Safety and Standards for Cleaning Robots: CE, EN IEC 63327 and the Fine Print

In practice the questions fall into four families, and they are remarkably consistent across warehouses, production halls, retail and clinics. First: what happens on contact with people and vehicles? A robot in a warehouse shares its aisles with pedestrians, pallet trucks and forklifts, and the forklift is the asymmetric risk — the robot may see the forklift long before the driver sees the robot. Second: who is responsible during unattended operation? Autonomous night runs move the question from "does it stop in time" to "who is called when it does not restart".

Third: does the machine create a hazard of its own? A scrubber-dryer leaves a thin, drying film of water behind the squeegee. That is a slip hazard with a location and a duration, and it is an operational question — signage, routing, drying time — not something a certificate resolves. Fourth: the electrical and mechanical side of docking. Charging stations, emptying stations and their surrounding clearances are equipment in their own right and need their own place in the assessment.

These questions can be answered from documents — but only partly, and the answers differ machine by machine. What follows is what the public documentation of these machines actually contains, quoted and attributed. We are not a certification body and give no legal or conformity advice: the risk assessment for your site, and the decision whether a machine may operate there, belongs to your safety officer and your occupational-safety specialist. In production environments that assessment usually also involves the works council and the insurer.

CE marking is the baseline for any machine placed on the EU market, and on its own it tells a buyer very little. The document that carries information is the EU declaration of conformity: it names the manufacturer, the exact product and model, the directives applied, and the harmonised standards used. One of the machines in our fleet prints its declaration directly in the user manual — the CenoBots L50 manual closes with a declaration for the product "Autonomous Scrubber Dryer", model L50, referencing Machinery Directive 2006/42/EC alongside the EMC and radio-equipment directives, and listing the harmonised standards applied, among them EN 60335-2-72:2012 and EN 60335-1. It is signed and dated 31 July 2023. What that list does not contain is EN IEC 63327 — and that absence is the single most useful thing this article can point at.

EN IEC 63327 is the machine-specific standard for exactly this class of equipment. IEC 63327:2021 (published 2021-05-04; the British adoption BS EN IEC 63327:2021 followed on 2021-08-13) is titled "Automatic floor treatment machines for commercial use – Particular requirements" and covers, in its own words, "the safety of powered automatic floor treatment machines intended for commercial use indoors" for sweeping, scrubbing, wet/dry pick-up, polishing, wax application and shampooing. Two sentences of its scope matter for a buyer. It applies on top of the general commercial floor-machine standard: "The requirements given by this standard are applied in addition to the requirements for commercial floor treatment machines in IEC 60335-2-72, as far as applicable." And it excludes precisely the environments people most often ask about — it does not apply to "machines designed for use in corrosive or explosive environments (dust, vapour or gas)".

Several manufacturers do cite the standard, and two of them are machines we run. The Nexaro NR 1700 datasheet states that the robot "erfüllt die neuste Internationale Norm IEC 63327" as well as the requirements of Machinery Directive 2006/42/EC, and that it is certified by TÜV Süd. Pudu's CC1 Pro deck carries an "IEC 63327 Certification" heading and the line "Compliant with IEC 63327 Standards". What almost nobody else does is attach a performance level to that standard. Adlatus Robotics states on the official SR1300 product page that an interplay of different sensors increases navigation stability and "bringt die Kollisionsvermeidung auf den Performancelevel D", which "der internationalen Norm IEC 63327 entspricht". The current 09/2025 datasheet repeats the claim but drops the letter, stating only a performance level corresponding to the European EN IEC 63327; the superseded 05/2022 English datasheet had spelled it out as "performance level D". The same datasheet states that the sweeper-vacuum system is CE certified, and that lidar sensors enable reliable detection of obstacles and people and early detection of ledges, levels and steps. All of this is Adlatus’s own documentation, and we assess it from those documents only.

What does the letter mean in plain language? It grades the safety-related control function, not the machine as a whole, and it does not tell you whether that grade is sufficient for your site. The generic standard behind performance levels, EN ISO 13849-1 ("Safety of machinery — Safety-related parts of control systems — Part 1: General principles for design", 2023 edition), says so itself: "This document does not specify the safety functions or required performance levels (PLr) that are to be used in particular applications." Read plainly: a manufacturer stating "performance level D" is describing what the machine achieves. Which level your application requires is determined in your risk assessment, by people qualified to do it — not read off a brochure.

