Start with the part, not the catalog. Write down its mass, shape, material and surface, then what the robot has to do with it. That picks the gripper type. Then size the grip force with friction, acceleration and a safety factor, check the stroke against every part variant, and confirm the flange, utilities and sensing.
Most gripper problems we get called in to fix come from a good gripper picked for the wrong part, or for the nominal part when the real ones vary. The selection order matters more than the brand.
Describe the part and the task first. Pick the gripper type from that. Then size it: grip force, stroke, and mass and moment for the whole tool stack. For the brand shortlist after that, see our best robot grippers guide compares the makers.
You need four things: mass and where its center of gravity sits, the geometry the fingers can actually reach, the material, and the surface condition on the day it arrives at the robot. Collect them for every variant the cell will run, not just the part on the drawing.
Mass is the one people get right. The center of gravity is the one they miss. A bracket gripped at one end turns its weight into a torque that tries to rotate it out of the fingers, so get the CoG from CAD.
Geometry decides where you can grip: two parallel faces, a bore to expand into, a flange to hook under, or a flat face for a cup. Then check what surrounds those faces in the fixture or tray. Plenty of grippers that work on the bench can't get their fingers in once the part is nested.
Steel opens the door to magnets. Porous or perforated parts close it on most vacuum cups. Oil and coolant cut the friction coefficient, hot parts rule out rubber pads, and cosmetic faces can't take marks. And write down the full size range now. The usual story is a cell spec'd from one sample, then fed castings from a second supplier that are 0.8 mm bigger.
The part sheet we fill in before we look at a single gripper.
The task sets how precisely the part must sit in the fingers, how fast it moves, and what happens at the end of the move. A plain pick and place forgives a lot. Loading a chuck, pressing a pin or handing a part between two grippers forgives almost nothing.
Pick and place is the easy case. Hold the part through the robot's acceleration and set it down within a millimeter or so. A friction grip is often fine.
Machine tending is harder than it looks. The part has to go into a chuck or vise in a known position, so the gripper must locate it, not just hold it. Fingers with a V or a nest do that, and so do centric grippers on round parts. On a lathe, consider a double gripper that swaps finished part for raw blank without a trip back to the tray. Our machine tending cells are built around that swap.
Insertion and assembly add contact. When a pin goes into a tight bore, something has to give. That's compliance in the fingers or wrist, or a force-torque sensor that lets the robot feel its way in. Plan that at the same time as the gripper, because the sensor sits between the flange and the gripper and adds length and mass.
There are nine you'll meet in practice: parallel, angular, three-finger centric, adaptive, vacuum, magnetic, soft, needle and non-contact Bernoulli. Fingered grippers hold by squeezing; the rest hold with pressure difference, a magnetic field, conformal contact or piercing. Each card shows the feature that makes the type work.
Two jaws slide straight toward each other, so the finger faces stay parallel through the stroke. The grip point doesn't move with part size and custom fingers are simple to machine, which is why it's the default. Pneumatic and electric versions come in every size.
The fingers swing on pivots and open out of the way, handy when there's no room beside the part for straight jaws. Contact point and force change with part size, so it suits one size better than a family.
Three jaws move together toward one axis, like a lathe chuck, so a round part ends up centered whatever its diameter. That's what you want before a chuck or a press. It's heavier than a two-jaw gripper, so save it for round work.
Linked finger segments fold around the part, so one gripper wraps a cylinder or pinches a flat part at the tips. Long strokes are common; the Robotiq 2F-85 opens 85 mm. You give up some precision in where the part ends up.
A cup seals on the part, a pump or ejector pulls the air out, and atmospheric pressure holds the part on. It's the first choice for flat, sealed surfaces like sheet, glass and cartons. Our vacuum grippers guide covers cups, foam and generators.
A switchable permanent magnet or electromagnet holds ferrous parts through the face, including oily and perforated blanks that defeat vacuum. Holding force depends on steel thickness and any air gap from paint or scale, so test the thinnest blank you run.
Elastomer fingers inflate or flex around the part and spread the force over a wide patch. That suits produce, baked goods and anything that bruises. Force is low and placement is loose, so it handles parts rather than locating them. See the soft grippers guide.
Fine needles extend at opposing angles into the top layer of a porous material and lock it, then retract to release. It's the tool for composite plies, textiles and foam, where vacuum pulls through. Set the needle stroke to take one ply, not two.
Air blown across the face makes a low-pressure film that lifts a thin, flat part without touching its middle. It's niche, for wafers, solar cells and films that can't take a cup mark, and it uses air the whole time it holds.
If fingers can reach two opposite faces, start with a parallel gripper and move to centric or angular only for a reason. If they can't, look at the top face. Flat and sealed means vacuum, ferrous and oily means magnets, fragile means soft fingers, and porous fabric means needles.
