Industrial Cleaning: What Hall Floors in Production and Logistics Really Demand

The search term is ambiguous, so the scope comes first. Special cleaning — degreasing production plant, cleaning extraction ducting, high-pressure work on facades and silos — is a separate trade with its own qualifications, its own permits and its own equipment. Nothing below applies to it. What follows is about the floor: maintenance cleaning of large hard-floor areas in production and in warehouses and logistics centres, and about which parts of that job an autonomous machine can take over.

The floors themselves are more varied than the word "concrete" suggests: power-trowelled and hardened concrete, epoxy or polyurethane coatings, screed, industrial tiles in wet areas, plus expansion joints, floor markings, dock levellers and ramps. The manufacturer restrictions follow that variety. The user manual of the CenoBots L50 names its intended surfaces as hard floors — "tile, wood floor, Concrete, terrazzo, Granite, etc." — in indoor areas. Coatings are not on that list, which is not a prohibition but a reason to test on your floor before signing anything.

Who does the work splits the market. Larger sites keep an in-house team for daily floor care and buy in a service provider for periodic deep cleaning; smaller ones outsource everything. Both models are under the same pressure, and it is a staffing pressure before it is a technology one — the shifts nobody wants to fill are the night and weekend shifts on empty hall floors. We have written about that pressure separately in our piece on the labour shortage in commercial cleaning. Robots do not solve it. They move a specific, repetitive, large-area part of it off the roster.

Walk the hall and sort what is on the floor into two piles. Loose material — dust, wood and metal particles, shrink-wrap remnants, strapping bands, cardboard scraps, pallet splinters, granulate — is a sweeping problem. Adhering material — the black tyre marks forklifts leave on light-coloured concrete, a thin film of oil or cutting fluid, dried-on spills — is a scrubbing problem. No sweeper removes anything that adheres, because no water and no brush pressure are involved. No scrubber usefully handles bulky loose debris; it pushes it, wraps it around the brush or blocks the squeegee. That is why sweeping runs before scrubbing rather than instead of it.

On the dry side we operate two machines. The Pudu MT1 sweeps 70 cm wide at up to 1,800 m²/h in standard mode and up to 6,000 m²/h in spot mode, holds 35 litres, runs 4–8 hours "differentiated by cleaning levels", charges in under 3 hours and needs 75 cm of clearance. Its manual restricts it to flat indoor environments. The MT1 Max does 2,200 m²/h (7,000 m²/h spot), 5–10 hours with standard mode stated at 8, charges in under 3.5 hours — and, unlike the base machine, is built for the yard side of the building: full IP54 sealing, −10 to 45 °C, 5 cm speed bumps, slopes up to 8°. Loading yards, gatehouse aprons and covered ramps are its territory. The full six-machine field, including the ones we only researched, is in our comparison of robotic sweepers.

On the wet side the CenoBots L50 is the hall-scale scrubber: a 510 mm disk brush, 55 litres of fresh and 55 litres of waste water, up to 6 hours of runtime, 67 dB(A), and a manufacturer figure of 2,203 m²/h that CenoBots itself labels maximum theoretical productivity. Two restrictions matter more than the throughput and both sit in the manual rather than the datasheet: indoor use only, and in automatic mode no gradient above 2 per cent — ramps are manual operation. What that means for planning is in our detailed piece on the L50. In most halls the honest allocation is a sweeper as the workhorse across the whole floor and the scrubber on defined zones: dispatch and staging areas, main forklift lanes, assembly walkways.

In a hall, dust is a process problem before it is a cleanliness problem. It settles on optical sensors and light barriers, it works into linear guides, it ruins coating and printing results, and depending on the substance it is a hazardous-substance question rather than a housekeeping one. The German hazardous-substances ordinance lists among its general protective measures, in § 8 (1) no. 5 GefStoffV, "angemessene Hygienemaßnahmen, insbesondere zur Vermeidung von Kontaminationen, und die regelmäßige Reinigung des Arbeitsplatzes" — appropriate hygiene measures and the regular cleaning of the workplace. It prescribes regular cleaning; it prescribes no interval, no method and certainly no robot. Which substances occur on your floor, and what that implies, belongs in your hazardous-substances assessment. We run cleaning machines, we do not give legal or compliance advice.

