Gausium Marvel: Sweeping and Scrubbing in One Pass — What the Figures Show

Gausium's official title for the machine is "Commercial Robot Floor Cleaner for Large & Complex Spaces"; the German brochure calls it a "Professioneller Roboter-Bodenreiniger für große und komplexe Räume". It has three cleaning modes — sweep, scrub and dust mop — and the point of the machine is that two of them run at the same time. Front-mounted dual side brushes collect fine dust and coarse debris into an integrated tray, and immediately behind them rear disc brushes scrub the same strip wet. There is no cylindrical roller involved in the sweeping; Gausium calls the arrangement a patented side brush system and states explicitly in its own comparison piece that the brushes "capture debris into the tray, with no roller involved".

That distinction matters more than it sounds, because it separates the Marvel from its own sibling. Gausium introduced the Marvel at CMS Berlin in September 2025 together with the smaller Mira, and its Interclean Amsterdam press release of 14 April 2026 lists both as "now commercially available following their introduction at CMS Berlin 2025". The Mira sweeps with a roller brush; the Marvel does not. And the two are not variants of one machine but genuinely different sizes: Mira is 680 × 550 × 900 mm and 130 kg with 22 and 20 litre tanks, a 400 mm scrubbing width, 25 kg brush pressure and a 660 mm minimum passing width. The Marvel is 1108 × 699 × 1168 mm and 220 kg with 80 and 70 litre tanks, a 600 mm scrubbing width, 55 kg brush pressure and needs 850 mm to pass. If your aisles are the constraint rather than your area, the smaller machine is the relevant one — we cover it separately in our Gausium Mira market overview.

Eight machines currently sit in Gausium's product navigation: Beetle, Phantas, PhanShop, Mira, Marvel, Omnie, Scrubber 75 and Vacuum 40. Read against that list, the Marvel is the large-format end of the sweep-and-scrub concept that Mira and Marvel introduced as a pair. The Phantas is the compact 4-in-1 at the opposite end of the range, Omnie and Scrubber 75 are scrubbers without a sweeping stage, Beetle is a dry industrial sweeper and Vacuum 40 a carpet-capable vacuum. Gausium recommends the Marvel for sites above 3,000 m² and names retail and supermarkets, manufacturing plants, warehousing and logistics, airports and subway stations, hospitals and shopping malls as its target environments.

One more piece of hardware is worth knowing about before you picture the machine only as a robot. The Marvel has a manual mode: a power-assisted handlebar and an optional stand-on pedal let staff drive it like a conventional walk-behind machine for spot work and repositioning between autonomous runs. In the German material Gausium calls this the "kraftunterstützten Griffbügel und ein optionales Stehpedal".

Start with what is solid, because for this machine most of it is. The English brochure of 7 April 2026 and the German brochure of 28 May 2026 are identical line by line, and the specification page agrees with both in every metric value. Cleaning widths: 600 mm scrubbing, 840 mm sweeping, 760 mm dust mopping. Maximum theoretical cleaning efficiency: 3,024 m²/h scrubbing, 4,234 m²/h sweeping, 3,830 m²/h dust mopping. Tanks: 80 litres fresh, 70 litres waste. Battery: 120 Ah lithium iron phosphate. Runtimes: 5 to 7 hours scrubbing, at least 8 hours sweeping, at least 10 hours dust mopping. Size 1108 × 699 × 1168 mm, net weight 220 kg, maximum speed 1.4 m/s, minimum passing width 850 mm, minimum U-turn width 1,300 mm, gradeability 10°. Brush pressure is 55 kg. Navigation runs on 3D LiDAR with a 360° vision system and SLAM localisation, and Gausium supports full remote deployment, so mapping and setup do not require someone on site.

Every one of those performance numbers carries the same footnote, and it is Gausium's own: "The specifications are derived from Gausium lab test results; actual performance data may vary in specific applications." Treat 3,024 m²/h as a ceiling under laboratory conditions, not as a planning figure — a machine that has to slow for obstacles, turn at aisle ends and dock for water will not hold it. Gausium publishes no practical range for the Marvel the way it does for smaller models, so there is no honest second number to give you here. Ask for one in writing, for a floor plan like yours.

