Knowledge

Guide · · 11 min read

Cleaning Robots at Night: What Unattended Operation Really Requires

For most buildings the night is the obvious cleaning window: empty floors, no obstacles that wander, closed doors, and a map that does not change between two runs. That is why the phrase "night operation" appears in almost every cleaning-robot quote — and why what actually hangs on it is so rarely spelled out. Unattended operation is not a property a machine automatically has because it drives autonomously. It requires a charging logic that carries further than one battery, a path by which a fault at three in the morning reaches a human being, an arrangement with the alarm system, and an honest answer to the noise question. This article works through those four points, sorts them by machine class, and ends with a checklist you can run against a building before you order anything.

Key takeaways

  • The night wins through stability, not through silence: an empty floor, a constant map, and no obstacles that move between two runs.
  • Noise is a question of machine class. Where a value is published for an autonomous scrubber-dryer in our fleet, it sits at or below 70 dB(A); for robotic sweepers not one manufacturer publishes a dB(A) value; among vacuums the published figures differ sharply from machine to machine.
  • Autonomous is not the same as unattended. What decides it is the return to the dock, the resume at the interruption point, and a route by which a fault reaches someone — the SoftBank Whiz, for instance, is started by hand.
  • The building decides too: alarm systems and motion detectors, doors and lifts, and whether anyone sleeps within earshot.
01

Why the night is the natural window — and what it is not about

What the night actually delivers is stability, and stability is what an autonomous machine converts into coverage. These robots drive a learned map and correct it live against their sensors. Every obstacle that was not there yesterday costs a detour, a re-plan, and in the end an uncleaned patch: a pallet parked in the aisle, a roll cage in front of the shelf, a chair pushed out. At night the floor is empty, the doors are closed, and the map from the last run still fits. That is worth more than any headline area figure — how the classes and the basic mechanics differ is covered in our introduction to cleaning robots. The second reason is just as banal: nobody walks across the floor while it is still drying.

Then comes the question every quote answers too quickly. For the autonomous scrubber-dryers in our fleet there are documented values, and wherever one is published it sits at or below 70 dB(A): Pudu states below 70 dB(A) for the CC1, the official EU datasheet for the CenoBots L4 gives a sound pressure level of 65 dB, and the CenoBots L50 user manual states 67 dB(A). Two caveats belong next to that. For the CenoBots L3 the available sources contradict each other across a range of roughly 60 to 66 dB(A), so we publish no figure for it. And for the CC1 Pro, Pudu contradicts itself: the official product deck of May 2025 says below 70 dB, the current official German product page says below 75 dB — never treat the Pro as the quieter machine, as our CC1 versus CC1 Pro comparison explains. The SH1 drops out of this discussion entirely: it is a walk-behind machine that always has an operator, so it is not a night machine in the first place.

For robotic sweepers, the honest answer is that there is none. Not one of the four manufacturers behind the six sweepers in our sweeper comparison publishes a dB(A) value for its machine. Figures circulate on dealer pages, but a shop listing is not a measured datasheet value. In practice that matters less than it sounds, because this class lives in halls, logistics and production, where the relevant window is night or the gap between shifts anyway. It matters a great deal the moment somebody wants to put a sweeper into a quiet zone.

Vacuums are the class where the answer is decided per machine and per mode, and where "loud" is too coarse a word. The Nexaro NR 1700 is documented at 49 dB(A) in Eco mode and 57 dB(A) in Balanced and Performance; the SoftBank Whiz datasheet states 62 dB in normal mode; for the Pudu MT1 Vac, Pudu publishes no value at all. What follows is a placement rule rather than a decibel rule: offices at night are unproblematic because nobody sleeps there, and hotel guest floors are not, because a machine passing bedroom doors at three in the morning is a complaint mechanism no matter what its datasheet says. Which vacuum belongs where is the subject of our vacuum robot comparison.