That makes filtration the specification to compare, and here the market splits in a way the brochures do not advertise. Among the machines in this comparison, only one carries a certified class: the datasheet of the Adlatus SR1300 — newly onboarded in our programme — states filters "der Klasse M" with 5.5 m² of bag filter or 6.4 m² of cartridge filter area and automated shaking built into the software. For the Kemaro K900 the class-M figure comes from a company blog post written in the Gen-I era, not from the current factsheet; what the factsheet documents is the patented Automatic Filter Cleaner, sold as a filter system that "significantly reduces dust levels". The CenoBots S5 brochure specifies HEPA filtration with a filter shaker and makes no class-M claim at all. For the two sweepers we actually operate, Pudu publishes a description and no class: "Utilizes high-flow negative pressure ventilation and an efficient filter to trap particles in the debris box, effectively preventing secondary pollution." The MT1 Max adds an internal vibration system that keeps the filter clear during the run. If your dust is fine, wood-based or metallic, that difference is the first thing to put on the table — and the honest answer for our fleet is that no filter class is published.

One operational detail from the same documents is worth carrying into any site visit, because it names the actual dust source. Adlatus writes that the side brush is normally switched off and raised after edge cleaning "um unnötige Staubaufwirbelung zu vermeiden" — to avoid unnecessary raising of dust. The side brush throws material toward the machine; that is exactly what makes it useful on edges and exactly what makes it a dust source in the middle of a hall. And one number nobody publishes: how many square metres one hopper filling covers. We went through the six machines in our comparison of robotic sweepers and found the figure for none of them, so nobody can give you an emptying interval honestly — it comes out of the first weeks on your floor.

Mixed traffic is the question every safety officer asks first, and the regulatory anchor is about people, not robots. The annex to the German workplace ordinance states for traffic routes: "Werden Transportmittel auf Verkehrswegen eingesetzt, muss für Fußgänger ein ausreichender Sicherheitsabstand gewahrt werden" — where transport equipment is used on traffic routes, an adequate safety distance must be maintained for pedestrians. The same annex requires floors to be free of unevenness, holes and trip hazards and to be "trittsicher und rutschhemmend", non-slip and safe to walk on. A cleaning robot is neither a pedestrian nor a designated means of transport, so nothing in that text answers the question for you: whether an autonomous machine may drive in your lanes, and under which conditions, is decided in your risk assessment. What the text does establish is that the wet film a scrubber leaves behind is a floor-condition topic with a name.

Against that, here is what the manufacturers actually document. For the MT1 Max, Pudu describes vehicle behaviour explicitly — the robot "can stop to let vehicles pass, pause at lane entrances, or edge aside" — and equips it with a 1.2-metre warning light plus a safety projection light and audible and visual alarms. For the L50 there is no manufacturer-confirmed forklift or industrial-truck detection anywhere in the official material; what is documented is a 3D LiDAR with up to 150 m detection range, glass detection and protruding-object detection. That distinction is not pedantry. A machine that names vehicle behaviour in its own documentation is a different conversation with your safety officer than one that names general obstacle avoidance. Among the machines we assess from documentation, Kemaro takes the most visible approach with its BlueSpot, a blue light that per the official factsheet alerts "employees and forklift drivers to the robot’s movements — even in areas with limited visibility or beneath conveyor systems".

