Cliff Sensors on Compact Floor Robots: A Store Checklist

How does a compact floor robot detect a drop?

A compact floor robot finds a ledge or stair edge by watching the floor beneath it. Most of these sensors are downward facing infrared emitter and receiver pairs. A strong return means floor is present, while a weak or missing return signals a drop, and the controller reads that as an edge.

Some designs skip light entirely. An LG patent describes a drop off detector with a contact bar on the floor, where lack of contact counts as a drop, and the stated aim is to avoid optical false drops on dark or reflective floors. A patent describes an approach, not how any scrubber performs, so read it as background.

For a store, the practical point is plain. A robot can report an edge only as well as its sensors read the floor in front of it. As of October 2026, the commercial listings we reviewed mention cliff sensing briefly, and few give a drop height, a test surface or a result.

Where do optical cliff sensors get fooled?

Consumer support material from eufy and iRobot lists the same main false drop causes: dark or reflective surfaces, rugs over stair edges, and dust on sensor windows. A dark mat can read as a drop and stall a route. A rug laid over an edge can hide the drop from the sensor altogether.

In a store these conditions are routine. A black rubber entry mat sits at a threshold, polished concrete throws glare down an aisle, and a rug covers a step into a stockroom. A contact bar avoids the light problem but adds a moving part that must touch the floor and stay clear of debris, so it trades one failure mode for another.

One iRobot support page gives a minimum cliff height of 3.5 inches (90 mm). We read that figure only from a search snippet, and it describes a household model, so use it as a prompt. A commercial scrubber has its own sensor layout and floor conditions, and the figure does not carry over.

What do vendors actually publish about cliff sensing?

Commercial listings vary widely. A listing that names a sensor tells you what hardware exists, but not the drop height the robot catches, the floor finish it was tested on, or how often it misses. Four listings, checked as of October 2026, show the range:

• CenoBots L4: a 32 beam 3D lidar, 3D depth cameras and bumper sensors, plus a cliff heading whose details did not appear in the copy we could read.
• Americhem distributor page for one autonomous cleaning robot: cliff detection listed beside protruding and low object detection, and a minimum edge distance of as close as 2 inches from walls for one model. That edge figure concerns walls, not drops, so it does not answer the cliff question.
• Gausium Scrubber 75: more than 20 sensors around the body, with no named cliff sensor in the copy we could read.
• Ecovacs commercial floor scrubber page: obstacle detection, cliff sensors and an emergency stop, with no specifications.

Which sensor questions belong on the vendor checklist?

Each listing above is a reason to ask, not an answer. Put these questions to every vendor in writing, because a vague reply is information too.

• Where are the downward sensors mounted, and do they read the floor ahead of the wheels or only beneath the body?
• What drop height is the cliff function tested to, on which floor surface, and where are the test records?
• How does performance change on black rubber, glossy tile, sealed concrete and carpet edges?
• How does the robot detect a dusty or wet sensor window, and does it stop when a window is blocked?
• Is there a contact based or redundant drop detector, and what happens if one channel fails?
• Are stairs and step downs handled the same way as ledges, and can a dock edge be mapped as a no go zone?
• Can we run our own drop test on our floors before go live, and will the vendor support it?

Which drops hide in a real store?

Stores rarely offer one clean edge. Expect threshold strips at entries, a step down into a stockroom, a raised display platform, a loading dock apron and the lip of a charging dock. Each one needs its own entry on the map.

Dock edges carry the most weight. A floor robot working near a loading dock can approach the apron at an angle the cliff sensor never sees coming, and a raised charging dock can catch a robot in mid approach. Map the dock edge as a no go zone rather than trusting a cliff sensor to catch it in time.

Overnight, the risk grows. Nobody is present to notice a rug curling at a threshold or a display platform moved after the last map was made. Pair the drop test with the route map, and recheck the map after every layout change.

What test steps should run before unsupervised work?

Treat the drop test as a site test. Run it on your floor and your surfaces, and log every result. Teams often skip the dark surface and dusty window cases, yet stores tend to produce both.

This is where one integrator can cut the back and forth. Service Robot Co. works as a vendor neutral robot integrator, so we compare sensor listings across manufacturers before we recommend a unit. Our site assessment maps the drops and dock edges described above, and our go live support runs the tests on your floor. The offer is a robot lease with one fixed payment, quoted after that site assessment on a three year lease, so the test results shape the quote instead of arriving after it. Many facilities start with one leased unit and a supervised pilot before they widen the route.

• Map every ledge, step, threshold, display platform and dock edge, and mark each as a no go zone or a tested drop.
• Drive the robot slowly toward each edge, forward and in reverse, and log whether it stops, turns or slows.
• Repeat the pass on black rubber, glossy tile, sealed concrete and carpet at the same edge.
• Lay a rug across a step down in a controlled test, and confirm the robot still stops at the drop beyond the rug edge.
• Dust the sensor windows with the debris your building makes, clean them the way staff clean them, and rerun the edge tests.
• Run a supervised shift and read the log for drop events, false stops and unexplained turns before any unsupervised run.
• Record a go or no go decision for each zone, with the reason written beside it.

Does a safety standard cover drop sensing yet?

UL 3300 covers service, communication, information, entertainment and education robots. According to UL's 2020 outline announcement, the standard classifies robots by height and mass, sets speed limits, requires object detection and avoidance, and calls for audible and visual indicators for vulnerable users.

The public outline does not mention a cliff or stair requirement, and the snippets we could see did not show one. The full text is sold by UL, so we could not confirm what it says about drops. Silence in a summary is not a pass, so ask the vendor for test records against the standards that apply to its product.