End-effector selection
Robot Grippers: Types, Selection Guide and Vendors
A robot gripper is the end effector that holds, releases or repositions an object at the end of a robot arm. Robot gripper selection starts with the object, contact geometry and load case, then the actuator and interface. The tables below separate gripper types from model-specific specifications so a payload rating is never mistaken for a measured grasp success rate.
Last updated:
Part of tactile sensing: compare sensing principles before selecting a model or end effector.
Published | Updated by RoboSkin.ai Editorial Team
November 4, 2026: reviewed the full guide and four manufacturer model sources; added contact-evidence distinctions and a receiving-test record linked to six source-reviewed studies. No hardware tests are claimed.
- 3
- sections
- 2
- questions
- 3
- next routes
Robot gripper models and selection fields
Four sourced models, checked 2026-10-30. Prices and protocols are marked Not published when the cited page does not establish them. Payload limits use different grasp conditions and are not a common performance ranking.
| Model | Type | Payload | Grip force | Interface | Contact points / sensing | Price | Source |
|---|---|---|---|---|---|---|---|
| Robotiq 2F-85 | Adaptive fingers | 5 kg form-fit; verify the application load case | 20–235 N | Modbus RTU (RS-485) | Two articulated fingers; optional tactile fingertips require compatible revision | Not published | Robotiq 2F-85 specifications |
| OnRobot RG2-FT | Adaptive fingers | 2 kg force-fit; 4 kg form-fit | 3–40 N | Integrated robot software; protocol not published on the reviewed page | Two fingers; each has 6-axis force/torque and proximity sensing | Not published | OnRobot RG2-FT specifications |
| OnRobot VG10 | Vacuum | 10 kg rated; 15 kg maximum under manufacturer conditions | Not published as grip force; 80% maximum vacuum | Integrated robot software; protocol not published on the reviewed page | Configurable suction cups on four arms; two independent vacuum channels | Not published | OnRobot VG10 specifications |
| OnRobot Soft Gripper | Soft fingers | Up to 2.2 kg; depends on shape, softness and friction | Not published | Electric actuation; protocol not published on the reviewed page | Interchangeable silicone tools in star or finger configurations | Not published | OnRobot Soft Gripper specifications |
Filter by gripper type
Open a type to view only matching rows. The complete comparison remains available above.
Adaptive fingers (2 models)
| Model | Type | Payload | Grip force | Interface | Contact points / sensing | Price | Source |
|---|---|---|---|---|---|---|---|
| Robotiq 2F-85 | Adaptive fingers | 5 kg form-fit; verify the application load case | 20–235 N | Modbus RTU (RS-485) | Two articulated fingers; optional tactile fingertips require compatible revision | Not published | Robotiq 2F-85 specifications |
| OnRobot RG2-FT | Adaptive fingers | 2 kg force-fit; 4 kg form-fit | 3–40 N | Integrated robot software; protocol not published on the reviewed page | Two fingers; each has 6-axis force/torque and proximity sensing | Not published | OnRobot RG2-FT specifications |
Vacuum (1 models)
| Model | Type | Payload | Grip force | Interface | Contact points / sensing | Price | Source |
|---|---|---|---|---|---|---|---|
| OnRobot VG10 | Vacuum | 10 kg rated; 15 kg maximum under manufacturer conditions | Not published as grip force; 80% maximum vacuum | Integrated robot software; protocol not published on the reviewed page | Configurable suction cups on four arms; two independent vacuum channels | Not published | OnRobot VG10 specifications |
Soft fingers (1 models)
| Model | Type | Payload | Grip force | Interface | Contact points / sensing | Price | Source |
|---|---|---|---|---|---|---|---|
| OnRobot Soft Gripper | Soft fingers | Up to 2.2 kg; depends on shape, softness and friction | Not published | Electric actuation; protocol not published on the reviewed page | Interchangeable silicone tools in star or finger configurations | Not published | OnRobot Soft Gripper specifications |
Short answer
What you need to know
- 1
Choose the contact method from object geometry, surface leakage, fragility and the required placement accuracy.
- 2
Check the exact payload conditions, available grip force, tool mass and controller interface together. A form-fit payload is not a friction-grip guarantee.
- 3
Tactile sensor data can support contact detection or slip feedback, but the controller must actually consume it at the required latency.
