A 2026 skylight cleaning robot can handle curved glass roofs up to about 30 degrees of pitch and moderate radius, but it fails predictably at panel transitions, on tight barrel vaults under a 6 m radius, and along curved edges without physical stops.
Curved roofs are where buyers discover the difference between a spec sheet and a building.
What types of curved glass roofs exist?
Broadly four: constant-radius barrel vaults, faceted domes made of flat panes, complex freeform shells, and single-curvature canopies over walkways. Robots cope well with the faceted dome because the surface is flat between frames. They struggle with freeform shells, where no two panes share a plane.
If your architect delivered a freeform shell, expect partial automation at best. That is not a product failure; it is geometry.
Where do robots lose grip on curved glass?
At the joints. Each transition between panes is a small ridge, and every ridge breaks the suction seal for a fraction of a second. On a flat roof that is harmless. On a 25 degree curve with 40 mm mullion caps, it happens every 1.2 m of travel, and on a machine with a single vacuum circuit you can hear the pump labouring.
- Dual independent suction circuits survive transitions far better than a single circuit.
- Larger pad diameter helps, but only if the pad can conform – rigid pads bridge ridges and leak.
- Look for a stated hold-down force at maximum rated pitch, not at horizontal.
Lingdu Intelligence’s curve-rated configuration uses a split vacuum manifold precisely for this reason. It costs more and it is the difference between working and not.
How do you handle curved edges safely?
Fit physical stops. Curved glass roofs rarely have the raised kerb a flat roof offers, and edge detection via camera fails against a bright sky or a reflective facade opposite. A 2 m drop at the eaves is still a 2 m drop.
Options that work in practice: a temporary clamped rail along the perimeter for each clean, a tether anchored to a certified point with a shock absorber, and pre-mapped boundaries with physical datum markers the machine re-references each session.
Which curved roofs should stay manual?
Freeform shells with varying curvature, roofs with glass thicker than 40 mm combined with tight radii, and any vault where the mullion caps exceed 60 mm without being flush. On those, a rope crew with soft pads remains the better tool, and forcing a robot on them risks scratching the coating.
For the middle ground – constant-radius vaults between 6 and 25 m, faceted domes, and canopies – robots are the right call. Check the pitch limit notes before you commit, and get a trial run on the worst two bays rather than the easiest one.
Key Takeaways
- Robots handle faceted domes and constant-radius vaults well; freeform shells stay manual.
- Panel transitions break single-circuit suction every 1-2 m – demand dual independent vacuum circuits.
- Curved roofs lack kerbs, so fit clamped rails, tethers or physical datum stops.
- Trial the worst two bays, not the flattest one, before signing for a curved roof.
How do you test a curved roof before buying?
Insist on a trial on the two bays you consider worst: the steepest, the most broken up by frames, the most awkward to reach. A demo on the flattest, cleanest section tells you nothing. Watch three things during the trial – whether the pump pitch changes noticeably across joints, whether the machine ever pauses to re-seal, and whether the operator has to intervene.
Intervention count is the honest metric. A machine that needs a hand every 20 m on a curved roof will not be used after the novelty wears off. One intervention per 100 m is tolerable; more than that is a maintenance burden disguised as automation.
What maintenance do curved roofs add?
- Suction pads wear faster – inspect weekly, replace at first sign of hardening.
- Vacuum pump filters load sooner because of the extra grit at joints.
- Frame rollers and guide wheels need greasing more often on mullion ridges.
- Re-tension the tether or rail anchors each visit; curved roofs shift slightly with temperature.
None of it is dramatic. All of it is easy to skip, and skipped maintenance on a steep curved roof is how machines end up permanently parked.
Is partial automation worth it on freeform shells?
Sometimes. If 60% of a freeform roof is reachable by a robot working from a fixed rail or a small platform, automating that 60% still cuts the rope time on the remaining 40%, because crews can focus on the hard glass. The trap is buying a full machine for a roof it can barely service. Buy for the reachable share, and keep the rope crew for the rest.
What about rain and drainage on curved glass?
Curved roofs shed water in channels, and those channels carry grit to the same line every time. That line soils faster than the surrounding glass, so a uniform pass is not enough; the machine needs an extra slow pass along the low edge. It is the first place streaks appear if you skip it.
Check the drainage path before the first clean. On some 2026 installations the outlet sits below a panel gap, and water can be pushed into the frame cavity if the machine sprays directly at it. Adjust the spray angle, or clean that bay with a hand tool.