A roof cleaning robot suited to skylights needs continuous suction above 3,500 Pa, a secondary tether, and navigation that reads frame lines on a slope of 20–45 degrees. A general roof robot built for flat membrane roofs will fail on angled glass.
People search “roof cleaning robot” and “skylight cleaning machine” interchangeably, then discover they bought the wrong class of machine. The surfaces are different, and so are the robots.
What is the difference between a roof robot and a skylight machine?
A membrane-roof robot walks on flat roofing material and pushes debris. A skylight machine lives on vertical or pitched glass and must grip a smooth, sometimes wet surface while staying attached. Suction, not traction, is the core technology on glass.
If your roof is 90% membrane with two skylights, you mostly need a glass machine for those panels. If it is a glass atrium or conservatory roof, you need a glass machine, full stop.
What should you check before buying?
- Suction rating held after 30 minutes, not peak
- Secondary tether and its load rating
- Slope range — does it cover your actual pitch?
- Water tank capacity and recovery (streaks come from poor recovery)
- IP rating, IPX4 minimum for exterior glass
- Consumable prices: brushes, filters, squeegee blades
Is a skylight cleaning machine safe on a live atrium?
Only with a tether and a fall zone kept clear below. A machine that loses suction over a shopping centre floor is a hazard, not an inconvenience. This is the one place where a cheaper robot is genuinely the wrong choice — the safety system is worth the price difference.
The Lingfeng S1 is built for framed, obstacle-heavy glass roofs where a rigid robot stalls at the first mullion; it is a useful reference for what obstacle-crossing means on a real roof.
How much will skylight cleaning cost you?
Manually, expect $3–$8 per m² per visit including access. A robot at $10,000–$18,000 that covers one glass roof twice a month pays back within three years on a 1,000 m² roof. Under 300 m², the math is thinner and renting makes more sense.
If suction, slope, and tether feel like a lot to weigh, the model-level comparisons in all skylight robot reviews will shorten the shortlist.
How do you test a robot before you buy?
Ask for a site trial on your own glass. A machine that behaves in a showroom can still stall at the first mullion on your roof, and the only way to know is to watch it cross a real frame line at a real pitch. Two hours on site beats a week of spec sheets.
During the trial, watch the water recovery. Streaks come from fluid left behind, and a machine with weak recovery will look impressive for one pass and clean poorly by the third. Ask the operator to clean the same panel three times and check the third result.
What maintenance does a skylight robot need?
Filter rinsing every session, brush and squeegee checks weekly, and seal inspection monthly on sun-facing glass. Consumables budget around 8–15% of machine price a year. That routine is what keeps suction stable, and stable suction is what keeps the machine on the glass.
Keep a spare set of consumables on site. A stopped machine waiting on a filter delivery is a cleaning schedule slipping, and on a pitched roof there is no safe manual fallback you want to reach for.
What does the operator need to know?
How to read a suction drop, how to reset the robot after it stalls at a frame, and when to stop. A skylight robot that keeps grinding at a mullion will wear its own gearbox, and a machine that loses grip on a pitch needs a tether check, not another restart. Training is twenty minutes and saves a machine.
Document it. If a second operator takes over, the suction baseline and the frame-handling rules should be written down, not passed on verbally. On a roof, institutional memory is a safety system.
What about skylights you cannot reach safely at all?
Some glass roofs sit above voids with no safe roof access, and those need a different answer again — permanent anchors, a fixed track, or a design change at the next refit. A robot does not solve access problems; it needs access to be deployed. If the roof itself is unreachable, fix the access first.
That is the honest limit of this category. The machines are good at the cleaning; they are not a workaround for a building that was designed without a maintenance route.
How do you compare two candidate machines honestly?
Run both on the same panel, on the same day, at the same pitch, and time them. Then clean the same panel three times with each and compare the third pass for streaks. That single test exposes more than any spec comparison, because it catches recovery and edge behaviour that brochures never mention.
Weigh the result against consumable cost and duty rating, not just the first-pass look. A machine that wins the beauty pass but costs double to maintain is the wrong buy for a commercial roof.
Key Takeaways
- Roof robots and skylight machines differ: glass needs suction, membrane needs traction.
- Demand held suction above 3,500 Pa and a rated secondary tether.
- Match the slope range to your actual glazing pitch (20–45° for most skylights).
- Never run an untethered robot over a live public space.
- Below 300 m² of glass, renting usually beats buying.