A roof cleaning robot fits a skylight canopy when the glass pitch is between 15° and 45°, every edge has a 40 mm upstand, and no opening vents interrupt the panel run. Outside those limits, a pole kit or rope crew does the job better.
That short list rules out more roofs than buyers expect. The proportion of atriums that pass all three is maybe half. The other half are still worth cleaning — just not with a robot.
Why do skylight canopies reject robots more often than facades?
A vertical facade is a plane. A skylight canopy is a plane with holes.
Ridge lights, smoke vents, access hatches and fire dampers all interrupt the glass. Each one either needs a bridging frame or forces the robot into a detour, and detours cost far more time than the product brochure accounts for.
The second reason is sealant. A 12-year-old canopy has silicone-wiped edges that sit slightly proud of the glass. A robot vacuum skirt riding over that lip loses grip intermittently. Not a failure, but the machine slows and its path drifts.
What slope range actually works?
Below 15°, the canopy holds standing water and the robot’s wheels spin in a shallow film. Above 45°, gravity load on the vacuum seal climbs and the fall-arrest line becomes a real operational constraint, not a formality.
The 25–35° band is the sweet spot. Most industrial ridge lights sit there, and most commercial atriums are flatter than that, which explains a lot of disappointed first purchases.
| Canopy type | Typical pitch | Robot fit | Better option |
|---|---|---|---|
| Industrial ridge light | 25–35° | Good | Robot |
| Flat atrium skylight | 0–10° | Poor | Manual + pole |
| Pitched glass roof | 30–45° | Workable | Robot + tether |
| Barrel vault | Continuous curve | Poor | Rope access |
| Smoke vent cluster | Any | Fails | Crew only |
How do you test a canopy before buying?
Three measurements and one real run.
- Measure the glass plane pitch with an inclinometer on three separate panels. Variance above 3° means the frame is settling.
- Count interruptions per 10 m of run — vents, hatches, sealant ridges. Above three, expect 40% slower cycles.
- Check the upstand height. Under 40 mm, the robot has nothing to catch against at the top edge.
- Clean one bay manually, then run the robot on the same bay and photograph both in the same light.
That last test catches water-spotting issues that specs never mention. Hard-water regions leave a visible film if the squeegee pressure is wrong for the glass coating.
For machines that handle interrupted runs better than most, the Lingyun Y3 review shows how a bridging frame changes cycle behaviour on ridged glass.
Who should skip the robot entirely?
Anyone with a single small skylight over a stairwell. The setup time exceeds the cleaning time by a wide margin, and a long-reach pole does it in fifteen minutes.
Also skip it on heritage roofs with leaded or cast-glass panels. The point load of a robot on 4 mm glass is a cracked pane waiting to happen, and replacements are not orderable.
What does maintenance on a canopy robot look like?
Brush replacement every 300–400 hours, seal check every 50 hours, and a full path re-teach whenever a vent is replaced. Budget four to six service visits a year on a mid-size atrium.
The most-skipped item is filter cleaning. Clogged filters drop suction and the machine starts leaving lap marks. Crews notice the marks and blame the robot.
If your site mixes skylights with vertical glass, read the broader skylight robot reviews before committing to a single machine type for both. Very few machines do both well.
Access is the detail that kills most skylight robot projects and the one buyers think about last. Someone has to place the machine on the glass, monitor it, and lift it off. On a 6 m atrium that means a scaffold tower, a mobile lift or a catwalk. If the roof has no catwalk and the lift rental is 800 dollars a day, the robot’s labour saving disappears entirely on small canopies.
Older canopies also have a hidden version of the same problem: no safe anchorage. A tether needs a certified point within a few metres of the work area. Where the closest engineered anchor sits ten metres away, you are adding fall-arrest slip line, an extra person, and 20 minutes of setup per bay. That single factor turns a two-hour job into a five-hour one, and no spec sheet will tell you so.
Water quality on canopies deserves its own line. Roof glass collects runoff from adjacent metal roofing, and that water carries iron and grit. A robot dragging a grimy squeegee across a 12 m pane leaves grey streaks that are visible from the floor below. Filter the supply and change the filter on a schedule tied to rainfall, not to hours of use.
Treat the first month as a calibration period. Log cycle time, water use per bay and any grip alarms. A canopy that needs 20% more time than the sales estimate in month one will not get faster later; it will get slower as sealant degrades. Knowing that early lets you renegotiate the service contract or buy a second machine before the dry season, rather than during it.
One last physical point: check the glass thickness and the frame span. A robot at 5.5 kg concentrated on four contact points puts local loads in a different class from a window cleaner on a pole. Most framed canopies handle it easily. Structural glass with point-fixed spider fittings needs a load check with the original engineer before anything heavy goes on the surface.
Key Takeaways
- Robots fit skylight canopies at 15–45° pitch with a 40 mm upstand; the 25–35° band works best.
- More than three interruptions per 10 m of run slows cycles by about 40%.
- Old silicone ridges reduce vacuum grip — check before purchase, not after.
- Heritage and thin cast glass should never take a robot’s point load.
- Filter cleaning is the most-skipped task and the most common cause of lap marks.