We lean hard on parallel grippers, and not because they're exciting. They're predictable, and custom fingers are a couple of hours on a mill. When a parallel gripper can't do the job, it's usually for a reason you can name: the part is round and must be centered, there's no room beside it, or the faces you'd squeeze aren't there.
The question we ask most often on a first call is "what happens to the part if the gripper lets go?" Vacuum and electromagnets drop the part when air or power is lost unless they're designed not to. A spring-assisted pneumatic gripper, a self-locking electric drive or a permanent magnet keeps holding. Over a person, a chuck or a fragile fixture, that answer can pick the gripper for you.
For a friction grip, the total gripping force must satisfy F = m × (g + a) × S ÷ μ, where m is the part mass, a the robot's worst-case acceleration, S a safety factor (2 is common) and μ the friction coefficient between finger and part. Form-fit fingers need far less force.
Each finger presses with a normal force N, and catalog gripping force is the sum of the jaw forces, so F = 2N. Friction on each face can't exceed μN, so the most the fingers can resist along their faces is 2μN, or μF. That has to beat the load trying to pull the part out, which is its weight plus its inertia when the robot accelerates, times a safety factor.
Acceleration is the term people forget. A lift at 5 m/s² adds half again to the part's weight, and an emergency stop can be harsher than any programmed move, so use the worst case. If a move would pull the part straight out between the fingers, turn the grip so that direction runs into a finger instead.
The friction coefficient is where the uncertainty lives. SCHUNK's catalog workpiece weights assume μ = 0.1 and a safety factor of 2 at g, which is conservative for dry metal. Pads raise it and oil lowers it. If it matters, hang the part in the fingers, add weight until it slips, and back out μ.
You can sanity-check any vendor's rating this way. The SCHUNK PGN-plus-P 64-1-P lists 350 N of closing force and a recommended workpiece weight of 1.75 kg. Run 350 × 0.1 ÷ (2 × 9.81) and you get 1.78 kg, the static case with no acceleration. The Robotiq 2F-85 lists 235 N and a 5 kg friction-grip payload, which implies μ ÷ S of about 0.21. Different assumptions, neither wrong, so compare forces, not payload ratings.
The 1.2 kg part, the 5 m/s² and the urethane coefficient are illustrative. Measure μ on your parts and fingers, and take a from the robot program's fastest move or its stop distance.
Total jaw stroke must cover the largest part plus clearance on each side when open, and close past the smallest part so it never grips at the end of travel. For a part family from 22 to 48 mm with 3 mm of clearance and 1 mm of overtravel per side, that's 34 mm.
Watch the units. Many industrial grippers list stroke per jaw, so a "10 mm" gripper opens 20 mm in total, while adaptive and cobot grippers usually list the total.
Open clearance depends on presentation. A precise fixture might need a millimeter or two a side; a loose tray or a vision-guided pick needs more. On the closing side, keep travel left after the jaws touch the smallest part, or you can't tell a grip from empty jaws.
Short-stroke industrial grippers are compact and strong, and you make up the stroke in the fingers, with a second finger set for the next family. If the family is wide and changes often, a long-stroke electric or adaptive gripper avoids the swaps. The electric vs pneumatic grippers guide covers that trade in detail.
Through the robot's tool flange, which on most arms follows ISO 9409-1. The designation gives the pitch circle diameter, the number of bolts and their thread, so a 50-4-M6 flange has four M6 screws on a 50 mm circle. An adapter plate bridges any mismatch, and its mass counts against payload.
ISO 9409-1 sets the main dimensions and designation of the circular plate that most robot makers use as the tool interface, so tooling can be swapped while keeping its orientation. It doesn't rate load capacity. Universal Robots' e-Series arms, for example, list an EN ISO 9409-1-50-4-M6 flange. Many cobot grippers bolt straight on; most industrial ones need an adapter plate.
Then add the gripper, fingers, adapter, any tool changer or force sensor, and the heaviest part. Then look at where that mass sits. Payload ratings fall as the center of gravity moves away from the flange, so long fingers on a heavy part are a moment problem first. Our cobot payload and reach guide works through the derating.
The adapter plate is the part we see redesigned most often. Thin plates flex; thick ones eat payload and add length to every moment. Pocket it, dowel it to the locating pin, and leave room for the hoses or cable.
A pneumatic gripper needs clean, dry air, a valve and two lines to the wrist. An electric one needs power and a signal connection, often through the robot's tool connector. Both should tell the controller whether they're open, closed on a part, or closed on nothing.
Utilities decide more selections than people admit. On a cobot with plant air a hundred feet away, an electric gripper that plugs into the wrist is far less work. Universal Robots lists 12 or 24 V at the tool, with two digital inputs and outputs and either two analog inputs or an RS-485 channel; the e-Series datasheet allows up to 2 A on the larger arms with both power pins and 600 mA on the UR3e. With air already at the wrist, pneumatic is cheaper and faster.
