A roof cleaning robot that grips flat glass perfectly can still slide on a 25° skylight. The five things that break grip in 2026 are frame lips, condensation, surface heat, grit and cable drag, in roughly that order.
Vendors demo on flat vertical glass because it is easy. Skylights are not flat and not vertical. They sit at 10° to 35°, they have frames, and the glass gets hot in the afternoon and wet in the morning.
The result is a category of buyer regret that is easy to avoid: units that clean a showroom pane flawlessly and then stall on the second row of a real atrium.
Why does slope change suction performance?
Suction holds a robot against gravity on a vertical pane. On an inclined pane, suction has to hold against a component of gravity that grows with the sine of the angle. At 25°, roughly 42% of the machine’s weight pulls it downslope. At 35°, 57%.
Good controllers compensate by raising vacuum, which raises motor load and heat. On a 30° roof at 38°C ambient, drive motor temperature is the real constraint, not grip. Machines rated for 3 hours of vertical work often do 70–90 minutes continuous on a pitched roof before a thermal pause.
| Slope | Share of weight downslope | Typical continuous runtime | Watch for |
|---|---|---|---|
| 10° | 17% | 2.5–3 h | Water pooling at the low frame |
| 20° | 34% | 1.5–2 h | Seal hardening in summer |
| 30° | 50% | 70–90 min | Thermal derate and tether drag |
| Above 40° | 64%+ | Not recommended | Fixed rail only |
Which slope traps catch buyers in 2026?
Frame lips. A 15–30 mm raised frame at the top and bottom of each glazed panel stops a robot dead unless it has a lift or a bridging roller. Many units rated for 20 mm handle it in the lab on one panel and fail where frames step 4 mm between panels.
Condensation and dew. Morning glass on a north-facing roof stays wet until 10 am in autumn. A wet rubber seal at 25° slides. Run logs from Scottish and Nordic installations show the first run of the day is the one that ends in a stopped machine.
Surface heat. Coated glass reaches 55–65°C in direct summer sun. Seals soften, adhesion drops, and the brush leaves drying marks that are harder to remove than the original dirt.
Grit. Wind-blown sand in dry climates becomes an abrasive layer. At slope, grit collects in the lower frame channel, and the robot’s lower wheels ride that channel first. Replace the wheels or scrape the channel before deployment.
Cable drag. A tether hanging over the top edge of a pitched roof pulls the machine uphill on one side. Over a 12 m run this is a constant offset that shows up as curved tracking and missed strips.
What suction figure should a skylight roof cleaning robot have?
Ignore millimetre-of-mercury numbers on their own. Ask for holding force at 30° slope with a 3 kg payload, and ask how many minutes it sustains before thermal derate.
A unit that holds 6 kPa on flat glass is not necessarily holding 6 kPa down a pitch under load. Real-world skylight work favours 4,500–6,000 Pa units with active slope compensation over peak-spec vacuum numbers on fixed hardware.
The Lingyun Y3 sits in that band and is honest about the derate curve, which is more useful than a single peak figure. Our Lingyun Y3 review walks the numbers and what they mean at 25°.
How do you test a machine on your own roof before buying?
Two hours, one panel, three conditions. Test dry midday, test wet, test the panel adjacent to a 4 mm step. If it makes all three, it is probably fine for your site.
Ask for the tether at full length during the test. Half the failures we see are tether-related and only appear on the long run.
Who should avoid a pitched skylight robot?
Roofs above 40°, roofs with heavy snowload frames, and roofs where individual panels step more than 25 mm. At that point you need a fixed rail system, and no current portable skylight unit handles it.
Owner-occupiers of a single 6 m² skylight should not buy. Hire a rope cleaner once a year. The economics only close above roughly 150 m² of glass, or where access equipment hire is expensive and recurring.
What does a site survey need to measure?
Six numbers: panel pitch in degrees, frame step height in millimetres, glass coating type, morning dew pattern, ambient summer maximum and the roof diagonal for tether length. Miss the frame step and you buy the wrong machine; miss the diagonal and you buy the wrong tether.
Take the readings at 2 pm in summer if you can. That is when heat derate shows up and when the tether is hottest and least willing to spool flat.
How much does a pitched-roof unit cost compared with a vertical facade unit?
Roughly 15–25% more for the same glass area, because slope compensation, lift rollers and steeper vacuum duty all add hardware. A 300 m² pitched skylight system lands in the $10,000–$16,000 range in 2026, against $8,000–$13,000 for an equivalent vertical facade unit.
Do not try to save that margin by buying a vertical unit. It will work for a season on a shallow 10° roof and then fail the first time summer heat softens the seal.
What are the risks to plan for?
Two failures repeat on real sites, and both are cheap to prevent once someone names them in the contract.
- A wet, cold seal on a north-facing pitch in autumn loses grip at far lower slope than the datasheet allows.
- Grit trapped in the lower frame channel grinds the drive wheels, and replacement sets are often a 3-week order.
Key Takeaways
- Slope is a suction problem, but heat is usually what stops the run.
- At 25°, 42% of the machine’s weight pulls it downslope.
- A 4 mm frame step between panels breaks more robots than a 15° pitch does.
- Above 40° pitch, portable units stop being the right tool.
Before you commit to a model, read our Lingyun Y3 review and all skylight robot reviews. Both are written for buyers running real rooftops rather than spec-sheet readers.