Table of Contents
Robot Gripper Types: Suction, Jaws and Fingers Compared
The three robot gripper types each fail at different things. Why Boston Dynamics dropped the pinky, and how to test grippers at home with household objects.
Boston Dynamics removed a finger. In an October 1, 2026 IEEE Spectrum piece, the company described its new Atlas hand: three fingers and a thumb, 13 degrees of freedom, up from 7 in the previous generation. The pinky is gone, because testing showed it wasn’t needed for most tasks. That’s the cleanest illustration available of how robot gripper types actually get designed: not by copying a human hand, but by asking which capabilities earn their cost. Alberto Rodriguez, the company’s director of robot behavior, put it in one line: “Hands are a ruthless design trade-off…you’re always giving up on something.”
Key Takeaways
- There are three main robot gripper types in commercial use: vacuum (suction), parallel-jaw (two-finger), and multi-finger dexterous hands. A fourth family, soft and jamming grippers, is growing.
- Suction force is pressure difference times contact area. That single equation explains everything suction can and can’t hold.
- Boston Dynamics’ new Atlas hand has 13 degrees of freedom, dropped the pinky, eliminated delicate tendons and cables, and uses fewer, larger direct-drive actuators in replaceable packs.
- The design driver is manufacturing, not biology. Rodriguez: “What we’re working on right now is figuring out what needs to change…if we want to make 100,000 of these hands a year.”
- Researchers measure grasping with shared physical standards. The Yale-CMU-Berkeley object set (Calli et al., IEEE Robotics & Automation Magazine, 2015) and NIST’s task boards exist so results from different labs can be compared.
Suction: the simplest physics in robotics
A vacuum gripper holds an object by removing air from a sealed volume between a cup and the object’s surface, letting atmospheric pressure outside push the object against the cup.
The whole behavior follows from one relationship: holding force equals the pressure difference multiplied by the sealed area. Atmospheric pressure is about 101 kilopascals, so even a modest vacuum across a 10-square-centimeter cup produces tens of newtons. That’s why a single suction cup can pick up a heavy box.
And it’s why suction fails in exactly predictable ways. Porous surfaces leak, so raw cardboard and mesh bags are hard. Curved or textured surfaces break the seal. Soft bags deform and crease around the cup. Dusty surfaces foul the seal. Shear loads are weak: a suction cup resists pulling straight off far better than it resists sliding sideways, which is why vacuum-equipped robots lift straight up and don’t drag.
Suction dominates parcel logistics for one reason: boxes are flat, rigid and clean. The International Federation of Robotics reported on September 30, 2026 that transport and logistics was the largest professional service robot application in 2025 at 117,000 units. A large share of robotic picking in that sector is pneumatic.
Parallel jaws: friction doing almost all the work
A parallel-jaw gripper holds an object by pressing two opposed surfaces against it, relying on friction, and sometimes on shape, to resist slipping.
Two ideas govern it. Force closure means friction is doing the holding: the grip force times the friction coefficient has to exceed the load. Form closure means the shape is doing the holding: a pin sitting in a V-groove can’t escape regardless of friction. Good gripper design chases form closure wherever possible, because it stops depending on a friction coefficient that changes when the part is oily.
Parallel jaws are the workhorse of industrial robotics, and they’re why the installed base is enormous: the IFR counted 5 million industrial robots in operation as of 2025, up 9 percent, with over 600,000 new installations in the year. Most of them hold things with two jaws or a vacuum cup. A parallel-jaw gripper has one or two actuators, a handful of moving parts, and failure modes you can reason about on a napkin.
What it can’t do is reorient an object once it’s held. If a part arrives upside down, a two-jaw gripper has to put it down, let it settle, and pick it up again. That limitation is the entire argument for fingers.
Fingers: expensive, fragile, and sometimes necessary
A multi-finger hand holds an object with three or more independently controlled contacts, which allows in-hand reorientation, moving the object relative to the palm without putting it down.
That capability is the prize, and it’s costly. More fingers means more actuators, more joints, more wiring, more things that break. The classic approach routed tendons and cables through the fingers, which is elegant and compact and, as Boston Dynamics’ redesign suggests, a reliability problem. Their new hand eliminated the delicate tendons and cables, and instead used fewer, larger, more powerful direct-drive actuators embedded in the joints, with each actuator pack easily replaceable.
