Commercial Scrubber Robots: Which Size Fits Which Floor

Mechanically, a scrubber is four assemblies working in sequence: a fresh water tank that meters water and detergent onto the surface; a rotating brush — disc or roller — that loosens the dirt with a defined contact pressure; a squeegee directly behind it that pulls the dirty water back up; and a recovery tank that holds it. Pudu's official manual for the CC1 calls the two tanks "solution tank" and "waste water tank", 15 litres each. Contact pressure is the number that decides how much cleaning actually happens: 18 kg on the L3, 20 kg on the L4, 25 kg on the L50, 27 kg on the walk-behind SH1. For the CC1, Pudu publishes no brush pressure figure at all.

The order of brush and squeegee is what makes the class work. The squeegee is wider than the brush — on the SH1 a 490 mm lip trails a 440 mm brush — so it also catches the water thrown out sideways. What stays behind is not a water film and not a puddle but a thin film of moisture. How fast that dries depends on the floor, the water setting and the air movement; none of the manufacturers in our portfolio publishes a drying time, and we do not invent one.

That is what separates this class from everything else driving across floors autonomously. Sweeper robots work completely dry: they sweep loose debris into a bin, they wash nothing. Vacuum robots pull dust out of carpet and joints. Only the scrubber dissolves adhering dirt with water and mechanics — the dry class is covered in our sweeper robot comparison. Inside the wet class there is one important exception: per Pudu, the CC1 is a four-in-one machine with the official modes sweeping, scrubbing, carpet vacuuming and silent mopping. The CenoBots machines are pure wet cleaners — for the L4 the manufacturer publishes no dry vacuum function at all. Coarse debris there gets swept up beforehand by someone.

The official floor lists are narrower than the brochures suggest. Pudu's manual gives the CC1's application scope as terrazzo, marble, floor tiles, epoxy resin, sandstone, artificial stone and low-pile carpet — hardwood and laminate do not appear on it. For the SH1, Pudu's datasheet explicitly does list hardwood flooring, and the L50 manual names tile, wood floor, concrete, terrazzo and granite. These lists do not contradict each other; they apply per machine. Anyone wanting wood or laminate wet-cleaned should therefore settle it per device and per surface, not as a blanket answer.

At the bottom sits the Pudu CC1: 500 mm cleaning width with side brush, 15 litres each of fresh and waste water, 75 kg, an official 700–1,000 m²/h. Pudu positions it for "small and medium-sized scenarios", and its manual states runtimes separately per mode: 5 hours scrubbing, 5 hours sweeping plus vacuuming plus mopping, 4 hours carpet vacuuming, 9 hours silent mopping. The frequently quoted "5 hours maximum runtime" is therefore the scrubbing figure, not a maximum — and the manual footnotes all of them as lab measurements at the lowest gear setting, from 100 % down to 10 % battery. With a 70 cm minimum path clearance the CC1 gets through doors and narrow aisles that are too tight for every larger machine. If you are weighing CC1 against CC1 Pro, the differences are in our direct comparison.

The mid-size is held by two CenoBots machines with quite different characters. The L3 is the narrow one: 400 mm working width, dual disc brushes at 18 kg contact pressure, 25 litres per tank, 127 kg, a 700 mm minimum passable width and an Nvidia chip rated at 100 TOPS. On the LiDAR, the manufacturer's own documents disagree: the product page and the L3 brochure state a 96-beam 3D LiDAR, the 2026 full-line catalogue a 128-beam unit — most plausibly a model-year change, which is worth asking about per delivery. The L4 is wider at 450 mm and longer-legged with 38 litres of fresh water, but its actual official differentiator is edge cleaning: CenoBots states less than 3 cm from wall surfaces. It fits through 810 mm and weighs 150 kg.

At the top sits the L50 with a 510 mm disc brush, 55 litres per tank, up to 6 hours of runtime and 220 kg of machine. It needs 900 mm of passable width (940 mm as an option), which makes it a hall machine, not a shop-floor one. The manufacturer datasheet states at least 67 dB(A) sound pressure and 6° gradeability; the user manual tightens that further — up to 10 % incline in manual mode only, and at most 2 % in automatic mode, with autonomous slope work explicitly not recommended. And it is an indoor machine for hard floors. Outdoors it has no business.

The ladder no longer ends at those four. Below the CC1 sits the compact ET1 at an official 300–720 m²/h, the CC1 Pro is the AI variant of the entry machine rather than a faster one, the BG1 sits above the L50 at 2,000 m²/h in covered cleaning mode, and the CR700 enters the class at 700 mm working width. All four are available through us; the ET1, the BG1 and the CR700 are newly onboarded, so their assessments are documentation-based and set out in their own articles rather than in the profiles above.

The SH1 belongs to the same technical class but not to the same operating model: it is not a robot. It is a battery-powered, walk-behind scrubber that always requires an operator — there is no autonomous mode, in any version. It scrubs at 440 mm brush width with 27 kg of pressure at 350 rpm, dries behind a 490 mm squeegee, and separates air, liquid and solids inside the recovery tank, which is what prevents drain blockages. Its tanks hold 4 litres each; Pudu states in its own web store that under standard mode refilling and draining are needed roughly every 20 minutes. The machine weighs 27 kg — Pudu's product page long said 17 kg, but we have weighed our own stock. On noise, Pudu's sources still conflict: the product page gives 72/69 dB(A), the official datasheet 76/71 dB(A); we name both rather than pick one.

