRB 660: 180 or 250 kg payload, 3.15 m reach |
| KUKA | KR QUANTEC PA | 140 to 240 kg payload, 3,195 mm reach; food and cold-storage (down to -30 °C) variants |
| Yaskawa | Motoman PL series | Six models, 80 to 800 kg payload |
For lighter loads, collaborative packages pair a cobot with a column lift, pallet stations and pattern software. An example is Robotiq's PE20, built on the Universal Robots UR20 and rated for cases up to 18 kg.
Vision and sensing
| Vendor | Role in palletizing |
|---|---|
| Cognex | Barcode identification on the infeed: DataMan image-based readers and multi-side scan tunnels that read cases moving on a conveyor. Also 2D and 3D machine vision for inspection and guidance. |
| Keyence | 3D vision-guided robotics for locating parts and cases, plus barcode readers and photoelectric sensors. |
| Zebra Technologies (Photoneo) | Structured-light 3D cameras (PhoXi, MotionCam-3D) widely used for depalletizing and mixed-case picking. Zebra completed its purchase of Photoneo in March 2025. |
| SICK | Safety light curtains built for pallet openings (the C4000 Palletizer line), safety laser scanners, and the photo-eyes and distance sensors found along conveyors. |
Gripper and vacuum makers
Vacuum specialists such as Schmalz supply foam-pad area grippers (the FXCB line) that hold a case even when only part of the pad is covered. Integrators often build clamp, fork and bag grippers to suit a specific product.
System builders and software
| Vendor | What they offer |
|---|---|
| Columbia/Okura | Complete robotic and collaborative palletizing systems (including the miniPAL cobot cell) and bagging machines, for food, beverage, building materials, chemicals, animal nutrition and pet care. |
| Mujin | MujinController, a vision and motion-planning platform that runs industrial robots from many makers for single- and mixed-SKU palletizing, depalletizing and truck unloading. |
| Symbotic | Warehouse-scale automation that stores cases in a dense buffer and builds tall, store-ready mixed-SKU pallets sequenced to each store's layout. |
Packaging-machinery groups (for beverage and consumer goods lines) and warehouse automation integrators also build palletizing cells as part of larger end-of-line or distribution projects.
Applications
The same robots, conveyors and grippers turn up in very different settings. What changes is the product and how often the pattern changes.
| Setting | What gets palletized | What makes it hard |
|---|---|---|
| Food and beverage end-of-line | Cases, trays and shrink-wrapped packs of cans and bottles | High speeds, washdown and hygiene rules, wet or glossy packaging that defeats vacuum and vision |
| Consumer goods (CPG) | Uniform cases of household and personal-care products | Many SKUs on one line, frequent pattern changes |
| Bags: feed, cement, resin, chemicals, pet food | 20 to 50 kg sacks | Soft, shifting loads; bag grippers and patterns built for settling; dust |
| Building materials | Pails, bundles, blocks, heavy bags | Very heavy picks, rough handling, outdoor or dusty plants |
| Pharmaceuticals and personal care | Shippers with serialized labels | Traceability: every case must be scanned and recorded against the pallet |
| Cold storage | Frozen food cases | Robots rated for sub-zero work, condensation on cameras and cups |
| Distribution centres (mixed-case) | Store orders built from many SKUs | Build order, stability and crush rules decided on the fly; 3D vision required |
| E-commerce and parcel | Totes, polybags and cartons onto pallets or carts | Huge variety of shapes and labels; often semi-structured |
Depalletizing is the reverse job on the same hardware: a robot unstacks inbound pallets onto a conveyor. It is often harder than palletizing because the robot does not know exactly where each box is. Layers shift, boxes get crushed, and slip sheets have to come off. Depalletizing cells lean heavily on 3D vision and motion-planning software.
Choosing and sizing a cell
Start from the line's peak rate and the worst case it will ever send, then work back to the robot.
Throughput
The robot must keep up with the line at peak, with margin for pallet changes:
picks per minute = cases per minute ÷ cases per pick
A line making 24 cases a minute needs 24 picks a minute with a single-case gripper, which is beyond most single robots. A gripper that takes 2 cases cuts that to 12, well within a typical palletizing robot. Size the cell against its slowest cycle, the far corner of the top layer. Then add the time lost while a full pallet leaves and an empty one arrives, unless the cell has two pallet stations.
Questions to answer before buying
- What is every case size, weight and packaging type the line runs today, and which are planned?
- How many patterns, and how often do they change?
- Peak cases per minute, and how many cases per pick can the product tolerate?
- One pallet station or two? Slip sheets? Pallet dispenser?
- Floor space and ceiling height for the tallest stack plus the arm above it.
- Fenced industrial cell or a collaborative cell with area scanners?
- Does the cell need to talk to MES or WMS, scan every case, or record pallet contents?
- Who will maintain it, and how fast can they get spare cups, sensors and belts?
Cobot or industrial robot
A collaborative package fits slower lines (a few cases a minute), light cases, small floors and sites that change products often. An industrial palletizer fits higher rates, heavy cases or bags, multi-case picks and tall stacks, at the cost of full guarding and more floor space.
Return on investment
The payback usually comes from labour saved across shifts and fewer lifting injuries. Consistent stacks also cut product damage, and the line keeps running when the palletizing crew is short. Packaging that won't grip and patterns that won't hold are the usual reasons a cell misses its projected return.
Sources
- FANUC M-410 palletizing robots
- ABB IRB 660
- KUKA KR QUANTEC PA
- Yaskawa Motoman PL series
- Robotiq PE20 (control.com)
- Cognex modular vision tunnels
- Keyence 3D vision-guided robotics
- Photoneo acquired by Zebra Technologies
- SICK C4000 Palletizer (MHL News)
- Schmalz FXCB area gripping systems
- Columbia/Okura palletizing systems
- Mujin Series C (Robotics 24/7)
- Symbotic FAQ