A window cleaning robot works by pumping air out of a chamber between its base and the glass, so atmospheric pressure holds it up while drive wheels move it along a taught path and a squeegee lifts the dirty water away. Remove the vacuum and it falls; that is the entire safety model.
Understanding that single principle explains most of the strange behaviour owners report: slowing down, slipping, stopping halfway, leaving a strip of water at the frame. All of those are pressure problems, not software problems.
What is the vacuum doing, exactly?
A small pump pulls air from under the chassis faster than it leaks back in. A 90 mm chamber with roughly 2800 Pa of negative pressure can hold several times the machine’s 1.2–1.6 kg mass against the glass, which leaves a large safety margin on a clean pane.
That margin shrinks fast. On a dusty pane, the gasket cannot seal. On a warped or wavy pane, the same thing happens in patches. On very smooth low-E glass, the machine sometimes needs a slightly wet pad to hold — a counter-intuitive detail that confuses new owners.
Why does the path look so slow and repetitive?
Because the robot is doing three jobs at once: moving, keeping seal, and wiping. Turn too fast and the chamber breaks contact at the corner. The fixed zig-zag path trades speed for grip.
Typical coverage is 60–90 m²/h on clean flat glass and closer to 40 m²/h on dirty or inclined glass. That is the honest number, and it is well below most marketing figures, which come from continuous-run tests on prepared panes.
Machines that handle interruptions well — vents, mullions, ridges — are worth a premium on real buildings. Our Lingfeng S1 review covers how obstacle logic actually behaves when a robot meets a frame mid-run.
Does it actually clean, or just smear?
Both, depending on setup. A wet pad followed by a squeegee pass removes dust properly. A dry pad dragging on a dirty pane redistributes it as a grey film that looks fine from inside and awful in raking light.
The most reliable improvement is a two-stage routine: spray and loosen, wait twenty seconds, then run a squeegee pass. That single change beats every accessory on the market.
| Glass condition | Seal quality | Real coverage | Notes |
|---|---|---|---|
| Clean flat glass | Strong | 80–90 m²/h | Baseline |
| Dusty exterior | Weak in patches | 40–55 m²/h | Filter clogs quickly |
| Wavy or tempered | Intermittent | 45–60 m²/h | Gasket cannot follow |
| Inclined skylight 30° | Reduced | 35–50 m²/h | Gravity load |
| Coated low-E | Good when wet | 60–70 m²/h | Dry pad slips |
What stops the robot mid-pane?
Three causes dominate. A gasket hairline crack, which owners never inspect. A clogged filter, which reduces airflow by 30–40%. And a low battery, where the pump sags before the drive motors do.
The first one costs almost nothing to check: run your finger along the rubber gasket before every session. A split is visible and cheap to replace. Most of the complaints that say the robot is broken end there.
Who should not rely on a robot?
Anyone whose glass has heavy oxidation, paint overspray or silicone contamination. Those need solvents and hand work; a robot will polish the contamination into the surface.
Also anyone in a market where service access is poor. These machines need gaskets, filters and pads, and none of those are universal parts.
How does the machine get water onto the glass?
Small internal reservoirs, usually 60–100 ml, that feed a spray nozzle or a pre-wetted pad. That is enough for roughly 6–10 m² per fill, which surprises most buyers.
On a large glazed wall, refilling becomes the bottleneck. A corded unit with an external supply line solves it. A portable unit does not, and pretending otherwise wastes half a day.
| Component | Typical spec | Failure mode |
|---|---|---|
| Vacuum pump | 2800–3000 Pa | Filter clog, seal leak |
| Gasket | Silicone, 8–12 mm | Hairline crack in cold |
| Drive wheels | 4, geared | Belt stretch after 12 months |
| Squeegee | Rubber blade | Nicks from grit |
| Reservoir | 60–100 ml | Nozzle scale in hard water |
Is the fall-arrest sensor enough?
No. It is a backup, not a primary safety system. It triggers when the machine senses free fall or a sudden pressure loss, and by then gravity has already started the job. The safety line is the thing that keeps the machine — and whoever is below — safe.
If you want the broader picture across machines and their real behaviour on skylights and facades, the skylight cleaning robot reviews archive breaks the field down by grip type.
Read the specifications knowing that all of them assume a clean, flat, dry pane. Your building is not that. Buy margin, not peak numbers.
Water chemistry is the part of the mechanism almost nobody explains. Pure water with a low mineral content sheets off glass cleanly; hard water leaves a mineral film within seconds of drying. If your site water reads above 200 mg/L total hardness, a simple filtration cartridge changes results more than a more expensive machine would.
Then there is timing. A squeegee pass over glass that has already dried removes almost nothing and can scratch. The workable rule is to keep the pad one pane behind the squeegee, never more. Crews that set a machine running and come back later get a dried film that looks like the robot never cleaned at all.
One more physical limit worth stating plainly: the machine cannot clean the outer frame or the glazing beads. Plan on a hand pass at the perimeter every cycle, which for a typical skylight is two minutes and for a curtain wall is more like twelve. Buyers who expect a complete finish from the robot alone are always disappointed.
Key Takeaways
- A window cleaning robot is a vacuum system first and a cleaning appliance second.
- Real coverage is 40–90 m²/h depending on dust and slope, not the continuous-run peak.
- Filter clogs and gasket cracks cause most grip failures; both are cheap to fix.
- Reservoir capacity of 60–100 ml limits most portable units to 6–10 m² per fill.
- The fall-arrest sensor is a backup — the safety line is the actual protection.