Read that as three separate decisions. Direct drive means no transmission to wear out. Larger actuators mean fewer of them. Replaceable packs mean a failure is a ten-minute swap rather than a return to the factory. None of those is about dexterity. All of them are about the hand surviving a year of work.
The capability result is interesting too: the hand does tool use with drills and welding torches, handle manipulation, and finger splaying beyond what a human hand can do. It’s not copying a hand. It’s beating a hand on the axes that matter and losing on the ones that don’t. Rodriguez said the team is “convinced that from a product perspective, this hand should be able to do everything we need it to do.”
How to Teach Your Kid About Robot Gripper Types
This is the rare robotics topic you can fully explore with things already in the house.
Ages 5–8: Suction cup hunt
Take a bathroom suction cup and find five surfaces it sticks to and five it doesn’t. Mirror, window, tile: yes. Wall paint, wood, fabric, paper towel, a crumpled plastic bag: no. Then ask the question that does the teaching: “what do all the good ones have in common?” Smooth, flat, solid. They’ve just derived the suction gripper’s spec sheet.
Ages 9–12: Build three grippers in an hour
Clothespin (parallel jaw), two chopsticks and a rubber band (a crude two-finger), and a suction cup taped to a stick. Then line up ten objects: a pencil, an egg, a tennis ball, a crumpled paper ball, a coin, a folded towel, a plastic cup, a key, a bag of rice, a sheet of paper. Have them score which gripper holds what, in a table. The pattern that emerges, no gripper wins everything, is the whole field in one afternoon.
Ages 13+: Run your own benchmark
This is what professional researchers do. The YCB object set exists so labs can compare grasping results on identical physical objects, and NIST maintains task boards for its Robotic Grasping and Manipulation Competition for the same reason. Have your kid define a benchmark: ten household objects, five attempts each, success defined precisely in advance (held for three seconds, lifted 10 centimeters, no drop). Then test two gripper designs and report success rates. If they build anything with a servo, this becomes a legitimate science-fair project with a real methodology.
The question to ask: “Which objects in this room could a suction gripper never pick up, and why?”
The four gripper families, compared
| Vacuum / suction | Parallel jaw | Multi-finger hand | Soft / jamming | |
|---|---|---|---|---|
| How it holds | Pressure difference across a sealed area | Friction and form closure between two faces | Multiple independent contacts | Conforms around the shape |
| Actuators | One vacuum source, often shared | One or two | Many, Atlas’s hand has 13 DOF | One or two |
| Best at | Flat, rigid, clean, non-porous | Rigid parts of known shape | Tools, handles, in-hand reorientation | Irregular, delicate, unknown shapes |
| Fails at | Porous, curved, soft, dusty, shear loads | Reorienting a held object | Cost, wiring, reliability | Precision placement, high force |
| Relative cost | Low | Low | High | Low to medium |
| Reliability | High, if the surface cooperates | Very high | Historically the weak point | Medium |
| Typical use | Parcel and case handling | Factory assembly and machine tending | Humanoid and research platforms | Food, produce, mixed goods |
| Repair | Replace a rubber cup | Replace a jaw or a cylinder | Replace an actuator pack, by design | Replace a membrane |
The honest summary of that table: suction and jaws do almost all the real work in the world today, and fingers are where the research money goes because fingers are what a general-purpose robot would need. Both statements are true at once.
What to do with this at home and when reading the news
Judge a robot by its gripper, not its face
The end effector tells you what a robot is for. A humanoid with a vacuum cup is a box-moving machine with legs. A humanoid with a 13-degree-of-freedom hand is aiming at tool use. The marketing may say “general purpose”; the gripper says what it can actually attempt.
Count the degrees of freedom
It’s a quick capability estimate. Two DOF is a jaw. Seven was the last Atlas hand. Thirteen is the new one. A human hand has roughly 20 or more depending on how you count. More DOF means more possible in-hand motions and more things to break, and the trend in shipping products is to pick the smallest number that does the job.
Watch for the dropped pinky pattern everywhere
Removing a capability after testing proved it unnecessary is the most underrated engineering move there is, and kids rarely see it modeled. Boston Dynamics tested, found the pinky wasn’t earning its cost, and deleted it. Point that out the next time your kid wants to add a feature to a project instead of finishing it.
Prefer kits with real grippers
A robot arm kit whose gripper is a decorative two-prong plastic claw teaches almost nothing, because it can’t hold anything interesting. One with a servo-driven parallel jaw and some rubber on the pads will actually pick up objects, which means the kid gets to run experiments. Our comparison of robotics kits and what they really teach covers which platforms do this well.