Two hard limits close the topic. Kemaro states in its official FAQ, without qualification, "The robot is forbidden from working in ATEX zones" and "The robot is forbidden from sweeping flammable or explosive substances" — and the machine-specific standard offers no way around it: EN IEC 63327:2021 applies to commercial indoor floor treatment machines in addition to IEC 60335-2-72 and explicitly "does not apply to machines designed for use in corrosive or explosive environments (dust, vapour or gas)". Explosion-hazard areas are therefore outside this machine class, full stop. Everything else about certificates, declarations of conformity and what a performance level does and does not tell you is in our article on safety and standards for cleaning robots.

The retail assumption that the building empties at night does not survive contact with a three-shift plant or a logistics centre that dispatches around the clock. In practice four windows exist, and they are all shorter than a night: the gap at shift changeover, the weekend, planned shutdowns and maintenance days, and mixed operation in defined zones while production runs. The last one is the realistic default in most halls, and it turns the question from "when is the hall empty" into "which zone is free when". Aisle by aisle, a robot that is allowed to clean the raw-material bay during the early shift and the dispatch area during the late shift covers more square metres per week than one that waits for an empty building it never gets.

Sizing that window means resisting the multiplication everyone does. The MT1’s official runtime is 4–8 hours "differentiated by cleaning levels" — the eight hours belong to the low-intensity end, so multiplying maximum area rate by maximum runtime produces a number the manufacturer never claimed. The same applies to the L50: six hours against a manufacturer-labelled maximum theoretical 2,203 m²/h arithmetically gives a ceiling of roughly 13,200 m², and real coverage sits below it, because turning, obstacles, edge passes and refills are all inside that time. How to read runtime figures properly, and which specification matters more, is the subject of our article on battery runtime in cleaning robots.

The wet and dry sides also differ in how much window they need. Behind a sweeper the floor is dry and a forklift can drive over it immediately — that is what makes dry machines compatible with running operations. Behind a scrubber-dryer there is a thin drying film that needs time, signage and a route plan, which is why the scrubber generally belongs in the quiet window and the sweeper does not have to be. If you are considering unattended runs, our guide to running cleaning robots at night works through what actually has to be true for that: docking and resuming, alarm systems and motion detectors, doors, and who gets the notification at three in the morning.

The honest list of what a robot does not take over is longer than the sales version. Edges and corners: CenoBots states edge cleaning to under 5 cm from the wall for the L50, which is good and is not zero. Everything under racking, roller conveyors and machine tables, because these machines are 490 mm tall (MT1) and 675 mm tall (MT1 Max) and simply do not fit — the low-build answers on the market, such as Kemaro’s K700 Compact at an official 25 cm of height, we assess on documentation alone. The immediate surroundings of machines, with their cable ducts, chip trays and pallets. Fresh oil spills and leaks, which are a binding-agent and a bucket, not a cleaning run. Sanitary and welfare facilities. And a category distinction that matters in food and pharma environments: these machines clean, they do not disinfect, and we make no hygiene claims for them.

Documentation is where expectations and reality drift furthest apart, so here is exactly what exists. For the sweepers, Pudu’s official material describes the PUDU Link app: remote monitoring by phone or PC with "real-time updates, cleaning reports, and overflow alerts". That is a run log — which route, when, how long, plus an alert when the hopper is full. It is genuinely useful in a hall where nobody sees the night shift, and it is not a hygiene certificate, not a heatmap and not a proof of cleaning result. Our fleet’s sweepers publish no camera-based result verification at all. Customers and auditors increasingly ask for documented evidence of cleaning performance, and it is worth agreeing early which of your requirements a run log can serve and which it cannot; we work through the options in our piece on documenting cleaning.

If you are still at the stage of deciding which machine class fits at all, which cleaning robot suits which operation runs through the decision from the floor upward rather than from the product sheet downward. What drives the cost of industrial cleaning — area, soiling type, window, edge share, deep-cleaning intervals — is a separate topic that we treat as a factor question rather than a price list, in our article on what cleaning robots cost. And if you want the specific answer for your hall rather than the general one, a walkthrough with the two piles from section two is a better hour than any spec sheet.