Now the two contradictions, both small, both worth knowing because they tell you how carefully the material was assembled. First, the specification page states the scrubbing efficiency as "3,024 m²/h 21,700 ft²/h" — but 21,700 ft²/h is the conversion of 2,016 m²/h, which is the Mira's scrubbing figure, and it appears verbatim on the Mira spec page too. The English brochure gets it right with 32,550 ft²/h. The metric values are consistent everywhere, so use metric and ignore the imperial column. Second, the German product page carries a feature block headed "Doppelte Effizienz" whose text reads "verkürzt Mira die Reinigungszeit … um die Hälfte" — on the Marvel's own page. The English page names the Marvel in the same sentence. The German text was evidently ported over from the sibling. Neither error changes a metric specification, but both are reminders to compare against the brochure rather than a web page.

What is missing matters more than what contradicts. There is no noise figure for the Marvel anywhere — not in either brochure, not on the spec page, not at any dealer we checked. There is no suction figure in pascal, no IP rating, no battery voltage or kWh, no stated volume for the debris tray, and no published CE or conformity documentation. One distributor, Robobee, states a charging time of 3 hours; no official source confirms it, so treat it as unverified until Gausium puts it in writing. And the headline marketing claim — that combining two steps into one pass "cuts cleaning time in half compared to machines that sweep and scrub separately" — is a manufacturer statement with no independent measurement behind it. The mechanism is plausible; the factor of two is not something anyone has demonstrated publicly.

Our reference in this class is the CenoBots L50, the industrial large-area scrubber we run on logistics and production floors — so the comparison runs between the Marvel as documented and a machine we know from operation. The L50 is specified at 2,200 m²/h with a 510 mm disc brush, 55 litres fresh and 55 litres waste water, up to 6 hours runtime and roughly 4 hours charging, 1.2 m/s, 67 dB(A) operating noise, gradeability up to 6° — and, as it happens, exactly the same 220 kg net weight as the Marvel. Same floor-loading class, same lift question, same "where does it live when it is not working" problem. One L50 figure we deliberately leave out of that list is the squeegee width: CenoBots' live product page gives 740 mm, the datasheet 770 mm with an 840 mm option, and the user manual 800 mm.

On paper the Marvel scrubs faster — 3,024 against 2,200 m²/h, roughly a third more — but do not spend that difference before you have it. Gausium's figure is an explicitly theoretical lab maximum, our L50 figure is a specified cleaning efficiency, and the two are not measured the same way. The tank gap is the more concrete advantage: 80 litres fresh against 55 means longer between refills on a large floor, and that is usually what decides whether a night shift completes unattended. In the other direction, the L50 has a published noise figure and the Marvel has none, which is not a small thing when you are planning where and when a machine may run.

The real argument for the Marvel is not speed, it is passes. On a warehouse floor that has both dry debris and wet soiling, our fleet does that job with two machines: the Pudu MT1 Max sweeps dry at up to 2,200 m²/h across a 700 mm width into a 35-litre debris container, with up to 10 hours of runtime and no water anywhere in the process, and the L50 scrubs afterwards. The Marvel folds both into one machine and one pass. That is a genuine capability our fleet does not currently have — and it is worth stating plainly rather than talking around. The counter-argument is equally plain: two machines are two independent battery budgets, two maintenance paths and two things that can fail without stopping the whole floor. One machine that does everything is also one machine whose breakdown stops everything.

Two practical consequences follow for anyone comparing. First, the wet trace: in sweep-and-scrub mode the Marvel leaves the same thin, quickly drying film any scrubber leaves, so the floor is briefly damp behind it — in sweep-only or dust-mop mode it is completely dry, like the MT1 Max. Plan the mode, not the machine. Second, mode-dependent runtime is a planning tool: 5 to 7 hours wet, at least 8 hours sweeping, at least 10 dust mopping. If you are still working out whether a scrubber is the right category at all, our guide to scrubber robots sorts that question, and the L50 machine profile covers what large-area scrubbing looks like in daily operation.

Begin with what is documented rather than what is suspected. The Marvel navigates with 3D LiDAR and a 360° vision system — vision means cameras, in the literal sense — and it is a cloud-connected machine by design: Gausium advertises full remote deployment and fleet management, and its technology page describes a cloud platform with over-the-air updates, 24/7 remote diagnostics and an open API for integration with other applications and IoT devices. A machine that can be deployed and diagnosed remotely is a machine that sends data off your premises. That is not a defect, it is the product; it just has to appear in your documentation.