CC1
NR 1700
CC1Scrubber-dryer, below 70 dB(A)NR 1700Vacuum, 49 dB(A) in Eco mode
Pictograms: vectorized 1:1 from our product reference photos — not illustrative icons.
02

Power, endurance — and who starts the robot

Start with the arithmetic, because it is the part that gets skipped. Pudu states runtimes per mode for the CC1: five hours of scrubbing, up to nine hours of silent mopping, four hours of carpet vacuuming, and under three hours of charging. A venue that closes at 22:00 and opens at 06:00 has an eight-hour window. Five hours of scrubbing do not cover eight hours on one charge — and they do not have to. What covers it is the loop Pudu documents officially: the robot returns to the charging station on its own when the battery runs low, remembers the location of the interruption and continues the task after charging. Without a docking station that loop does not close, and Pudu lists the dock as optional equipment. In our deployments it ships wherever the use case needs it; the full-service model is built exactly around that kind of question.

The Nexaro NR 1700 is documented for the same loop — the datasheet states automatic recharging explicitly — with runtimes of 250 minutes in Eco, 110 in Balanced and 105 in Performance. One detail matters more at night than during the day: the machine connects over its own 2G/LTE-M mobile link with global roaming, not over the building network. Nobody has to release a corporate Wi-Fi at two in the morning. The corresponding caveat comes from the manual, which requires mobile reception at the docking location, so this is a question for the site survey rather than for the contract. The unit also carries a warning light, which is a practical detail in a dark building.

The counter-example is the more instructive case. The SoftBank Whiz runs three hours on a single charge and has a swappable battery — and its datasheet contains no docking station at all. SoftBank's datasheet documents SMS support delivering near-time texts of "assist notifications" from specific robots; that these also report the start and the end of a route is described in the Whiz Connect portal glossary rather than in the datasheet. No remote start appears anywhere in the official material: the run is started at the machine. A robot that a person has to start and whose battery a person has to change is autonomous during the run, but it is not unattended across a night. That distinction is worth testing in every quote, because both properties are sold under the same word.

Two machines document the night case particularly clearly — the Kemaro K900 Gen II and the Adlatus SR1300, which is available through us but newly onboarded. We say plainly that what follows is a reading of their documentation. Kemaro states "automated round-the-clock operation with configurable time settings" for the K900 Gen II, up to four hours of runtime against a 90-minute charge, and one property that is genuinely a night property: the machine needs no lighting. Adlatus writes that the SR1300 starts its cleaning programmes fully autonomously via a time schedule on any weekday and hour, and pitches operation outside working hours as the normal case.

03

When something goes wrong at night — and what the alarm system says about it

Every autonomous machine eventually meets something its map does not contain. A pallet parked in the aisle, a cable across the route, a door that used to be open. At night nobody notices, so the useful question is not whether the robot can get stuck — it can — but what happens in the next few minutes. Four examples of what the manufacturers document: the NR 1700 reports into the Nexaro HUB with dashboard and cleaning reports, the Whiz by SMS alert, the Kemaro K900 into a web app the manufacturer describes as real-time monitoring with tracking and reporting, and the Adlatus SR1300 as an automatically generated cleaning protocol emailed after the run. Our machines report into our monitoring, and in the full-service model reacting to that report is our job, not yours — that is the part of our service that only ever shows up at three in the morning. What none of it replaces: somebody with a key has to be able to reach the machine in the morning.

The alarm system is the conflict that gets discovered too late. A robot moving through an armed building is precisely what a motion detector is built to notice, and there are only three practical answers. Schedule the cleaning window outside the armed hours. Exclude the cleaned zone from arming. Or use a system that genuinely talks to the alarm — and in this field exactly one manufacturer documents that: Adlatus writes that the SR1300 can be integrated into IoT systems and enter a kind of dialogue with fire-alarm or intruder-alarm systems. That statement comes from the manufacturer datasheet. For everything else the answer is scheduling, and it belongs settled with whoever is responsible for the alarm before the first night, not after the first false alarm.

Doors, gates and lifts set the last boundary, and it is a hard one: a machine cleans the floor it can reach. Without a lift interface, a multi-storey building means either one machine per floor or a person who carries it. What is documented: Pudu lists intelligent lift control for the CC1 and, for the CC1 Pro, lift control plus electronic gate control as IoT compatibility, and Adlatus states that the SR1300 can operate automatic doors and gates. Two points of caution belong with that. An interface is a project, not a checkbox — it has to be built with whoever maintains the lift. And for every machine where the manufacturer documents nothing, assume it opens nothing and plan the route inside one accessible, lockable area.

