A painting robot pays for itself on repetition. The moment the part changes every hour, the programming time — not the robot — becomes the cost. Three approaches exist to attack that, they solve different problems, and they are routinely confused with each other in sales conversations. The SR series from Zhihui Robot Technology (Wuxi) offers all three, which makes it a convenient way to explain the differences.
1. Vision-based automatic path planning
A camera measures the shape and position of the part as it arrives, and the system generates the spray path directly on the captured point cloud. Nobody teaches a path; the geometry produces one.
This is the right answer for flat parts and simple curved surfaces arriving in a mixed sequence — panels, doors, sheet-metal covers — where writing a program per part would take longer than spraying it by hand.
Where it stops: the camera gives geometry, not process. Fan width, fluid flow, gun distance, overlap and the number of passes still come from the paint data sheet and a trial panel. A generated path with the wrong flow gives you a consistently wrong film build.
2. Vision positioning and path correction
Here the path is taught once, for the part in its nominal position. In production a camera measures where the part actually is, and the taught path is shifted before it is sent to the robot.
This solves a different problem: not part variety, but part position. Work hung on a conveyor swings. Parts laid on a trolley are never square. Hand-loaded fixtures repeat to a few millimetres at best. A fixed taught path sprays where the part should be; correction sprays where it is.
Where it stops: it corrects position, not shape. A part that is bent, or dimensionally different from the one you taught, is a fixture and tolerance problem, and no camera fixes it at the gun.
3. Lead-through teaching from a hand-sprayed part
An operator sprays the first part by hand while the robot records the motion. Every part after that is sprayed by the robot repeating what the operator did.
This is the fastest route to a first working program on a part nobody has programmed before, and it puts the knowledge of the person who actually knows the part into the robot.
Where it stops: what you have recorded is one person’s technique on one day, including the places where they were tired, rushed, or compensating for a gun that needed cleaning. Treat it as a first draft to be measured and tuned, not as a finished program. Film-build readings on the first robot-sprayed part are what tell you whether the recording was any good.
Which one you actually need
| The problem you have | The approach that fits | What it does not fix |
|---|---|---|
| Many different parts, geometry known, flat or simply curved | Vision path planning | Process parameters — still yours to set |
| Same part, never in the same place | Vision position correction | Shape variation and bad fixtures |
| New part, nobody has programmed it, needed this week | Lead-through teaching | Consistency — the recording inherits the operator’s faults |
| Same part, same place, high volume | None of the three — a taught program is faster and more repeatable | — |
The honest summary
These three options widen the range of work a painting robot can take on; they do not remove the paint engineering. Every one of them still needs the film-build target, the paint data sheet, the trial panel and someone who can read the result. Where a supplier presents vision as a replacement for process knowledge, that is a sales position, not an engineering one.
Frequently asked questions
Does vision path planning work on complex 3D parts?
The manufacturer positions it for flat parts and simple curved surfaces. Deep recesses, undercuts and shadowed features are where generated paths struggle, because the geometry the camera sees is not the geometry the paint has to reach.
Is lead-through teaching the same as a collaborative robot?
No. Lead-through teaching here means recording the path an operator sprays with the gun in hand. A collaborative robot is a machine class defined by how it behaves near people. The two are often sold in the same sentence and are not the same thing.
Can these be retrofitted to a robot we already have?
Ask us with the make, model and controller. Vision correction is often possible on an existing cell; generated path planning is usually tied to the supplier’s own software stack.
Which SR models offer this?
The SR series is offered with these options by the manufacturer. Which options apply to a given model and controller version is confirmed at quotation — we ask them for your model rather than assuming.
Request a quote Send drawings and we check the selection before quoting.