A window cleaning robot holds itself to glass with a vacuum pump that keeps the pressure inside its suction chamber below atmospheric — typically 2,500-4,000 Pa of clamping force. Two drive belts walk the body across the pane while a pad cleans and a backup battery keeps the seal for 20-30 minutes if power dies.
That is the whole trick, and it is old. The interesting part is not how it works but how narrow the safe operating window is, and why two machines that look identical on a product page behave differently on the same pane.
What actually keeps the robot on the glass?
Three systems, in order of importance.
- The pump. A small vacuum pump evacuates the chamber under the chassis. The pressure difference — atmospheric outside, reduced inside — presses the machine onto the pane.
- The seal. A flexible cup or skirt on the underside keeps the chamber sealed. This is the weak link. A 0.2 mm nick in the rim or a speck of grit under the lip is enough to leak.
- The backup battery. If mains power fails, an internal battery keeps both the pump and an alarm running long enough to recover or lower the machine. Rated 20-30 minutes on most units.
The friction drive is a separate matter — two belts or tracks with rubber tread walk the body across the glass. They do not hold the machine on. If the pump loses seal, the treads are irrelevant. That confusion is common, and it leads operators to blame tread wear for a fall that was actually a seal leak.
How does it navigate around the pane?
Two philosophies. Simpler units use a pattern path: down, across by one body width, up, across, repeat — sometimes with a second crossing pass. Smarter units use edge-detection sensors to find the frame and adjust the pattern to fit, which cuts the number of wasted turns at the top and bottom rails.
The difference is worth about 15-25% on runtime and coverage, and it is the main thing you are paying for at the higher price bands.
Why do some robots fall off, and what stops them?
Falls happen when the seal breaks and the pump cannot recover before gravity wins. Common causes, ranked by how often I see them in the field:
| Cause | What you notice first | Fix |
|---|---|---|
| Contaminated cup rim | Repeat suction alarms on one pane only | Clean rim, replace if nicked |
| Wet glass at high pitch | Sliding before the alarm | Reduce water, work below 35° |
| Tether pulling sideways | Drifts off path in one direction | Re-anchor, add slack |
| Aged backup battery | Machine drops instantly on power cut | Replace pack, test twice a year |
| Frame overrun | High-centres at a kerb | Measure kerbs, or buy a chassis that clears |
Does the glass type change how it works?
Yes. Low-E coated glass and self-cleaning glass have very low surface energy — water beads instead of sheeting, and friction between the tread and the pane drops. On a vertical pane this is manageable. On a pitched skylight, low friction plus a water film is the exact combination that produces slides.
Textured or patterned glass is worse in a different way: the seal cannot sit flush, so the pump runs continuously to maintain vacuum and battery runtime collapses. If your skylight has pattern-embossed glass, most suction-based robots are simply the wrong tool.
Does the water system matter to the mechanism?
More than people expect. A single-tank unit sprays water ahead of the pad and vacuums nothing back. On a vertical window the residue runs down and away. On a skylight, residue stays on the pane. Two-motor units that pull dirty water into a recovery tank leave a cleaner dry cycle, which then means less re-wetting and fewer passes — and fewer passes means the seal is under load for less time.
Ratings on suction are usually quoted dry. Our walkthrough of how the mechanism behaves on a skylight covers the wet-glass numbers. For the practical difference between two well-tested chassis, the Lingyun Y3 review and the Lingkong K3 review are worth reading side by side.
What should you actually test before buying?
Ask the supplier for three numbers that marketing pages rarely give: suction measured with a wet pad fitted, rated pitch with a wet pad fitted, and the minimum frame height the chassis clears. If any of those is missing, assume the best case in the spec sheet is the dry-glass case.
Then do a single test yourself. Ground floor, one pane, half-wet, and watch the machine’s behaviour at the frame edge. Edge behaviour tells you more about build quality than any spec.
If you want the wider comparison first, the full window cleaning robot review index sorts units by what they are actually good at.
Key Takeaways
- Suction, not tread friction, keeps the machine on the glass — 2,500-4,000 Pa of vacuum is typical.
- The cup rim is the weakest component; a small nick or grit spec can break the seal.
- Backup batteries are rated 20-30 minutes and are almost never tested by owners.
- Low-E and patterned glass change friction and seal behaviour enough to break a machine’s stated envelope.
- Ask for wet-pad suction, wet-pad pitch and frame clearance before you buy.