04

Four building patterns — and one exception

In supermarkets the night is the cleanest case there is. After closing the sales floor is empty, the aisles stand where they stood yesterday, and the window between closing and the first delivery is defined to the minute. The limiting factor is not noise but goods: pallets staged overnight for the morning shift change the map, and anything staged in the aisle is a detour. The real gain is the one the day cannot deliver — a full-coverage pass across the whole sales floor without a single customer in it.

In offices the night is where the vacuums belong. This is the case the noise objection does not survive: nobody sleeps in an office, so a documented 49 dB(A) in Eco mode on the NR 1700 is not even the decisive argument — the decisive arguments are access and the alarm system. The typical window starts late in the evening and is finished long before the first employee arrives; carpet and hard floor in the same building usually mean two machine types rather than one compromise.

In warehouses the correct term is not "night" but the gap between shifts. Dry sweeping fits it exactly, because the floor is driveable the moment the machine has passed — no drying time blocks the next shift. Kemaro documents that the K900 Gen II needs no lighting, which makes a dark hall a legitimate operating environment rather than a compromise. Do the arithmetic before you promise anything, though: against four hours of runtime and a 90-minute charge, an eight-hour gap holds one full run, a charging pause and a second run that gets cut short. Two complete cycles need nine and a half hours, not eight — two full runs mean either a longer window or a second machine. Those are the manufacturer’s own figures, and how the class compares is in our sweeper comparison.

Hotels are the exception, and it is the most important one in this article. Public areas at night are a good fit for scrubber-dryers: lobby, restaurant, ground-floor corridors are closed, empty and free of sleeping guests. Guest floors are not, and the reason is not a datasheet prohibition — it is that a machine passing bedroom doors at three in the morning generates complaints regardless of its published value, and that for one machine in the vacuum class no value is published at all. Guest floors are a daytime job, inside the housekeeping window, and that is exactly where a machine that runs visibly during opening hours earns its keep.

05

The day is not the rival of the night

Night operation and day operation are usually presented as alternatives, and they are not. The night does full coverage: every square metre, in a defined window, without interruption. The day does the things full coverage cannot do — the spill that happens at eleven in the morning, the entrance dirt on a rainy afternoon, and the visible signal that the floor is being looked after while people are standing on it. A building that needs both does not choose between two shifts; it chooses machines whose day capability is documented.

For the scrubber-dryer class, Pudu documents that capability precisely: the CC1 Pro carries an AI spot-cleaning mode that patrols at 1,500–3,000 m²/h, detects dirt visually and cleans exactly there. That figure is a mode, not a faster machine — in full-coverage cleaning the Pro is rated identically to the base CC1, which is the single point most dealer pages get wrong and which our CC1 versus CC1 Pro comparison takes apart. The base CC1 has no such mode; its daytime tool is the silent mopping mode with its nine-hour runtime.

The practical consequence is a sequencing decision rather than a machine decision. Put the full-coverage pass in the night window, where stability is highest and interruptions cost nothing, and keep the day for targeted response — either through a machine with a documented spot mode, or through the cleaning staff you still have. Which class covers which of these two jobs in your building is the question our guide which cleaning robot fits which operation walks through from the beginning.

06

Checklist: is your building night-ready?

The window first, and honestly. When does the building actually empty, and when does the first person come back? The official closing time is rarely the answer — deliveries, cleaning staff, security rounds and night shifts all sit inside the supposed window. Write down the real start and the real end, then subtract the time the floor needs to be walkable again. Scrubber-dryers leave a thin drying film, so the pass has to end before the first person arrives, not with them.

Then power and reach. Is there a place for a docking station with a permanent socket, inside the cleaning zone and reachable without steps? For a scrubber-dryer, add the water question: a dock with automatic water exchange needs a supply and a drain, otherwise the mobile water tank takes over. Does the route lie on one level and inside one lockable area? Every door that has to be opened during the run is a person, and every person is the end of unattended operation — unless the machine documents lift or gate control, which most do not.

