Which Cleaning Robot? A Decision Guide for Commercial and Industrial Floors

Scrubber-dryers are the wet class. A brush loosens dirt with water and detergent, a squeegee pulls the dirty water straight back off, and the floor is left with a thin film that dries within minutes. This is the only class that actually washes a floor — greasy films, drink spills and ground-in soiling need water. Our scrubbers cover a wide size band: the CC1 with 15 litres of fresh and waste water each and an official 700–1,000 m²/h, the L4 at 450 mm working width with a 38-litre solution tank, and the L50 with 510 mm, 55 litres and up to six hours of runtime for industrial halls. Which one, and on what basis, is covered in our scrubber guide.

Dry sweepers are the waterless class. The MT1 and the MT1 Max work with dual disc brushes and active dust control — no tank, no water, no drying time, no slip risk, and the floor is walkable the second the machine has passed. Pudu rates the MT1 at max. 1,800 m²/h in standard mode with a 35-litre waste container, and the MT1 Max at max. 2,200 m²/h. The limit of the class is equally clear: sweepers pick up, they do not wash. A film stays a film. The two machines and their real differences are compared in our sweeper comparison.

Vacuums are the dry pick-up class, and the only one that properly handles textile floors. The Whiz covers up to 1,500 m² per charge over three hours, stores up to 600 taught routes and filters 99.97% of PM2.5 particles. The MT1 Vac is officially a three-in-one sweep, vacuum and dust-mop machine that recognises carpet and hard floor in real time, with 55 cm vacuuming width, 70 cm sweeping width and H11 filtration (H13 optional). The NR1700 sits at the other end: 280 mm width, HEPA H13, a 1.4-litre dustbin — and no camera anywhere in its sensor suite.

The fourth class is not a robot at all. The SH1 is an operator-guided upright scrubber-dryer: 44 cm brush width, 27 kg brush pressure, four-litre tanks, 70 minutes of runtime in standard mode. It exists because every real building has areas an autonomous machine will never clean well — behind counters, under fixed tables, in small side rooms. Planning a robot deployment without planning who covers those areas is how robot projects lose their credibility in month two.

Question one: can this floor be wet-cleaned at all? This is where most bad deployments are decided, and it is answered by the manufacturer, not by intuition. Pudu's official CC1 operation guide names the application scope explicitly: terrazzo, marble, floor tiles, epoxy resin, sandstone, artificial stone and low-pile hard carpet. Parquet and laminate appear nowhere in that list — which does not make them impossible, but it does mean nobody is standing behind the result. Carpet is never scrubbed by any machine in our fleet; it is vacuumed dry, and on the CC1 that even needs the optional carpet vacuum attachment. If the floor covering on site is unclear or mixed, the honest sequence is: confirm the build-up with the property, and clean dry until you have.

Question two: how large is the area, and how is it cut up? Working width is the obvious lever — 280 mm on the NR1700, 400 mm on the L3, 450 mm on the L4, 500 mm on the CC1, 510 mm on the L50, 700 mm on the MT1 and MT1 Max. Twice the width is roughly half the passes. But the second number matters just as much and is usually forgotten: the width the machine needs to get through. Pudu specifies 75 cm of narrow-path clearance for the MT1. CenoBots specifies 810 mm passage width and 1,100 mm turning width for the L4, and 900 mm minimum passable width for the L50.

That pairing is what actually decides the machine. A 12,000 m² hall with three-metre aisles is an MT1 or L50 job. A 2,500 m² specialist retail floor with display islands and 90 cm gaps is not — there the L4 wins, not because it is faster, but because it fits and cleans to within less than 3 cm of the wall. Area size alone has misled more selections than any other single number.

Question three: day or night — which is really a question about noise. The published figures are thin and unevenly distributed, so here is the full honest picture for our fleet. Pudu states below 70 dB(A) for the CC1. SoftBank states 62 dB in normal mode for the Whiz. Nexaro publishes per-mode values for the NR1700: 49 dB(A) in Eco, 57 dB(A) in Balanced and Performance. CenoBots states "at least as 67 dB(A)" for the L50 in the official manual. And for the MT1, the MT1 Max and the MT1 Vac, Pudu publishes no noise value anywhere — dealer figures circulate, we do not adopt them. Beyond the numbers there is an operating rule we do not break: vacuums do not run at night next to sleeping people, in hotel corridors or beside patient rooms. A steady vacuum sound in a silent corridor at 2 a.m. is a different experience from the same sound in an empty open-plan office, whatever the datasheet says.

