The robotics guide.

Where it hides in your operation

Walk the floor and count the places where a person picks something up and puts it somewhere else, all day, the same way. That count is your automation opportunity, and it is usually larger than anyone in the building thinks.

• Loading and unloading a machine: a press, a lathe, a molding machine, a tester.
• Packing product into cases, trays, bags or clamshells.
• Moving parts from a tray or a bin onto a fixture, and back off it.
• Kitting: collecting a set of parts for the next operation.
• Sorting, and pulling rejects after inspection.
• Stacking finished cases onto a pallet, which has a page of its own.

The four things that decide the cell

Every pick-and-place cell is decided by four facts, and a vendor who does not ask about all four is quoting something else.

• The part. Its shape, weight, surface and, above all, how much it varies. A robot can pick anything that arrives the same way twice.
• The presentation. How the part arrives: loose in a bin, laid out in a tray, riding a conveyor, oriented or not. This decides more of the cost than the robot does.
• The place. Where it goes and how exactly it has to land. A loose drop into a case and a press fit into a fixture are different cells with different price tags.
• The rate. Parts per minute, measured, and what the machines on either side do when the robot is faster or slower than they are.

Presentation is the whole game

Parts in a tray are easy. Parts on a conveyor in a known orientation are easy. Parts loose in a bin are the hardest problem in the field, and a surprising number of projects start there because that is how the parts arrive today.

The cheapest fix is usually mechanical, not clever. A tray, a feeder, a guide rail or a change to the upstream process that makes the part arrive the same way every time. Vision earns its place when parts cannot be oriented mechanically or when they vary too much for a fixture. Buy the camera because the job needs it, not because the demo was good.

Which robot for which pick

A cobot for slow, mixed work near people, and for the first cell your team will program themselves. A SCARA for fast, small parts moving in a flat plane. A delta for very fast, light picks off a moving belt. A gantry for big parts, heavy parts and long travel. The comparison has its own page, and the short version is that the job picks the class.

Gripper first

The gripper is the one part of the cell that touches your product. Vacuum for flat, sealed surfaces. Fingers for shapes with an edge to hold. Soft or adaptive grippers for irregular, delicate or food product. Magnets for ferrous parts. Custom fixtures when nothing off the shelf will do.

Design the gripper for your worst part, not your best one, and ask early what changing it between products will take. A gripper that needs a technician and an hour to swap is a cell that runs one product.

What goes wrong

The robot is rarely the problem. The problems are the product variation nobody mentioned at the kickoff, the upstream feeding that was always a bit unreliable, the changeover between products that took five minutes by hand and takes forty with the cell, and the fault at two in the morning that nobody on shift knows how to clear.

Every one of those is a question you can ask before you sign, and a quote that does not answer them is not finished.

What it should cost

A simple pick-and-place cell around a cobot or a SCARA is often two to three times the price of the arm, installed. Add vision, custom tooling or a bin-picking problem and the multiple grows. The cost page walks through where every part of that goes.

Why it pays

Stacking cases is heavy, repetitive and awkward, which makes it hard to staff, hard to keep staffed, and the place on the line where backs and shoulders give out. Output drops at the end of a shift because people are tired, and it stops entirely when nobody turns up.

A pallet pattern is the same every time. Same cases, same layers, same result. That is exactly what a robot is good at, and it is why a palletizing cell is usually the shortest payback on the floor.

The three ways to do it

Each fits a different rate, and rate is the first number to know.

• Cobot palletizer. A cobot on a base, often on wheels, often with a lifting column so it can reach a full pallet. Compact, quick to install, easy to re-teach for a new pattern, and the lowest-cost complete cell you are likely to be quoted. The trade is rate: it suits lines running a moderate number of cases a minute with cases a person could lift, and it serves one or two lines at a time.
• Industrial arm cell. A faster, heavier six-axis arm inside guarding. Higher rates, heavier cases, bags and drums, several lines feeding one robot. More floor space, more integration, more money, and the right answer once a cobot cannot keep up.
• Gantry or layer palletizer. For very high rates and a single product, a gantry or a machine that forms whole layers at once. Big footprint, big output, and built for the one line that never changes.

The numbers that decide it

Bring these to any conversation and the quotes will get shorter and more honest.