What not to do
Don’t tell a kid that robot hands are “almost as good as human hands.” They aren’t, in a specific and interesting way: robot hands are better at force, precision and splay, and dramatically worse at tactile sensing and at handling objects they’ve never seen. Hands are where robotics is most visibly unfinished, which we also cover in why robot hands are harder than legs.
What to Watch For Over the Next 3 Months
- Week 4: Watch for any manufacturer publishing a grasp success rate on a standard object set. It is the single most informative number in manipulation and almost nobody discloses it.
- Month 2 red flags: Hand demos filmed only with rigid, high-friction objects, blocks, bottles, tools with knurled handles. The hard objects are thin, slippery, deformable and transparent: a plastic bag, a wet glass, a sheet of paper. If those never appear, the demo has told you what the hand can’t do.
- Month 3 self-check: Ask your kid why a robot would delete a finger. If the answer involves reliability or manufacturing rather than “it didn’t work,” they’ve understood the actual logic of product engineering.
Frequently Asked Questions
Why don’t all robots just use suction?
Because the object has to cooperate. Suction needs a surface that is flat, rigid, clean and non-porous, and it resists sliding poorly. A warehouse of identical boxes is a perfect suction environment. A kitchen is not.
Is a five-fingered hand better than a four-fingered one?
Not automatically. Boston Dynamics tested and found the pinky unnecessary for most tasks, so the new Atlas hand has three fingers and a thumb with 13 degrees of freedom. Fewer digits means fewer failure points and lower cost. The right number is whatever the task list requires, which is a question you answer with testing, not intuition.
What’s the hardest object for a robot to pick up?
A thin, transparent, deformable one. A single sheet of plastic film defeats nearly everything: depth cameras can’t see it, suction crumples it, jaws slide off it, and fingers can’t find an edge. Plastic bags are a genuine open problem in logistics automation.
Do robot hands feel what they’re touching?
Some do, poorly. Tactile sensing lags far behind vision in robotics: there’s no equivalent of a cheap, high-resolution, durable skin sensor. This is why robots often grip too hard or too softly, and it’s one of the most active research areas.
Can my kid build a working gripper cheaply?
Yes. A single hobby servo, some craft sticks or 3D-printed links, and rubber bands make a functioning parallel-jaw gripper for very little money. The design lesson arrives immediately: the first version will drop everything, and fixing that means adding friction material and rethinking the jaw geometry.
Why does manufacturability matter so much here?
Because a hand that can’t be mass-produced isn’t a product. Rodriguez framed the current work as figuring out what has to change to make 100,000 hands a year. Many research hands are beautiful one-offs that no factory could build at volume, and that gap is where most robotics startups actually die.
About the author
Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.
Sources
- Ackerman, E. (2026, October 1). “Atlas Robot’s New Hand May Outperform Humanlike Designs.” IEEE Spectrum. https://spectrum.ieee.org/robust-robot-hand
- Calli, B., Walsman, A., Singh, A., Srinivasa, S., Abbeel, P., & Dollar, A. M. (2015). “Benchmarking in Manipulation Research: The YCB Object and Model Set and Benchmarking Protocols.” IEEE Robotics & Automation Magazine, 22, 36–52. https://arxiv.org/abs/1502.03143
- National Institute of Standards and Technology. “Robotic Grasping and Manipulation for Assembly.” Intelligent Systems Division. https://www.nist.gov/el/intelligent-systems-division-73500/robotic-grasping-and-manipulation-assembly
- International Federation of Robotics. (2026, September 30). “Global Sales of Professional Service Robots Surge 24%.” https://ifr.org/ifr-press-releases/news/global-sales-of-professional-service-robots-surge-24-percent
- International Federation of Robotics. (2026, September 24). “Five Million Robots now Operate in Factories Globally.” https://ifr.org/ifr-press-releases/news/five-million-robots-now-operate-in-factories-globally
- Occupational Safety and Health Administration. “Industrial Robot Systems and Industrial Robot System Safety.” OSHA Technical Manual, Section 4, Chapter 4. https://www.osha.gov/otm/section-4-safety-hazards/chapter-4
- Ackerman, E. (2025, September 11). “Reality Is Ruining the Humanoid Robot Hype.” IEEE Spectrum. https://spectrum.ieee.org/humanoid-robot-scaling