The first number everyone looks at is area performance, and it is the least reliable one. On the CenoBots machines the advertised value is pure arithmetic: working width × top speed × 3,600. 0.40 m × 1.4 m/s gives the L3's 2,016 m²/h, 0.45 m × 1.2 m/s the L4's 1,944 m²/h, 0.51 m × 1.2 m/s the L50's 2,203 m²/h. The manufacturer says so openly — the label in its own datasheets reads "Max. Theoretical Productivity". That also shows how little the number tells you: the narrowest machine ranks above the wider one because it is allowed to drive faster. No turns, no overlap, no refills and no obstacles are included. Pudu's 700–1,000 m²/h for the CC1 is a band rather than a peak, which makes the two figures not directly comparable.

The second number decides harder but is rarely quoted: minimum passable width. 70 cm for the CC1 and the L3, 81 cm for the L4, 90 cm for the L50. That is not a comfort value but an exclusion criterion. A shelving aisle with 85 cm of clearance, a pallet standing half in the way, a fire door at 80 cm — any machine that does not fit through has an area performance of zero. Measuring the tightest points comes before measuring the total area, not after.

The third number is the tank, and here there is an honest gap. 15 litres on the CC1, 25 on the L3, 38 on the L4, 55 on the L50 — the ladder is unambiguous. How many square metres one filling covers, however, is published by no manufacturer in this portfolio, neither Pudu nor CenoBots. Figures circulating online are back-calculations, so we state none. What does exist officially is an order of magnitude from the walk-behind SH1: Pudu puts its 4-litre tanks at roughly 20 minutes before refilling in standard mode — against 70 minutes of battery. On that machine the tank runs out first, not the battery. For the autonomous machines the comparable figure simply is not published, which is why the docking station is the real answer.

That poses the fourth and, in practice, most important question: does a station change the water automatically, or does a person? For the CC1, Pudu's manual lists automatic charging and automatic water refill and drainage as supported — in conjunction with the docking station, which the same table lists as an optional function. For the L3, CenoBots offers the WS3 workstation handling charging, fresh water, waste water and detergent, plus the WT3 mobile tank with 60 litres of fresh and 50 litres of waste water for sites without a fixed water connection. The L4 gets the same four-in-one workstation. On the L50 the charging station charges, and the larger workstation adds water and detergent. Without a station, someone fills and empties by hand — and in daily operation that is exactly the recurring labour the machine was supposed to remove. Our full-service model exists to keep that off your staff.

The standard pattern is night operation in closed venues, for two reasons: nobody walks across the damp film, and nobody hears the machine. The published values sit close together — below 70 dB(A) for the CC1, 65 dB for the L4 per its EU datasheet, at least 67 dB(A) for the L50 per the manufacturer datasheet. For the L3 the picture does not resolve: the CenoBots catalogue prints "≥ 64 dB(A)", which as a minimum value is obviously a typo, and the Czech distributor states 66 dB. Until that is settled we publish no figure for this machine. Worth keeping in mind for planning: all of these are manufacturer sound pressure statements, not measurements taken in your building.

Daytime operation in occupied venues works, but differently from how it is usually sold. An autonomous scrubber drives a programmed zone; it is good at completing planned passes while staff keep working. For reacting to a bottle that just broke in aisle four, the walk-behind unit is faster: someone takes the SH1 and is there immediately. For stubborn areas, CenoBots lists a DeepClean mode in its catalogue — a selectable two-stage intensive pass in which the machine first soaks with the squeegee raised, then scrubs and extracts. And in any daytime operation one rule holds: the damp film is a slip hazard until it has dried. Signage is not paperwork, it is part of the procedure.

What actually creates work in operation is wear and water. Brushes, squeegee blades and filters are consumables, and the manufacturers have responded: CenoBots uses magnetic quick-release brushes and a tool-free squeegee on the L3, and states 30 seconds for a squeegee change on the L4. On the SH1, the air-solid-liquid separation in the recovery tank prevents the blocked drain that is the most common failure mode of this device class. And the recovered water is waste water — it belongs in a drain connected to the sewer, not in a yard gully. In a full-service arrangement, wear parts, detergent, water handling and maintenance are part of the service, precisely so they do not end up as one more task for a person on site.

Supermarkets are the classic case for the mid-size: hard surfaces, fixed aisle widths, plannable night windows. CenoBots officially lists retail and grocery as target segments for the L3, alongside offices, hotels, schools and hospitals. The L4 plays its edge cleaning here — shelf plinths and skirting boards are exactly the places otherwise finished by hand.

In hotels, the area per zone decides, not the total: lobby, corridors and conference rooms are separate, often small units with doors in between. A CC1 at 70 cm passable width fits that structure; a 220 kg hall machine does not. In hospitals, results additionally have to be documentable — which makes it worth asking which machine records what it cleaned, not just how fast it cleans.

Warehouses and production areas are the L50's territory, with three sober limitations. First: 900 mm passable width rules out narrow racking aisles. Second: indoor only, hard floors, and no autonomous ramp work. Third: there is no manufacturer-confirmed detection trained specifically on forklifts. What is officially documented is dynamic obstacle detection including protruding objects and glass; the phrase "360-degree protection" is CenoBots' own marketing wording, not a measured coverage figure. In mixed traffic with material handling equipment, none of that replaces traffic rules on site.

Which class is right in the first place — wet, dry or vacuuming — is the subject of our overview of all robot types. Once the class is settled, choosing within the scrubbers comes down to measurement: tightest point, area, night window, water connection. Four numbers you can have within an hour on site — and going through them together is what a demo is for.