What Gausium says about that data is thinner than a German buyer needs. The company privacy policy states that collected personal information "will be stored on Amazon Cloud servers in your country or region" and that users "can choose the cloud server area where your personal information will be stored based on your location", and it contains a clause about transferring domestically collected personal information to overseas institutions. It names no concrete EU location and does not mention the GDPR. Separately, a statement of 25 February 2026 attributed to CTO Dr. Baoxing Qin says Gausium "is fully GDPR-compliant, collecting training data exclusively from the China market, validating models there, and then deploying validated software upgrades to overseas fleets without collecting any customer data from international markets". That claim does not stand unopposed. The Clingendael Institute, the Netherlands Institute of International Relations, reported that data from Gausium robots is stored in Singapore and questioned whether it enjoys GDPR protection there. Manufacturer and think tank contradict each other, nothing in the public record resolves it, and we are not going to resolve it here — which is precisely why this belongs in a contract rather than in a brochure. It is also worth knowing the structure behind the machine: Gausium has one European office, in Zaltbommel in the Netherlands, and no German subsidiary — the German market runs through independent dealers.

None of that is a contract, and that is the point. For any cloud-connected cleaning robot carrying cameras — irrespective of manufacturer or country of origin — insist on five documents before deployment: a data processing agreement under Art. 28 GDPR, the named storage region for operational and map data, the sub-processor list, the handling and retention period for camera imagery, and the technical and organisational measures. For public tenders, hospitals and works-council-sensitive sites this is not paperwork, it is a knock-out criterion. We apply the same standard to every machine in our own fleet; our article on data protection with cleaning robots sets out what to ask and why.

One conformity gap belongs in the same conversation and is specific to this machine. We found no published CE declaration, no conformity documentation and no IP rating for the Marvel — not on the official site, not in either brochure, not at any dealer. For a 220 kg autonomous machine that Gausium markets for operation in mixed forklift and pedestrian traffic, that is the first document to request, not the last. Treat the manufacturer's "zero-blind-spot perception" and its safety claims for mixed traffic as marketing language until a declaration of conformity and the applicable standard are in front of you in writing.

Shortlist the Marvel if your site is large, hard-floored and dirty in both senses at once: warehouses and production halls where dust and tyre marks arrive together, big-format supermarkets, car parks with dry grit and oil film on the same slab. Gausium's own recommendation of more than 3,000 m² per site is a sensible floor, and 850 mm minimum passing width with a 1,300 mm U-turn is the geometry your layout has to satisfy. The 10° gradeability is a genuine differentiator for ramped car parks: our own L50 is specified up to 6°, so a steeper ramp puts it outside its rating.

Do not shortlist it where the aisle is the constraint rather than the area — below 850 mm the Mira is the machine to look at, not the Marvel. Do not shortlist it for anything outdoors, since no IP rating is published. And be careful with the two verticals Gausium names that we would not promise without a measurement: hospitals and airports have night-time corridor work near sleeping or resting people, and there is no noise figure for this machine at all. If you are earlier in the process and still sorting categories rather than brands, start with our guide to choosing the right cleaning robot.

Then there is the commercial reality of buying one in Germany, which is not the same as the machine being ready. Gausium's Interclean press release of April 2026 calls the Marvel commercially available, but the German distributor Terra Robotics still lists it as "Coming Soon" with a "Preis anfragen" button and a leasing enquiry form. So the first question in any sales conversation is not about specifications, it is: what is your delivery date, and where does the machine physically come from? Ask about the workstation in the same breath. Two DACH distributors, SEBOTICS and Robobee, advertise an optional workstation for the Marvel with self-docking, charging, fresh-water refill, waste-water discharge, tank rinsing and detergent dosing — but Gausium's own workstation page lists only WS-01 for the Scrubber 50, WS-02 for the Scrubber 75 and WS-03 for the Phantas, and nothing for the Marvel. If unattended water handling is what makes the business case work, get the model designation, its footprint and its plumbing requirements in writing before you count on it.

And in the demo, test what no datasheet can settle. Drive the 850 mm passing width and the 1,300 mm U-turn through your narrowest real aisle, not your average one. Bring a sound level meter and take your own dB(A) reading at the operating point, because no published figure exists to fall back on. Watch the tray fill: no volume is published, so measure how long it lasts on your worst floor before someone has to empty it. Run the machine until the fresh water is gone and note the area actually covered — Gausium publishes no coverage-per-fill figure, and the 80 litres are only an advantage if they last your shift. Check what the side brushes throw when they meet real debris rather than test dust, look at the strip left along your worst wall, and ask for the CE declaration, the sensor list and the data processing agreement before the pilot, not after. If you want that same list worked through with us in the room, that is what our demo appointments are for.