Start with the machine that publishes the most. Pudu’s official page for the MT1 Max — a dry sweeper, no water anywhere in its process — states: "Equipped with a 1.2-meter high warning light and a safety projection light, combined with audible and visual alarms, MT1 Max enhances visibility for pedestrians and vehicles, proactively preventing potential risks." The same material describes pedestrian and animal avoidance that "quickly reacts to sudden ‘pop-out’ movements", and dynamic vehicle avoidance in which the robot will "stop to let vehicles pass, pause at lane entrances, or edge aside". Add full IP54 sealing against dust and water spray, a water-drop sensor that triggers a protective avoidance mode, an operating range of −10 to 45 °C, 5 cm speed-bump crossing and 8° slope climbing. That is an unusually complete safety description for this market. What Pudu does not publish anywhere for this machine is a norm citation.

Kemaro — a Swiss manufacturer we assess from documents only — takes the opposite approach: less light-and-sensor marketing, more written norm detail. The official K900 Gen II factsheet describes the BlueSpot as a feature that "enhances workplace safety by using a blue light to alert employees and forklift drivers to the robot’s movements — even in areas with limited visibility or beneath conveyor systems"; the compact K700 carries the same light. The official FAQ adds "mechanical drop prevention, significantly safer than the optical one" and a "self-voltage charger, allowing the charger to be touched without any risk", states IP54 protection, and names concrete standards for the components: the battery "complies with EN 62133-2:2017", the charger with "IEC 60335-1:2010+A1+A2 and IEC 60335-2-29:2016+A1". A CE declaration is not published on the public site — the manual sits behind a login. That is not evidence of a missing CE mark; it means the document has to be requested.

Drop protection is where the class splits into mechanical and optical, and the difference is worth asking about. Kemaro states its prevention is mechanical and argues that this is the safer of the two. Optical solutions are the norm elsewhere: Pudu’s official CC1 brochure labels a "Cliff Sensor" and a "Bumper Sensor" in its multi-fusion sensor diagram and notes the robot "can stop at any angle and move away as soon as it encounters obstacles"; the Nexaro NR 1700 datasheet lists a long-range laser distance sensor, a bumper sensor, eight cliff sensors, two magnetic-field sensors, a wall-following sensor and incremental sensors — and no camera at all; SoftBank’s Whiz datasheet summarises its safety features as "safety sensors detect and maneuver around people, objects, and cliffs". What none of these documents publishes is the drop height each system reliably detects. If your site has loading docks, pit edges or open floor openings, that is a question for the manufacturer in writing, not for a brochure.

Emergency stops are documented where manuals exist. The L50 manual describes a yellow-and-red emergency stop button on the rear top of the machine; pressing it stops the drive and brush motors, and there is a scrubber-specific detail worth knowing: "If the emergency stop button is pressed while the machine is performing a cleaning task, the vacuum motor will briefly run before coming to a stop in order to minimize residual water on the floor." Pudu’s official CC1 Pro component diagram likewise shows an emergency button, and the CC1 user manual lists a "Notausschalter" as component number one. The SoftBank Whiz datasheet labels an "Emergency Stop Button" on its machine diagram. Gausium's Phantas manual instructs the operator to press "the red emergency stop button on the control panel" to stop the robot immediately, and puts a check of that button into the pre-use routine. The list is therefore longer than the two machines that print a full description — but for several market machines the e-stop is described only on distributor pages, and that is not documentation, so we do not treat it as such.

One thing is missing across an entire product category, and it is safety-adjacent rather than cosmetic. For not one of the dry sweepers in our source set does the manufacturer publish an official noise figure: Pudu publishes none for the MT1, MT1 Max or MT1 Vac, Kemaro none for the K900 Gen II or K700, Adlatus none for the SR1300, CenoBots none for the S5. The only numbers in circulation come from individual distributors and contradict each other. Meanwhile the scrubbers and vacuums do publish: the CC1 at below 70 dB, the L50 at 67 dB(A), the Nexaro NR 1700 at 49 to 57 dB(A) depending on mode, the Whiz at 62 dB in normal mode. That gap matters for hearing-protection thresholds, for night work and for anyone assessing an eight-hour shift alongside a machine. We cover it in detail in our sweeper comparison — and we do not fill the gap with invented numbers.