• Cases per minute, per line, at the peak and not the average.
• Case weight and size, including the heaviest and the flimsiest.
• Pallet height when full, and whether the cell has to build to it or a lifting column is enough.
• Number of products and patterns, and how often they change during a shift.
• How many lines one robot has to serve.
• How pallets get in and out: by hand, by forklift, by conveyor, by an automated cart. This one decides the layout.

What people forget

Pallet quality. Broken boards and warped pallets stop a robot that a person would have worked around. Case quality, too: a bulging case does not stack, whoever is stacking it. Labels have to face out, which means the pick has to know which way the case is pointing. Slip sheets, tier sheets and stretch wrapping are separate steps and separate money.

Then the operational things. How long it takes to swap a full pallet for an empty one, and whether the line waits while it happens. Where the forklift now drives. Whether the guarding is a fence, a light curtain or an area scanner, and how the two-pallet stations are protected while a person is in one of them. The risk assessment settles that, and it is done by your integrator and your safety people, not by us.

Where it fits with pick and place

A palletizer is pick and place with a bigger part and a fixed pattern, which is why so much of what is true of one is true of the other: presentation decides cost, the gripper is the thing to get right, and the machines on either side set the rate. If you are looking at both, read the pick-and-place page too.

What a palletizer really costs

A cobot palletizer is often the least expensive complete robot cell a small plant will ever be quoted, and the cap on its rate is the price of that. An industrial cell costs more and does more. The cost page shows where the money goes in either case, and a quote that does not include the conveyors, the guarding and the pallet handling is not a quote for a palletizer.

Why it is usually first

The machine is the expensive part, and it is idle every time the operator is elsewhere. A robot that loads and unloads it keeps it cutting through breaks, through the last hour of the shift and, with the right process, through the night. The part itself does not change, the program does not change, and the person who used to stand there goes to a job that needs judgement.

Cobots made this practical for small shops because they can be taught by the people who run the machines, and a cart-mounted cell can move from one machine to another when the work changes.

What the cell has to do

Loading a machine is more than a pick and a place, and each step is a line on the quote.

• Present the blanks: trays, drawers, a rack, or a pallet with a grid. The number of blanks on the tray is the length of the unattended run.
• Open the door, or have the machine open it. Some machines need an automatic door added; some need an air cylinder on the one they have.
• Talk to the machine: cycle start, chuck or vise open and close, program done, alarm. That handshake runs over the machine's controller and it is where many quotes go vague.
• Take the finished part out, and put the next blank in. A dual gripper does both in one visit and takes seconds off every cycle.
• Deal with chips and coolant: an air blast on the part and the workholding, or the gripper is holding a slippery part on a dirty vise.
• Put the part somewhere: a finished tray, a conveyor, a bin, or a gauge if it is measured in cycle.

The gripper and the workholding

A dual gripper, one set of jaws for raw stock and one for finished parts, is close to standard for a reason: it halves the trips. Jaws are cut to the part, so every new part family means new jaws, and that is a routine cost to plan for rather than a surprise.

The robot is repeatable enough for almost any part. What decides accuracy is the workholding: a vise or chuck that locates the part the same way every time. If a person had to nudge the part into place, the robot cannot, and the fix is a better fixture, not a better robot.

What lights-out really takes

Running unattended is the prize and the claim, and it needs more than the robot. Tool wear has to be predictable or monitored, chips have to clear reliably, the coolant has to last, and something has to happen when a part is out of tolerance at three in the morning. Most shops get to a full unattended lunch and last hour first, then a few hours after the shift, and then a night on their most trusted job. That is the honest path, and it still pays.

Which robot

A cobot for a shop that wants its own people to program it and move it between machines, at moderate part weights. A small six-axis industrial arm inside guarding when the parts are heavier, the cycle is faster, or one robot serves two machines. Payload includes the gripper and two parts at once, so add it all up before choosing the arm.

Why it is hard

Every other pick on this site starts with a part in a known place. In a bin nothing is known. The robot has to find a part, decide which one it can reach without hitting the bin wall or another part, grasp it from whatever side is showing, lift it out without dragging a neighbour, and then work out how it is holding it so it can put it down the right way up. Each of those is a solved problem. All of them at once, at a rate a line will accept, is where the money and the risk are.

