Vacuum Robots for Industry and Commercial Use: Whiz vs. NR 1700

Commercial vacuum robots work dry. A rotating brush lifts loose material out of the covering, a blower pulls it through a filter into a bag or bin, and the floor is usable again the moment the machine leaves it. That is the entire process — and it also describes the limit. Whatever sticks to the floor stays there: drink rings, grease in front of the coffee point, de-icing salt smears in winter. Those need a wet machine, and our scrubber-dryer guide sorts out which one. Coarse debris across large hard-floor areas, in turn, is a sweeper job, which we work through in the robotic sweeper comparison.

The reason the class is needed anyway is carpet. On textile coverings a wet process is not an option — water does not belong in the pile — so the vacuum robot remains the responsible machine, and carpet is common in hotel corridors, office floors and conference areas. What autonomous carpet care looks like day to day is a topic of its own, covered under autonomous carpet cleaning.

The third point is where most plans come apart: a vacuum robot is a running suction unit. The Whiz is specified at 62 dB(A) in normal mode, the NR 1700 at 49 dB(A) in Eco and 57 dB(A) in Balanced and Performance mode. That is roughly the level of a normal conversation — entirely uncritical in an empty office at night, defensible during the day in occupied rooms depending on mode. Outside hotel room doors at three in the morning it is not. At night, offices are this class's home turf; in hotels the machine runs in the daytime housekeeping window. What else has to be settled for unattended night runs is in our night-operation article.

The Whiz is a cobot in the literal sense: it does not build a map on its own, it is shown its route once by a person. A staff member pushes the machine along the desired lane by its steering handle, the robot stores that lane and repeats it autonomously afterwards — up to 600 routes according to SoftBank Robotics. Navigation runs on BrainOS, Brain Corp's platform; on board, per the datasheet, are a LIDAR sensor, a 2D/3D camera and safety sensors that detect and manoeuvre around people, objects and cliffs.

The datasheet states a cleaning productivity of 465 to 557 m²/h and a run time of three hours on a single charge; the launch press release states up to 1,500 m² per charge on a four-hour battery charge. Those figures are consistent with each other: 465 to 557 m²/h over three hours works out to 1,395 to 1,671 m². The dust bag holds 4.0 l, filtration is HEPA — the datasheet states that 99.97 % of PM2.5 particles including pollen and dust are captured. At 30 kg machine weight plus a 5 kg battery, the Whiz is not something anyone carries up a floor on the side.

Two things have to be planned around. First, recharging during a shift means a battery swap. For the European market SoftBank supplies two removable batteries with up to three hours of life each; a fresh battery replaces the empty one without shutting the robot down — but a person has to be standing at the machine for that. Second, starting is manual too. Whiz Connect is a monitoring and analytics application: fleet status, route and location management, SMS notifications, reports. No function for starting a cleaning run remotely appears in any official material. That makes the Whiz a machine for buildings where someone is on site anyway — classically a hotel with daytime housekeeping, or a site with its own cleaning shift.

Part of the picture is the age of the platform. The Whiz has been on the market since 2019, and SoftBank reported more than 10,000 units sold worldwide as of the end of June 2020 — it is the most proven machine in this comparison and is still regularly available in Europe. In May 2026, however, SoftBank Robotics presented a new Whiz generation in Japan (Whiz P, Whiz B, Whiz V); there is no European announcement for it so far. So we sell the Whiz as what it is: proven, not new.

The NR 1700 differs from the Whiz fundamentally in exactly one respect: it has no camera. The official sensor list in the datasheet names a long-range laser distance sensor (LR-LDS), a bumper sensor, eight cliff sensors, two magnetic-field sensors, a wall-follow sensor, incremental sensors and an inertial measurement unit — no optical image anywhere. Mapping runs on what Nexaro calls AI-Mapping, at up to 600 m²/h mapping speed. Where the works council and the data protection officer have a vote, that is not a side note but the argument; the legal side is covered in our data protection article.

Nexaro is a company of the Vorwerk Group based in Wuppertal. The NR 1700 is developed and designed there, and the blower motor is produced at Vorwerk's Wuppertal motor plant. Final production takes place at a Vorwerk-owned plant in China — Nexaro states that openly in its own FAQ, and we state it the same way. "Engineered in Germany" is the accurate phrasing; "Made in Germany" would not be. Customer data is stored on European servers according to Nexaro, and the robot meets Machinery Directive 2006/42/EC and the IEC 63327 standard and is certified by TÜV Süd.

