End-of-arm tooling, vision and force control
Many projects fail not because the robot was wrong, but because the tooling and sensing were. Whether the cell “works” for the customer depends on reliable gripping, accurate positioning and controlled contact force. Treat workpiece → gripping → localisation as one chain.
- Recognise 4 common EOAT types
- Explain 2D vs. 3D vision
- List the must-ask subsystem questions
End-of-arm tooling (EOAT)
The EOAT, also called the end effector, mounts on the flange and acts as the process “hand”.
| EOAT type | Good for | Not for / risks |
|---|---|---|
| Pneumatic gripper | High force, low cost, fast cycles | Needs site compressed air; force and position hard to fine-tune |
| Electric gripper | Programmable force and position; many part sizes and fragile parts | Higher price; usually less force than a pneumatic gripper of the same size |
| Vacuum cup | Cartons, sheet metal, glass, smooth plastics | Porous, oily, soft or thin surfaces leak; check film wrap and deformation |
| Magnetic gripper | Ferrous steel sheet | Useless on non-ferrous materials; watch residual magnetism and double-sheet picks |
| Soft / adaptive gripper | Irregular, delicate food and produce | Low payload, weak at precise positioning |
▶Watch
Worked example: how much force do cups and grippers need?
The two most common EOAT calculations are both simple in principle.
Vacuum cup holding force comes from the pressure difference:
Theoretical force F (N) = vacuum level ΔP (Pa) × effective cup area A (m²)
A 50 mm cup has an area of about 0.00196 m². At −60 kPa it gives a theoretical force of about 118 N, equal to the weight of about 12 kg. The load it can actually handle is lower, because you divide by a safety factor: at least 2 for a vertical lift, and 4 or more for horizontal moves, high speeds or side loads on the cup. So this cup suits loads of roughly 3–6 kg. If the carton surface is porous, film-wrapped or uneven, the vacuum level drops and capacity falls further.
Gripper force relies on friction. For a two-finger gripper the required grip force is roughly:
Required grip force ≈ mass × gravity × safety factor ÷ (friction coefficient × number of fingers)
To hold a 2 kg steel part with a steel-on-steel friction coefficient of 0.2, two fingers and a safety factor of 3, you need about 2 × 9.8 × 3 ÷ (0.2 × 2) ≈ 147 N. With custom fingers shaped to the part, which cradle it instead of relying on friction, the required force drops sharply. That is why engineers so often ask, “Can we see the part drawing?”
Machine vision and force control
Two-dimensional
Planar locating, code reading, presence checks, dimensional checks. Parts need to lie roughly in one plane.
Three-dimensional
Structured light, time-of-flight or stereo; sees height and orientation, for uneven surfaces and bin picking of random parts.
Force/torque sensing
A six-axis force/torque sensor measures contact force, for precise insertion, surface-following grinding and polishing, and constant-force assembly.
Vision succeeds or fails on lighting, calibration (aligning camera and real-world coordinates), occlusion and cycle time; real site lighting and sample parts matter more than demo videos. Separate two uses of vision: guidance (telling the robot where to pick) and inspection (judging quality). Force control has an operating window; it cannot make up for inconsistent incoming parts.
▶Watch: vision and force control
Subsystem question table
If a quote lists only the robot unit price and leaves out the special gripper, force-control flange or 3D camera, the customer will see the solution as incomplete when it comes to delivery. At a minimum, ask the questions below and pass the unknowns to application engineering.
| Subsystem | Must ask | Common mistake |
|---|---|---|
| Gripper / vacuum | Material, weight, size, surface; are marks allowed? Are parts presented randomly or in fixed positions? | Sizing on weight only, ignoring CoG, surface and cycle time |
| Torch / spindle | Process parameters, consumables, cables and dress pack, interference | Ignoring cable life and collisions |
| Force control | Assembly tolerances, contact-force range, speed | Treating it as a universal fix |
| 2D / 3D vision | Guidance or inspection? Cycle time? Site lighting? | Underestimating site lighting and calibration work |
| Safety sensing | Human–robot distance, required safety level (PL / SIL) | Mixing up collaborative and fenced operation |
A three-step questioning method
- Profile the workpiece
- Material, weight, size, surface roughness; are clamp or suction marks allowed?
- Is it presented in a fixed pose, or random in a bin (bin picking)?
- Choose the gripping method
- Rigid parts → gripper; flat surfaces → vacuum; porous or oily → test with samples
- Is the cycle tight? Do several products need a quick changer?
- Decide on vision / force control
- Uncertain position → vision guidance; quality decisions → vision inspection
- Precise insertion, surface-following polishing → force control; confirm tolerances and contact-force range
Key points
- Wrong EOAT and even the best robot fails; never quote the bare arm alone.
- Vision depends on lighting and calibration; force control has a window and can't fix inconsistent parts.
- Whether people can work close by depends on safety ratings and the risk assessment; put the sensing solution in scope.
Exercises
- For a live opportunity, draw four boxes: workpiece → gripping → localisation → force/vision, and mark the unknowns.
- Write 5 questions you must confirm with the customer.