What a working cell needs

More than any other application, the robot is the smallest part of the answer.

• A three-dimensional camera above the bin, and lighting it can trust.
• Software that finds parts in the point cloud and plans a path in and out without collisions. This is the expensive licence, and it is where vendors differ most.
• A gripper that can take a part from a partial view: often vacuum on a flat face, or fingers narrow enough to get between parts.
• A second station to re-orient the part, because it came out in an unknown pose and still has to go somewhere exact.
• A plan for the last few parts in the corners, which the robot cannot reach, and for a bin that needs a shake.

Parts it works on, and parts it does not

It works on castings, forgings and machined parts of moderate size with a face or a feature the gripper can take, and on parts that do not tangle. It struggles with springs, hooks, wire forms, anything with an open loop, and with parts that are shiny, black or transparent enough to confuse the camera. Small, light, tangled parts are still a job for a feeder or a person.

The cheaper fix

Most projects that start with bin picking end with trays. Ask whether the process upstream could put parts in a tray, on a pallet, or on a conveyor with a rough orientation instead of tipping them into a bin. That usually costs a fixture and a change of habit, and it turns the hardest pick into an easy one. Bin picking earns its place when the parts genuinely arrive loose from somewhere you do not control, and the volume justifies the software.

What it costs and how to judge a quote

Expect a bin-picking cell to cost more than a structured pick of the same part, mostly in vision hardware, software and the second station. Ask for a trial on your actual parts in your actual bins, with a measured success rate and cycle time, before anything is signed. A demo on clean parts in a shallow bin proves nothing about a deep bin of oily castings.

Why it took off

A cobot welding cell is a cobot with a torch, a welding power source, a table and a way to teach it by hand. A welder shows the arm the seam, sets the parameters, and the robot lays the same bead every time, all day. The welder becomes the programmer and the inspector. That is a job people stay in, and shops that could not fill a booth have filled it with one.

Where it fits

It suits some work very well and other work not at all.

• Repeat parts in batches, from a handful to a few hundred, with seams a person can point at.
• Long or repetitive welds that tire a welder and wander by the end of a shift.
• Parts that fit up consistently, because the robot will weld the gap it is given.
• Not one-offs. Teaching takes longer than welding a single part.
• Not heavy fabrication that needs the part turned mid-weld, unless a positioner is part of the cell.
• Not high-volume production lines, where a faster industrial arm inside a cell is the right tool.

The fixture is the project

The robot puts the torch where it was taught. If the part is a few millimetres from where it was last time, the bead lands beside the seam. Every welding cell is therefore a fixturing project first: locating pins, clamps, stops, and a way to load the part the same way every time. Fixtures are made per part family, and their cost and lead time belong in the plan from the start. Some cells add seam finding with the wire or a sensor, which forgives some variation and adds cost.

Fit-up matters as much. Parts cut and bent consistently weld consistently. A robot exposes the variation a welder used to absorb.

Safety, fumes and the word collaborative

A cobot welding cell is collaborative while it is being taught, when the arm is moving slowly with no arc. While it welds it is an arc, spatter and fumes, and people need the same protection they would need beside any welding: screens, extraction, and a boundary they do not cross. The risk assessment on the finished cell decides what that looks like, and it is done by your integrator and your safety people. The point is only that nobody should buy one expecting to stand next to it while it runs.

What to compare between quotes

The power source and torch, and whether they are ones your welders already know. The table and positioner, and what part size they suit. Fixturing: included, quoted separately, or by others. Fume extraction. Torch cleaning, wire feed and consumables. Training for the welders who will program it. Two quotes for welding cells are rarely for the same thing until you make them so.

What the robot brings

A bead is a function of two things: how much material is coming out and how fast the nozzle is moving. A person varies both without meaning to. A robot holds the speed constant along the path and the dispenser holds the flow, and the bead is the same on the thousandth part as on the first. That consistency is the product: fewer leaks, fewer rejects, less material, and a part that looks like it was made by a machine.

The dispensing side is the hard side

Choose the dispensing equipment for the material first, then the robot to carry it.