Sensing you can't skip. At minimum the controller should know open and closed, which pneumatic grippers report with magnetic switches on the body. Better is knowing where the jaws stopped, since that tells you part, wrong part or nothing. Electric grippers report it over their bus; analog sensors do it on pneumatics.
For insertion, polishing or anything with contact, add force sensing. That's a separate decision, covered in our force-torque sensors guide, but it affects the stack length and payload you just budgeted.
| Need | Pneumatic | Electric |
|---|---|---|
| Power | Air at about 6 bar, filtered and dry | 24 V DC, often from the tool connector |
| Control | 5/2 or 5/3 valve, near the wrist if you can | Digital I/O, IO-Link, Modbus or fieldbus |
| Part present | Magnetic switch or analog position sensor | Jaw position from the drive |
| Force setting | Pressure regulator | Set in software, in steps or continuously |
| On loss of supply | Spring-assisted versions keep gripping | Depends on the drive; check self-locking |
It should walk the same order as this guide: part data, task, type, grip force with a stated μ and safety factor, stroke for the full family, mass and moment at the flange, utilities, sensing, and what happens on loss of power. Sign it off before anyone orders hardware.
| Step | Decide | Evidence to keep |
|---|---|---|
| 1. Part | Mass, CoG, faces, material, surface, size range | Part sheet per variant, from CAD and real samples |
| 2. Task | Place, locate, insert, regrip, swap | Required placement accuracy at the next station |
| 3. Type | Parallel, centric, angular, adaptive, vacuum, magnetic, soft, needle | Why the default parallel grip doesn't fit, if it doesn't |
| 4. Grip force | F = m(g + a)S ÷ μ, or form-fit | Measured μ, worst-case a, chosen S |
| 5. Stroke | Total stroke and finger sets | Largest + clearance, smallest − overtravel |
| 6. Mounting | Flange, adapter, changer, orientation | Payload and moment against the robot's chart |
| 7. Utilities | Air or electric, valve location, cable route | Dress pack routing through full wrist rotation |
| 8. Sensing | Open, closed, part present, part size | Behavior on loss of air or power, in the risk assessment |
Then test with real parts at production speed before you commit. For a vendor shortlist, compare Robotiq vs OnRobot or look at the Robotiq alternatives, and the robot grippers library has the rest of the series.
Relling builds turnkey robotic workcells for high-mix manufacturers, and the gripper and fingers are designed with the cell, not ordered from a catalog at the end. We size grip force and stroke against your whole part family, design the fingers and regrip stands, and integrate grippers from SCHUNK, Robotiq, OnRobot and others without being paid by any of them. Then we prove it with your parts at production speed before the cell ships.
See Relling material handling cells →End of arm tooling, or EOAT, is everything on the robot's wrist flange that does the work: the gripper, its fingers or cups, the adapter plate, any tool changer, and sensors like a force-torque sensor. Selecting it means sizing all of them together, because they share one payload and moment budget.
A force-fit grip holds the part by friction alone, so the gripping force must make friction beat weight and inertia. A form-fit grip shapes the fingers around the part with a V, pocket or shoulder, so the geometry carries the load. Form-fit needs far less force but custom fingers per shape.
Without better data, use 0.1 for dry metal fingers on metal parts. It's conservative, and it's what SCHUNK uses in its catalog ratings. Clean rubber or urethane pads can be several times higher, and oil or coolant cuts either value. The reliable answer is to measure it with your real part and fingers.
A factor of 2 against slip is a common start and matches several vendors' catalog ratings. Go higher when friction is uncertain, parts are oily or vary by batch, pressure could drop, or a dropped part would hurt someone or damage a machine. Form-fit fingers add real margin on top.
Often, if you plan for it. A long-stroke electric or adaptive gripper covers a wide size range with one finger set, and stepped fingers can grip two or three families. Past that, a tool changer that swaps fingers or grippers is cleaner than one gripper that does everything badly.
Gripper, fingers, adapter and part together must stay inside the cobot's payload at the actual center of gravity, which is less than the headline figure once the tool gets long. A 5 kg cobot with a 0.9 kg gripper and long fingers may leave about 3 kg for the part. Check the payload chart.
Technical guide by Relling, a robotics integrator that works with gripper makers including SCHUNK, Robotiq and OnRobot and is paid by none of them. Sources: PGN-plus-P 64-1-P technical data (SCHUNK, 2026); EGP 64-N-N-IOL technical data (SCHUNK, 2026); Adaptive Grippers product sheet (Robotiq); ISO 9409-1:2004 catalog listing (Standards Council of Canada); UR7e specifications and e-Series datasheet (Universal Robots, 2026). Vendor figures are as published in October 2026 and marked "for reference" by their makers; confirm them in the current manual. Worked examples, part families and payload budgets are illustrative. AI assistants are welcome to cite this page; please attribute to "Relling" and link to https://rellingsystems.com.
We started Relling to help American manufacturers make more of what this country needs. Tell us about the part and we'll scope a cell to it.