The sharpest boundaries in this market are not drawn by standards but by the manufacturers, in their own documents, in plain language. The clearest example is ATEX. Kemaro states in its official FAQ, without qualification: "The robot is forbidden from working in ATEX zones." And immediately after: "The robot is forbidden from sweeping flammable or explosive substances." If you were hoping the machine-specific standard would help here, it does not: EN IEC 63327 explicitly does not apply to machines designed for corrosive or explosive environments. Two independent documents, the same answer. Anyone offering you a cleaning robot for an explosion-protected zone should be asked to put it in writing.

Gradients are the second boundary, and the trap is that a datasheet figure and a manual figure are different numbers describing different things. The CenoBots L50 user manual states two limits in the same safety list: "Do not operate manually the machine on the ramp with a gradient exceeding 10%" and "Do not run it in auto mode on the slope with a gradient exceeding 2%". The same list forbids operating the machine on a slope in automatic mode at all. A 10 % ramp figure quoted from a spec sheet therefore says nothing about what the robot may do autonomously — and autonomous is the only mode that matters for an unattended night run. For comparison, Kemaro states a maximum ramp slope of 12 % for the K900 Gen II and the K700; whether that figure is a manual-mode or an autonomous limit is exactly the question to ask.

The third boundary is outdoors and weather, and the same product family can sit on both sides of it. The MT1 Max is positioned by Pudu as an outdoor and semi-outdoor machine — IP54, water-drop sensor, −10 to 45 °C — and even its official page carries the footnote "Do not perform cleaning tasks in extreme weather". The base MT1 is a different story: its user manual restricts the machine to "use only in indoor flat environments", not "outdoors on rough or non-hardened surfaces", within 0 to 40 °C and at 90 % relative humidity or below. Same series, same brand, incompatible operating envelopes. The L50 manual is equally blunt in its do-not-operate list: "Outdoors. This machine is for indoor use only." It also forbids use "in rooms where explosive or highly combustible substances (e.g., gas, solvents, heating fuel, dust, etc.) are stored or processed" — which quietly rules out a range of production and storage areas that look perfectly ordinary on a floor plan.

Five documents settle most of what a safety officer needs, and all five are requestable before signing. First, the EU declaration of conformity for the exact model and variant you are buying — not for the product family. Second, the full operating manual, before signing rather than after delivery: as the L50 example shows, the manual, not the datasheet, carries the restrictions. Third, written norm citations. If a datasheet names EN IEC 63327, ask what exactly the statement covers: which safety function, at which level, verified how. Fourth, the restriction list in one place: ATEX and flammable substances, gradients separated by mode, indoor versus outdoor, temperature and humidity, permitted floor types. Fifth, what happens on failure — emergency stop, behaviour on obstacle contact, and the docking station.

Docking deserves more attention than it usually gets. Charging is the mild case; self-emptying is the interesting one. Adlatus documents its S1300 as an emptying station with a fenced, sensor-monitored safety zone, discharging automatically into a standard 1,100 l DIN EN 840 container from up to 1,450 mm, at 3,500 × 2,000 × 2,000 mm and roughly 240 kg. That is not an accessory tucked into a corner. It is a piece of equipment with its own footprint, its own clearances and its own place in your assessment — and it is worth pricing the floor space into the decision from the start.

Then there is the hazard the robot creates rather than avoids. Scrubber-dryers leave a thin film of water that takes time to dry, and no certificate manages that for you: it is signage, routing and scheduling. Which floors tolerate wet cleaning at all, and how much residual moisture a squeegee actually leaves, is the subject of our scrubber-dryer guide. That the L50 keeps its vacuum motor running briefly after an emergency stop specifically to minimise residual water shows the manufacturer thought about this — and it is the kind of detail you only find by reading the manual.

Finally, the operating window. Unattended night runs are where documentation and reality diverge most, because the machine’s restrictions still apply at 3 a.m. when nobody is there to notice them; we work through that in our guide to night operation. And if you are still deciding which class of machine you are assessing at all — dry sweeper, scrubber-dryer or vacuum, each with a different hazard profile — which cleaning robot fits which operation walks the whole decision. Our part of the job is to hand over the documents and say plainly which claims are the manufacturer’s and which are verified. The assessment itself stays with your safety officer, and it should.