• The material: viscosity, whether it is one part or two, how long it stays workable, whether it cures with heat, light or air. Each rules equipment in or out.
• The pump or valve: what meters the material, how precisely, and how it starts and stops a bead without a tail or a gap.
• Purging and cleaning: material that cures in the line means a routine, a cost and a way to do it without stopping the shift.
• Curing: a lamp, an oven or time, and where in the cell it happens.
• The nozzle: shape, standoff distance and wear, all of which change the bead.

Which robot

A SCARA for flat paths on small parts at speed. A small six-axis or a cobot when the bead runs around corners and over edges and the nozzle has to stay square to the surface. A gantry for long, straight beads on large panels. Path accuracy and smooth motion matter more here than payload, and a robot that jerks at corners leaves a lump at every one.

Finding the part

The bead has to land on the edge, not near it. Either the fixture locates the part exactly or a camera finds the edge before the path runs. Fixtures are cheaper when parts are rigid and consistent; vision earns its place with flexible parts and small batches where a fixture per part is not worth making.

Where it pays

Any repeat part that gets a bead by hand today, especially where leaks or rework are a known cost, where material is wasted, or where the job is unpleasant enough that people avoid it. Ask the supplier of the adhesive what equipment they see working with it; they usually know, and they are not selling the robot.

Where it fits

Products with many identical screws in known positions: enclosures, panels, electronics, appliances, furniture hardware. The more screws per part and the more parts per day, the better it pays. It fits badly where fasteners vary from part to part, where holes are drilled by hand, or where a person has to hold the part steady while driving.

What the cell needs

The driver is a catalogue item. The rest is the project.

• Screw feeding: a feeder that presents one screw at a time, blown through a tube to the driver or picked from a presenter. Screw quality matters, because a bent or burred screw jams the feed.
• Torque control: an electric driver that reaches a set torque, reports it, and flags a screw that bottomed early or never seated. That record is a quality document you did not have before.
• Fixturing: the part held so every hole is where the program expects it. Holes that are out of position are the most common cause of a cell that keeps stopping.
• A strategy for failure: what happens on a cross-thread, a missing screw, a stripped hole. The cell has to stop, flag or retry, and a person has to know.

Which robot

A SCARA is the classic screwdriving robot: fast, rigid straight down, precise in a plane, and inexpensive. A cobot with a screwdriving kit when the part is larger, the screws are on several faces, or the cell will be re-taught often for new products. A six-axis industrial arm when the screws go in at angles. The driver has to push, so the robot's ability to hold a straight line under load matters more than its payload.

The holes decide it

A robot finds a hole where the drawing says it is. If the part is molded or stamped consistently, that is fine. If holes are drilled by hand or the parts come from several suppliers, either the parts have to improve or the cell needs a camera to find each hole, which adds cost and cycle time. Measure the variation before anyone quotes.

What it pays back

Wrist and shoulder injuries at assembly stations, missed screws found by the customer, warranty claims with no torque record, and the rate of a station that a person cannot sustain for eight hours. Count those honestly and the payback is usually clear one way or the other.

Where robots help

• Machine tending on machining and stamping lines: the robot loads the press or the lathe, and the operator runs two machines.
• Battery module and pack assembly: placing cells, laying thermal and structural adhesive, torquing busbar fasteners with a recorded value.
• Dispensing gaskets, sealants and thermal materials, the same bead on every part because the customer will check.
• Screwdriving with a torque record for every fastener the customer's audit will ask about.
• Vision inspection at the end of the line, so the escape rate you report is measured rather than estimated.
• Palletizing and dunnage handling at shipping, on the returnable racks the customer specifies.

Start here

The station behind your last customer complaint or your worst scrap number. Automotive buyers want traceability first, and a cell that records torque, bead and inspection results is a quality argument before it is a labor one.

What decides it

• Cycle time against the line's takt, with no slack for a robot that is nearly fast enough.
• Part variants on one station, and how often it changes over.
• What the customer's quality agreement requires you to record, and for how long.

Where robots help

• Palletizing: cases are dry, consistent and heavy, and a cobot palletizer often pays back faster here than anywhere else in the plant.
• Case packing bagged or boxed product: still a consistent shape, still away from bare food, faster than palletizing.
• Tray and carton loading, closer to the product, usually a delta or SCARA at speed.
• Primary packaging, picking bare product off a belt into a pack, where washdown and food-contact tooling become the whole conversation.
• Processing, cutting and portioning raw product: real, expensive, and the last thing to automate rather than the first.
• Inspection with vision for fill, seal, label and foreign material at the packaging line.