Question four: is the problem debris, dust or washing? These need different machines, and mixing them up is expensive. Coarse debris — cardboard, film, screws, packaging, leaves at entrances — belongs to a dry sweeper. A scrubber sent into it smears rather than cleans and blocks its own pickup. Dust and fibres on textile floors belong to a vacuum. Only a film — grease, drink residue, ground-in traffic soiling on hard floors — genuinely needs water. Large industrial and retail areas are usually two passes, not one: dry first, wet second, and the dry pass is the one that protects the wet machine.

Question five: what does the site actually provide? Dry machines need power and nothing else, which is why they are the easy answer in halls without a water point. Wet machines need a water and drainage concept. The CC1 can run off a docking station with automatic water exchange or a mobile 30/30-litre tank. The L3 has the WS3 workstation, which handles charging, refilling, draining and detergent dosing in one unit, extended by the mobile WT3 tank with 60 litres of fresh and 50 litres of waste water. And some constraints are layout, not plumbing: Nexaro specifies max. 1,000 m² per cleaning job for the NR1700 and max. 12 × 12 m of free open space without characteristic objects to navigate by — a hall that is too empty is as much of a problem as one that is too full. Which of these levers actually moves your running costs is the subject of our cost article.

Retail. Supermarkets and specialist retail are mostly hard, water-tolerant floors with narrow, changing aisles — the CC1 covers the mid-size formats, the L4 the cut-up ones, and on very wide-aisle big-box formats an MT1 dry pass before the wet pass pays for itself. Whether the CC1 or the CC1 Pro is the right version of that machine depends on exactly one question about your operation, which we work through in the CC1 versus CC1 Pro comparison. Furniture stores are the mixed-floor case: hard floor and carpet islands in one route, which is what the MT1 Vac's real-time floor recognition is built for. DIY stores are the debris case — dry first, always.

Industry and logistics. Warehouses and production halls are two-pass sites: the MT1 or MT1 Max for dust, packaging debris and abrasion, the L50 for the ground-in film, with 55 litres of fresh water and up to six hours of runtime. Where the two dry machines differ in practice — and why the Max is not simply "the bigger MT1" — is set out in our MT1 versus MT1 Max comparison. Car parks are the one semi-outdoor zone we can serve today: the MT1 Max is the only machine in our fleet with full IP54 sealing and an official outdoor and semi-outdoor positioning including proactive rain avoidance. The wider market has since produced dedicated candidates for this zone — the CenoBots S5, the Adlatus SR1300 and the CR700 are positioned for parking decks and industrial areas — and all three are now available through our programme. They are newly onboarded, though, so we assess them from documentation until we have run them ourselves.

Offices, public buildings, healthcare. Offices are the easy vacuum case: empty at night, so noise is not the constraint, and the deciding factor is usually data protection rather than cleaning. That is where the camera-free NR1700 belongs — what it does and does not do in daily operation is covered in our NR1700 field report. Large connected carpet areas favour the Whiz with its taught routes. In hospitals, camera-free navigation is a procurement argument long before it is a technical one; corridors get wet cleaning, and vacuums stay away from ward corridors at night. Schools, sports halls, gyms, airports and petrol stations each come down to the same first question — sports floors in particular need both the floor build-up and the point load checked before anything drives on them, given that the L50 alone weighs 220 kg.

Hotels are where the classic mistake happens, so they deserve their own line. Hard-floor public areas — lobby, restaurant, conference floors — are a CC1 job at night. Carpeted corridors are a Whiz or MT1 Vac job during the daytime housekeeping window, never at night. And whatever the layout, the edges stay human: behind reception, under the buffet, in the back-of-house corridors, the SH1 is the honest answer, always with an operator. The broader reason this whole conversation is happening — a labour market that no longer supplies enough cleaning staff — is the subject of a separate article.

