October 5, 2026

by

Sonair team

What it takes to make humanoid robots safe around people

Humanoid robots have to perceive people at every height, not in a single scan plane. In an interview with ROBOTIK UND PRODUKTION, Sonair CEO Knut Sandven explains what 3D ultrasonic sensing adds, and what it will take to make it work on humanoids.

Sonair CEO Knut Sandven looking at a screen with a colleague

A humanoid robot doesn't work in a single plane. It reaches, bends and turns among people who do the same. That puts perception at the centre of making humanoids safe to work beside.

In its October issue on humanoid robotics, ROBOTIK UND PRODUKTION asked Sonair CEO Knut Sandven what 3D sensing brings to humanoids, AMRs and industrial robots. Below is a shortened English version.

Read the full interview (in German).

How did you extend safety-rated perception to the full three-dimensional space?

A rotating laser sees one horizontal plane. Anything above or below that plane is invisible to it, and no amount of software fixes that. It is a limit of geometry, not of processing, so it had to be solved in hardware. Sound does something light cannot. A laser traces a single line through a room. A pulse of ultrasound expands to fill the whole volume, and every surface it reaches sends part of it back. Nothing in the room goes untouched, which is exactly the property a safety function needs.

We developed tiny MEMS ultrasonic transducers, placed closely enough together to work as one array. The principle is similar to medical ultrasound: send out sound, capture the echoes, build an image. Doing it through air rather than tissue is the hard part. Air is a far more challenging medium for ultrasound. Almost all the energy is lost at the boundary, and what does get through attenuates quickly. Producing a usable 3D image from those echoes, in real time and to a level that satisfies a notified body, is what took us years.

SIL 2 / PL d wasn't added at the end. Safety requirements shaped the design from the start, from how the elements are arranged to writing the firmware in Rust, since IEC 61508 and ISO 13849 require deterministic real-time behaviour.

Where does 3D ultrasound add the most value over 2D laser scanners?

People, first of all. A scanner mounted at shin height sees legs. It does not see an arm reaching into a path, someone leaning over a conveyor, or a person crouching to pick something up. ADAR One sees the whole body, including whichever part of it is actually closest to the machine. The same geometry applies to low and overhanging obstacles. Forklift tines, pallets, dropped tools, low shelving, and overhanging loads all fall outside a 2D scanner's plane. With 180×180° coverage, ADAR One detects people and objects across their full height.

Sound also behaves differently from light. Dust, moisture, steam, glass and lighting conditions defeat cameras and LiDAR in ways they do not defeat ultrasound. A glass partition returns a sound echo; to a laser it is close to invisible. Cleanfix, the Swiss manufacturer of professional cleaning machines, was the first to put ADAR One into a commercial robot. More than 100 companies are now testing the sensor. Cleanfix saw fewer false stops in dusty and wet cleaning environments. On the Cleanfix RA660 Navi XL, a single ADAR One replaces 14 separate linear ultrasonic sensors, which means less cabling, fewer failure points and true 3D coverage. It can also be used in stationary robot cells to create distance-based, dynamic safety zones.

How does 3D sensing improve human-robot collaboration and fleet operation?

It moves you from avoidance zones to real spatial understanding. A fleet that sees one plane has to work with large margins and abrupt stops, because an overhanging object is not something it can see at all. It can only assume something might be there. Full 3D perception lets robots move more smoothly and predictably around people, with fewer unplanned stops.

ADAR One also allows gradual responses. A robot can slow down instead of stopping outright, because it can tell an object in a warning zone from a real intrusion into the stop zone. ADAR One supports a stop zone and warning zones across up to 128 preconfigured zone sets. That makes robot behaviour easier for people to read, and more consistent across larger fleets. People work around machines they can predict. A robot that slows as you approach and resumes as you step away is one that operators understand, and that is where the productivity actually comes from.

Which 3D sensing approaches are most promising for humanoids?

Cameras and ADAR together, and the division of labour between them is clear.

Cameras and visual SLAM are very good at understanding an environment: what an object is, where the robot is, what it should do next. They are poor at guaranteeing that a person has been detected, because an optical system fails on glass, on dark clothing, in low light, and in ways that are hard to quantify for a safety case.

ADAR does not interpret the scene at all. It reports what is physically present in the volume around the robot, deterministically, to a certified failure rate. Let the cameras understand the world and let ultrasound handle safety.

Functional safety matters here too. ADAR One is currently the only one of these approaches available as a certified, ready-to-use 3D sensor. Building with uncertified components, the robot maker has to prove safety at system level themselves. That takes time, cost, functional safety expertise, and changes to the safety architecture can trigger recertification.

What makes real-time processing of 3D sensor data hard?

Cameras and 3D LiDAR generate large volumes of data, which is often downsampled or compressed on the device. That risks losing exactly the details that matter.

For a safety function, we don't think that works: you cannot take the risk that the one detection that should have stopped the robot was the detail the compression discarded.
Ultrasound avoids the problem by design.

ADAR One delivers a sparse point cloud focused on the nearest points, the ones that matter for safety, so there is little data to process in the first place. It also does not need to recognise anything. A perception model has to have seen something resembling an object before it can be relied on to react to it, which is a difficult property to defend in a safety case. ADAR measures distance to whatever is there, whether or not it has encountered it before.

What will make 3D sensing cheaper and more reliable for humanoids?

It's a chicken-and-egg problem.

Humanoid production is still in the thousands, not millions. As volumes grow, sensing becomes a smaller share of total component cost. Because our sensors are MEMS-based, the manufacturing process is inherently a volume technology, consistent and scalable.

Form factor is the second lever. MEMS transducers are small, and that is an advantage no competing safety technology has. A certified safety laser scanner is a mechanical instrument with a rotating mirror, and there is a floor on how small it can be made.

An ultrasonic element has no moving parts and can shrink to millimetre scale. Longer term, that allows many small elements distributed across a humanoid's outer shell rather than one large module, giving all-round coverage without asking the designer to find room for a box.

‍

First published in German in ROBOTIK UND PRODUKTION, issue 5 (October) 2026, in the magazine's humanoid robotics focus section, and online on 25 September 2026. Questions by editor-in-chief Frauke Itzerott. Translated, edited, and shortened by Sonair. Read the original article.