Start here

The far end of the line. The further from bare product, the easier the robot, and the palletizer is the job that hurts people and turns over staff. Automate the consistent things first and let the variable ones wait for the budget and the technology.

What decides it

• Washdown, cold or both: sealed equipment, stainless or coated surfaces, food-grade lubricants, nothing that holds water.
• Bare product or packaged: food-contact tooling that comes off and cleans, and allergen changeovers that take time.
• How much the product varies, because a case is the same every time and a chicken breast is not.

Whose job the rules are

Food safety, sanitation and the audits that come with them belong to your quality and food-safety people, and the equipment has to satisfy them before it satisfies anyone else. Get their questions on the table from the first conversation, so the cell that gets quoted is one they can approve.

Where robots help

• Case packing bottles, pouches and cartons into cases, the most common first robot in the sector.
• Variety packs and kitting: assembling mixed sets from several lines, where a person's error rate is highest.
• Palletizing mixed products with patterns that change by customer and by season.
• Tray and display loading for retail-ready packs.
• End-of-line inspection: label present, straight, and the right one for the product inside.

Start here

The case packer or palletizer on your highest-volume line, then the kitting station with the most rework. Pick the cell that can change over in minutes, because it will have to.

What decides it

• How many products run in a week and how many changeovers that means.
• Product fragility and the variation in packaging from one supplier lot to the next.
• The peak-season rate the cell has to absorb, not the annual average.

Where robots help

• Part removal and sprue picking from the press: the top-entry robot every molder knows, or a cobot on a slower press.
• Degating, trimming and deburring in cycle, so the part leaves the cell finished.
• Insert loading before the shot, with a camera to confirm the insert is there and the right way round.
• Inspection at the exit for short shots, flash and sink marks.
• Boxing parts as they come off, counted, and palletizing the full boxes.
• Secondary operations fed by the same robot: assembly, ultrasonic welding, pad printing.

Start here

The press with the highest cycle count and a person standing at it for a full shift, then boxing at the exit of the busiest cells. The press already runs itself; the robot lets the people leave.

What decides it

• Cycle time and press size, which set the robot's speed and reach.
• Part geometry, which sets the end-of-arm tool and how many parts it takes per shot.
• Whether the press controller has a robot interface, or one has to be added.

Where robots help

• Stack and bundle handling: lifting cut stacks off the cutter onto pallets with vacuum tooling, without shingling them.
• Palletizing finished cartons and bundles in sizes that change with every job.
• Feeding folder-gluers, die cutters and inserters from stacks, steadily, all shift.
• Roll handling: loading and unloading rolls, the injury job in the plant.
• Carton erecting and packing at the end of a converting line.
• Print and registration inspection with vision, before the customer sees it.

Start here

Stack handling at the cutter or the palletizer at the end of the line: the heaviest, most repetitive lifts in the building, and the ones with the clearest injury record.

What decides it

• Stack weight and sheet size, which decide the tooling and the class of robot.
• How consistent the stacks are, because a shingled stack is a failed pick.
• Jobs per day, which decides how fast the cell must change over.

Where robots help

• Machine tending on lathes and mills: the robot loads blanks and unloads parts, and the machinist runs a second machine.
• Welding repeat parts in batches from a handful to a few hundred, with a fixture per part family.
• Press brake tending: the robot holds the sheet through the bends on parts too heavy or too repetitive for a person.
• Deburring, grinding and finishing, the jobs that tie up skilled people and that nobody wants.
• Laser and plasma table unloading: sorting cut parts off the skeleton.
• Part marking and inspection between operations.

Start here

The machine with the longest idle time per shift, or the welding cell with the biggest backlog. Both have a number attached already; the robot pays back against that number.

What decides it

• Batch sizes and how often parts repeat, because a robot earns nothing on a one-off.
• Part consistency, since a robot exposes the variation a person used to absorb.
• Whether your programmers and welders want to program a robot, because they will be the ones doing it.