Then the building systems. Which zones are armed at night, who arms them, and can the cleaning window be scheduled around the arming? Are there motion detectors inside the zone to be cleaned? This is a conversation with the person responsible for the alarm, and it belongs before the first night. Clarify at the same time who receives a notification at three in the morning and who can be in the building at six if the machine is standing somewhere with a fault.

And finally the two questions that decide placement rather than technology. Does anyone sleep within earshot of the route — and if so, does the machine class have a published dB(A) value at all? If the answer to the first is yes and the answer to the second is no, that route does not get a night slot. Second: what changes the map between the run and the morning? Pallets staged for the early shift, chairs moved, goods parked in the aisle. That is not a reason against night operation; it is the reason why a night route is worth defining precisely once instead of being improvised nightly. We walk this list through the building in a site assessment before anything gets ordered.

07

Frequent questions

Yes, if three conditions hold. The machine has to close its own energy loop: Pudu documents for the CC1 that it returns to the charging station on low battery and resumes the task at the interruption point after charging, and the Nexaro NR 1700 datasheet states automatic recharging. There has to be a path by which a fault reaches a human being at night. And building access has to be settled, because a machine that cannot open a door cleans only what it can reach. The counter-example shows what the word hides: the SoftBank Whiz is started at the machine, has a swappable battery, and no docking station appears in its datasheet — autonomous during the run, but not unattended across a night.

That depends entirely on the machine class. For the autonomous scrubber-dryers in our fleet, wherever a value is published it sits at or below 70 dB(A): Pudu states below 70 dB(A) for the CC1, the CenoBots L4 EU datasheet of 2024 gives 65 dB, the CenoBots L50 manual 67 dB(A). For robotic sweepers there is no answer at all — not one of the four manufacturers behind the six machines in our sweeper comparison publishes a dB(A) value. Among vacuums the figures diverge: the Nexaro NR 1700 is documented at 49 dB(A) in Eco mode and 57 dB(A) in Balanced and Performance, the SoftBank Whiz at 62 dB in normal mode, while Pudu publishes no value for the MT1 Vac. Where a manufacturer publishes nothing, we state nothing.

A machine moving through an armed building is exactly what a motion detector is built to notice, so this has to be settled before the first night rather than after the first false alarm. Three practical answers exist: schedule the cleaning window outside the armed hours, exclude the cleaned zone from arming, or use a system with a documented alarm interface. Among the machines in our sweeper comparison, only Adlatus documents the third option — the SR1300 datasheet states that the machine can be integrated into IoT systems and enter a kind of dialogue with fire-alarm or intruder-alarm systems. Future Machines offers the SR1300; that statement comes from the manufacturer datasheet. For all other machines the answer is scheduling, agreed with whoever is responsible for the alarm system.

It stops and reports. Each of these machines has a documented reporting channel: the Nexaro NR 1700 into the Nexaro HUB with dashboard and cleaning reports, the SoftBank Whiz by SMS alert, the Kemaro K900 Gen II into a web app the manufacturer describes as real-time monitoring with tracking and reporting, the Adlatus SR1300 as an automatically generated cleaning protocol emailed after the run. Our machines report into our monitoring, and in the full-service model reacting to that is our job rather than yours. What none of that replaces is physical access: somebody with a key has to be able to reach the machine in the morning, which is why access is part of the night-readiness assessment and not an afterthought.

Usually not, and that is not the problem it sounds like. Pudu states runtimes per mode for the CC1: five hours of scrubbing, up to nine hours of silent mopping, four hours of carpet vacuuming, with charging under three hours. The Nexaro NR 1700 datasheet gives 250 minutes in Eco, 110 in Balanced and 105 in Performance. A night is covered not by one charge but by the loop of return, charge and resume at the interruption point — which is why the docking station is the component that decides whether the night runs, and why machines without that loop need a person in the building.

08

Sources

See it instead of reading?

We show you the machines in 30 minutes at your site — free and non-binding.

Request a demo
Pudu CC1View machine →Nexaro NR1700View machine →