The machine is small: 343 × 343 × 175 mm, 6.5 kg, 280 mm cleaning width, a maximum of 0.3 m/s. It has three modes with 250, 110 and 105 minutes of runtime (Eco, Balanced, Performance), returns to its charging station by itself, and connects to the Nexaro HUB through a built-in 2G/LTE-M mobile module — no customer Wi-Fi needed, though mobile reception at the dock location is. The bin holds 1.4 l; the washable HEPA H13 filter has 530 cm² of filter area. On charging, the official sources contradict each other: the February 2025 datasheet gives roughly 120 minutes for 15 to 95 %, the December 2024 operating manual roughly 195 minutes. We give both figures rather than picking one.

The 100 m²/h from the datasheet is a maximum, and Nexaro qualifies it in its own footnote as dependent on floor plan, furnishing, floor covering and soiling. What survives of it in a furnished office we have written up with our own deployment figures: the NR 1700 in the field puts the manufacturer figure and our practice side by side. Anyone sizing the machine should read that article before doing the area arithmetic.

The decision hangs on three questions: how large and how contiguous is the carpeted area, who is in the building while the machine runs, and when is the machine allowed to drive.

Large, contiguous carpet areas and an operation with staff on site speak for the Whiz. A hotel with long, uniform corridors is the ideal case: housekeeping teaches the route once, starts the robot in the morning window, and swaps the battery when it runs down. The same logic carries in congress and event areas and on office floors with their own cleaning shift. What the Whiz cannot do is run unattended through the night: the shift ends with the battery, and the next one is put in by hand.

Cellular floor plans, small to medium areas and the wish to let the machine work alone at night speak for the NR 1700. It maps by itself, returns to its dock by itself and reports over mobile data — in an office with individual rooms, corridors and meeting rooms, that is exactly the right operating mode. Then there is weight: one person carries 6.5 kg up to the next floor without a lift. And where the camera question blocks the rollout, it is the only answer in this comparison.

When neither profile fits, the reason is usually area size. Above what these two machines sensibly cover, the range of the large dry machines begins — the MT1 Vac, for instance, is officially a three-in-one sweeper, vacuum and dust mop rated at a maximum of 1,400 m²/h, described in its own article. Which class belongs in your building at all is worked through end to end in which cleaning robot fits which operation.

The search term exists, and the honest answer to it is, in most cases: you are not looking for a vacuum robot, you are looking for a sweeper. A hall is dry, large, open and usually finished in concrete or industrial screed. Sweepers are built for exactly that — the two models we operate ourselves are rated at 1,800 and 2,200 m²/h with a 700 mm sweeping width and a 35 l debris hopper, and the robotic sweeper comparison puts six available machines side by side with verified figures.

For the NR 1700 the answer is printed in the datasheet. Nexaro specifies a maximum of 12 × 12 m for free, open areas without distinctive objects and limits the distance from the charging station to 35 m. An empty hall is precisely the geometry that specification excludes: the laser distance sensor needs walls, edges and objects to orient itself against. That is not a defect of the machine, it is its documented operating limit.

For the Whiz it is a matter of arithmetic. At 465 to 557 m²/h per datasheet, a 5,000 m² hall works out to nine to eleven hours of pure driving time — at three hours of runtime per battery, that is three to four battery charges, each swap performed by hand. With the two batteries supplied, six hours of continuous work are possible; after that the four-hour charging time sets the rhythm. It is telling that SoftBank Robotics built a separate, much larger machine for warehouses and factories: the Whiz B, presented in Japan in May 2026 with 40,000 m² per night. It is not announced for Europe.

There is a sensible use case for vacuum robots in hall buildings, but it is not on the hall floor. It is the adjoining areas — supervisor offices, break and training rooms, reception and showroom, often carpeted or exposed to fine dust. In warehouses and in production that is the realistic division of labour: sweeper on the hall floor, vacuum robot in the attached office wing. And when the hall floor carries something that sticks — oil, tyre marks, dried-on soiling — neither class helps, and it takes a wet machine.