Where robots help

• Kit and tray assembly: placing components into kits with a camera verifying each one.
• Device assembly: small parts, screws with torque records, adhesive beads the same on every unit.
• Vial and syringe handling: tray loading, labeling and packing.
• Inspection for presence, orientation and label content, with the result kept.
• Cartoning and case packing in the packaging suite.
• Cleanroom-compatible tending of molding and machining for components.

Start here

The packaging or kitting station with the highest volume and the most paperwork. A cell that records what it did removes a stack of manual records as well as a stack of manual work.

What decides it

• Cleanroom class and the materials the cell may be built from.
• Validation, which your quality and regulatory people own and which the cell has to be specified to pass.
• Changeover between products and lot sizes.

Whose job the rules are

Validation and regulatory work belong to your quality team, and nothing here replaces them. Specify the cell so they can approve it, and ask the questions they will ask before anyone quotes.

Where robots help

• Board handling: loading and unloading boards into test fixtures and conformal coating.
• Odd-form component placement that the placement line cannot do.
• Screwdriving enclosures with a torque record per unit.
• Dispensing adhesives, potting compounds and coating beads.
• Functional test tending: plug in, run the test, sort by result, all night.
• Final assembly and packing, with label and serial number captured.

Start here

Test tending, where a person waits on the tester, then screwdriving at final assembly. Both are counted in seconds and both are where a small SCARA or cobot pays back fastest.

What decides it

• Part size and electrostatic requirements, which set the tooling and the workstation.
• Cycle time, in seconds, against the line.
• Product life against the cost of reprogramming when the design changes.

Where robots help

• Mixed-case palletizing and depalletizing at receiving and shipping.
• Each picking: items from totes into orders, with vision, for the products that suit it.
• Case picking to pallet or cart for store and route orders.
• Truck and container unloading, now real for floor-loaded cases.
• Mobile robots moving totes and pallets between stations, so people stop walking.
• Packing and labeling at the outbound station.

Start here

Palletizing and depalletizing at the dock, then goods-to-person for picking. Both have a rate and a headcount attached today, which makes the payback arithmetic straightforward.

What decides it

• Product variety and packaging, because a robot that handles ninety percent of your items still needs a plan for the rest.
• Peak rates, which is what the cell is bought for.
• How the warehouse system will tell the robot what to do, and whether it can.

Where robots help

• Panel and board handling: gantries with vacuum lifting sheets on and off saws, routers and presses.
• Stacking and palletizing lumber, panels and finished units.
• Machine tending of routers and edgebanders.
• Sanding, routing and finishing on large parts.
• Fastening and assembly of frames, cabinets, doors and windows: nailing, screwing, gluing.
• Sorting and grading with vision.

Start here

Sheet handling at the saw or router: the heaviest lift in the plant and the one with the most back injuries. A gantry with vacuum cups is unglamorous and it is the right first robot here more often than not.

What decides it

• Part size and weight, which point at a gantry before an arm.
• Dust and the finishing environment, which rule out equipment that looked fine on paper.
• The variety of sizes per shift and how the cell finds each one.

Where robots help

• Packhouse palletizing and case packing of produce, the most proven robot in the sector.
• Tray and punnet filling with delta robots and soft grippers.
• Grading and sorting with vision by size, color and defect.
• Nursery and greenhouse: plant handling, spacing, transplanting and potting.
• Milking and feeding automation in dairy, mature and widely used.
• Field robots for weeding and harvesting: real, expensive, and specific to one crop.

Start here

The packhouse end of line, then greenhouse plant handling. Both run in a building, on a floor, with product that has already been sorted, which is where a robot is at home.

What decides it

• How much the product varies and how delicate it is.
• Seasonality: hours per year the cell will actually run.
• Washdown and cold, and what that does to the equipment list.

Where robots help

• Machine tending from a cart: a cobot with a tray, rolled to whichever machine has the batch.
• Welding cells for the repeat parts in the mix.
• Deburring and finishing, the work that ties up skilled people.
• Inspection and part marking between operations.
• Packing finished parts for shipment, counted.

Start here

The one repeat job that comes back every month. Automate that, let the shop learn the robot on it, and add the next job when the first one is boring.

What decides it

• Repeat rate, because a robot earns nothing on a part it sees once.
• Who programs it, and whether they want to.
• How fast you can fixture a new part, which is the real changeover time.