A mid-range skylight cleaning robot in 2026 covers 300-600 m2 per hour on a clean rectangular bay, runs 4-6 hours per battery charge, and carries a 15-25 litre tank that lasts roughly 45-90 minutes on wet mode. A water feed line removes the tank limit entirely.
These numbers get quoted in every 2026 brochure, and every one of them is technically true. What the brochures omit is how fast they collapse on a real roof with mullions, valleys and a parapet in the way.
What actually slows a robot down?
Frame breaks. Every mullion the robot crosses costs a few seconds of hesitation and repositioning. A bay with mullions every 1.5 m can cut throughput by 40% against the datasheet figure. Deep valleys and steep transitions cost more still.
The second drag is edges. A robot spends disproportionate time near the perimeter turning and re-anchoring. On a 100 m2 bay that is minor; on a 12 m2 decorative skylight it dominates the cycle entirely.
| Roof type | Throughput | Battery | Tank |
|---|---|---|---|
| Open rectangular | 450-600 m2/h | 5-6 h | 20-25 L |
| Mullioned panels | 280-380 m2/h | 4-5 h | 15-20 L |
| Valley and curve | 180-280 m2/h | 3-4 h | 12-18 L |
How long does a battery really last in 2026?
Cold weather erodes runtime. Lithium packs lose 20-35% of their useful capacity below 5 degrees C, so a winter cycle that claims 5 hours delivers closer to 3.5. Plan winter jobs with a shorter shift or a spare pack.
Charge cycles matter too. Most 2026 packs are rated for 800-1,500 full cycles before capacity drops to 80%. At one charge per working day that is several years, but fast-charging between cycles shortens it. Charge overnight on the slow setting when you can.
Is a water feed line worth it?
Yes, if you clean more than 800 m2 per visit. A tank forces a pause and a refill, and the refill trip is where cycles lose 20 minutes. A 10 mm feed line to the roof, plus a simple drain for runoff, turns a stop-start cycle into a continuous one.
If a permanent line is impossible, carry two tanks and a quick-coupler. It is crude but it halves the reload downtime without any building work.
How does water quality affect the numbers?
Hard water leaves mineral spots that force a second pass, and a second pass halves the effective throughput. If the site has hard water, treat it or use a spot-free rinse, or the machine’s hourly figure quietly collapses in real use.
Soft or treated water lets a single pass finish the job on most panels. That is why experienced operators care more about the rinse water than about the cleaning speed on paper.
Does the operator affect throughput?
More than any spec sheet admits. An operator who plans the route before starting, keeps the water topped up, and clears the edges early will beat one who reacts as they go by 20-30% on the same machine. Skill, not hardware, is often the limiting factor.
This is another reason training pays for itself quickly. A well-taught operator squeezes more out of the machine every single cycle, and the difference compounds over a year.
What is the realistic daily output?
On a decent roof, plan for 1,500-2,500 m2 in an eight-hour shift including setup, water handling and one battery change. That figure holds far better across real buildings than the raw hourly rate, because it already prices in the interruptions a brochure never mentions.
Use the daily figure for staffing and scheduling. Use the hourly figure only for comparing machines on the same roof. Mixing the two is the fastest way to a missed deadline.
How do you set a realistic contract target?
Quote clients a daily output range, not an hourly peak, and leave room for one slow bay and one weather stop. A contract that promises a rate no roof delivers is how suppliers lose money on their own machine.
Track actual daily output for the first month, then set the contract number from real data. Machine work rewards measured promises far more than optimistic ones.
Which efficiency claim should you ignore?
Ignore any single number with no conditions attached. “600 m2/h” means nothing without the bay layout and slope. Ask what the figure was measured on, and ask to see a cycle on your own worst bay.
The Lingfeng S1 tends to hold throughput better on tight-bay roofs because its body is light and it repositions faster. On large open atria the Lingyun Y3 gains more from its longer runtime. Match the machine to your layout and the datasheet flatters itself less.
Send your roof layout to Skylight Cleaning Robot and ask for a throughput estimate on your actual bay dimensions.
Key Takeaways
- Expect 300-600 m2/h; mullions and valleys can cut that by 40%.
- Battery runtime of 4-6 hours drops 20-35% in cold weather.
- A 15-25 litre tank covers roughly 45-90 minutes of wet cleaning.
- A water feed line pays off above about 800 m2 per visit.
- Any throughput number without roof conditions is meaningless.
Does cleaning speed change with glass condition?
Badly soiled glass always slows the machine. A first clean on a roof that has not been touched in two years can run at half the normal rate because the robot passes over the same area twice. After the first cycle, throughput climbs to its expected range and stays there.
This matters for the very first job. Do not judge the machine on its first deep clean of a neglected roof. Judge it on the third cycle, when the glass is at its steady-state dirt level.