Topic 01
Robot gripper types and contact tradeoffs
The type comparison is an engineering selection framework. Numerical limits belong to individual products in the model table, not to the whole category.
| Type | Contact method | Useful object conditions | Main constraint | Tactile integration |
|---|---|---|---|---|
| Parallel or adaptive fingers | Two opposing contact regions; adaptive links can wrap | Rigid parts and repeatable insertion geometry | Reach, friction, finger stiffness and jaw travel | Fingertip sensors or instrumented fingers |
| Vacuum | Pressure difference across a sealed cup | Flat or sealable surfaces with accessible pickup area | Leakage, porosity, cup placement and loss of vacuum | Vacuum monitoring is not a spatial tactile map |
| Soft fingers | Compliant fingers conform around the object | Fragile or variable shapes | Deformation changes pose and force transmission | Embedded or surface sensors require calibration under bending |
| Magnetic | Magnetic attraction at the contact surface | Compatible ferromagnetic parts | Material compatibility, residual magnetism and release | Separate contact or force feedback may be needed |
| Dexterous hand | Multiple articulated finger contacts | Reorientation and in-hand manipulation | Control, calibration and contact planning complexity | Distributed fingertips or skin; coverage is model-specific |
Topic 02
Selection sequence
Record object mass, dimensions, surface condition, centre of mass and the planned accelerations. Include gripper mass in the robot wrist load and moment budget. Check the manufacturer load case instead of translating a force rating directly into a payload.
Select jaw stroke or suction coverage, then confirm electrical supply, controller integration and mounting. Record force and slip observations with timestamps during a representative grasp test. Check release behaviour and dropped-part handling as well as successful pickup.
| Decision | Record before ordering | Reject the candidate when |
|---|---|---|
| Payload | Mass, centre of mass, acceleration, orientation and tool mass | The documented load case does not cover the application |
| Grip force | Adjustable range, contact area and permitted object deformation | Minimum force already damages the object or maximum force cannot retain it |
| Interface | Protocol, controller support, update rate and power budget | Required feedback cannot reach the controller in time |
| Contact points | Finger or cup geometry and accessible contact area | The object cannot be constrained without collision or leakage |
| Price | Dated quote including tooling, controller, sensors and replacement pads | The integration cost exceeds the project budget |
Topic 03
Separate contact mechanics, sensing and demonstrated control
A compliant surface can improve contact without measuring it. A sensor can measure a useful signal without feeding the controller. A model can predict motion accurately without having executed a grasp. Identify which claim a supplier or research prototype actually supports before comparing success percentages or choosing an integration path.
| Evidence layer | Illustrative research case | What remains a separate decision |
|---|---|---|
| Passive adaptation | Granular jamming changes enclosure and retention mechanically. | Whether the actual reaching distribution lands inside a useful capture region, and whether transfer and release succeed. |
| Contact measurement | A Hall-based tackiness sensor distinguishes compression from pull-off; TacGooseBumps encodes shear-dependent patterns. | Calibration, replacement effects and whether the required output is a pattern, stress, force or separation event. |
| Active perception | FibTac observes contact while a pneumatic sequence moves its fibers. | Whether classification or regression has been connected to a tested feedback controller; timed actuation alone does not establish this. |
| Motion prediction | JFM models a tendon-driven finger; a membrane array uses a compact deformation action space. | Prediction or positioning criteria are different from measured contact, damage-free handling and completed manipulation. |
Define one receiving test from pickup through release
For example, a proposed test of a fragile part should identify the object, pad revision, surface condition, orientation, motion and permitted deformation before any trials. Record first-attempt pickup, retention through acceleration, final placement, complete release, visible damage and interventions separately. Repeated use of one object should retain its identity so accumulated damage is not mistaken for independent fresh-object trials.
Attach acquisition timestamps, controller input and executed command to the same trial when evaluating feedback. Define the physical reference used for force or slip claims, the pass criterion and the failed-attempt denominator. After changing a pad, membrane, suction cup or sensor, repeat the relevant calibration and load case. This is suggested engineering acceptance documentation, not a claim that the research cases below or RoboSkin completed that combined test.
- Tackiness: pull-off measurement and release
- TacGooseBumps: retrofit and damage criteria
- Granular gripper: reaching versus capture
- FibTac: timed actuation and classification
- JFM: recorded prediction versus executed control
- Membrane array: a different manipulation surface
- Plan a tactile calibration record
- Scope a research and integration evidence review
Common questions
FAQ for this topic
Does a soft gripper automatically measure touch?
No. Mechanical compliance and tactile measurement are different properties. Check whether sensors are installed, which signals are exposed and how the controller uses them.
Why are prices marked Not published?
The reviewed manufacturer pages request a quote or do not disclose a usable model price. A reseller listing or an older quote is not silently substituted.