Wrong class for the floor. This is the most common and the most expensive. A scrubber put onto a covering the manufacturer never listed; a wet machine sent into coarse debris it will only push around; a dry sweeper expected to remove a greasy film. All three end the same way — a customer who concludes that "robots do not work", when in fact the wrong class was bought. The fix is boring and reliable: read the manufacturer's application scope before choosing, and where the floor is unclear, clean dry.

A vacuum running at night next to sleeping people. Datasheet decibels tempt people into this. The Whiz at 62 dB and the NR1700 at 49 dB(A) in Eco look harmless on paper, and in an empty office at 1 a.m. they are. In a hotel corridor at the same hour they are a guest complaint, and beside a ward they are a clinical problem. Vacuums belong in daytime housekeeping windows in occupied buildings; the night belongs to scrubbers, on hard floors, in areas nobody is sleeping next to.

Underestimating obstacle density. Every machine here is a floor machine, and floors in real buildings are occupied. Pudu is explicit for the CC1: thresholds up to 8 mm can be crossed while cleaning, up to 20 mm while driving through without cleaning, cables must be removed beforehand, heavily greasy floors are not recommended, and the robot is not designed for environments with steps, steep gradients or tight spaces. A floor that is 30% covered by pallets, roll cages or stacked chairs gets cleaned to 70% at best. Robots do not tidy up first — and if nobody clears the floor before the run, coverage falls, not because the machine failed but because the shift did.

Treating catalogue rates as planning figures. CenoBots labels its own numbers "max. theoretical productivity" — 1,944 m²/h for the L4, 2,016 m²/h for the L3, 2,203 m²/h for the L50 — and those are width times speed with no turns, no obstacles, no refills. SoftBank is unusually honest about the gap: alongside the headline 1,500 m² per charge, its Whiz datasheet states an actual cleaning productivity of 465–557 m²/h. In our own fleet the NR1700 cleans at an effective 50–75 m²/h against a datasheet 100 m²/h, and covers 200–300 m² per charge rather than the roughly 400 the arithmetic suggests. Plan the run window with a realistic factor and the deployment holds; plan with the ceiling and it fails in week one.

Stairs. None of these machines climbs a step, and none is meant to. The cliff sensors on the NR1700 and the Whiz exist so the robot stops at the edge — that is a protective function, not a capability. Pudu states outright that the CC1 is designed for level surfaces and is not suitable for environments with steps, steep gradients or tight spaces. Stairwells stay manual, and any plan that quietly assumes otherwise will be corrected by the first landing.

Ramps and gradients. Pudu permits a maximum of 8° for the CC1 and only when it is not cleaning. CenoBots is stricter still for the L50: the manual gives a gradeability of 10% but states that running on slopes in automatic mode is not recommended, and that gradients above 2% are off-limits in auto mode entirely — manual operation only, up to 10%. In practice this means ramps between levels are a transport route, not a cleaning area.

Heavily cluttered and constantly occupied floors. A robot cleans the floor it can reach. Storage areas where pallets stand wherever they were dropped, event halls between set-ups, retail back rooms in peak season — these are not robot areas until someone owns the job of clearing the floor. That is an organisational decision, not a technical one, and it is worth making honestly before the machine arrives rather than after.

Outdoors and wet environments. The MT1 Max with IP54 and proactive rain avoidance is the only semi-outdoor machine in our fleet, and semi-outdoor means a covered yard, an entrance apron, a parking deck — not a street and not a landscaped area. For swimming pool surrounds no manufacturer in our fleet publishes an IP rating or chemical resistance for that environment, so we do not deploy there. CenoBots states plainly in the L50 manual that the machine is for indoor use only. And the boundary that applies everywhere: the robot cleans the floor. Sanitary facilities, glass, surfaces, waste bins and everything above ankle height remain a human job — which is exactly why our service model pairs the machines with a plan for the rest, and why every deployment starts with a walk through the actual building. If you want that walk-through, book a demo; if you want to see the machines side by side first, the full fleet is online.