The accelerating rate of robot deployment in manufacturing is obvious. All sectors are growing—robots, cobots, AMRs, even humanoid and cute doggyoid robots have been videoed while working in manufacturing environments. In order to move about, these robots need to “see,” and in many cases lidar (light detection and ranging) devices help them do this by reflecting light. There are limitations, of course, and we’ll discuss a couple during our interview with Knut Sandven, the CEO of Sonair (Oslo, Norway). Sandven and his team are devoting their energies toward ADAR, acoustic detection and ranging, a new way of navigation and safety for robots like AMRs that move (and machines that don’t, but need the extra avoidance safety brought by ADAR). Come and read along and get to know this entrepreneur, his company, and their goals.
Fifth Wave Manufacturing: We call this series The Story of Making Tomorrow. And you and your company certainly fit that description. Today we’re speaking with Knut Sandven, and he is the CEO of Sonair. Sonair is a company in Norway that has really taken a break from the conventional. We talk about non-visual imaging sometimes, and that is a fitting description. It takes their technology of ultrasound and uses that to see what’s happening in the floor of a manufacturing plant or a warehouse, and avoid collisions. And when that becomes very important, I think the more robots and cobots are accepted, the more people worry about being plunked by one of those things. It’s a valid concern. They’re <robots are> very strong. Some of the robots are able to lift a car. Knut, welcome. Thank you so much for taking some time with us, and if you could maybe give us a bit of your background, then we’ll get into the questions.
Knut Sandven: Thank you, David. It’s a pleasure being here and a pleasure meeting you. Thank you for the introduction of of Sonair, the company I’m currently building and the technology we’re building, which we’re calling the ADAR, Acoustic Detection And Ranging.
Sonair is really my third adventure in the startup world. I built two companies previously, and what’s nice about Sonair is that I’m building this company together with people who were also part of my first company. In my first company, we made safety certified gas detectors for oil and gas installations. That company Company was acquired by by the German company, Dräger, after only five years.
The gas detectors we developed used combined detection principles. They used both ultrasonic or acoustic detection and optical detection, and what we learned from acoustic there we have brought into Sonair. More specifically, we are using what’s called a MEMS transducer. MEMS is Micro Electro-Mechanical System. It’s a silicon-based sensor. That is the core of our technology platform, this MEMS transducer. You can compare it to the parking sensor you have on your car. You see those button-sized sensors on the bumper. Our sensor has the same performance, more or less, but it’s millimeter-sized. This has allowed us to build a hardware platform where we can do not only ranging in the form of distance measurements, and we can do what you called imaging. But it’s not really imaging—and I’ll get back to that—but we can also do localization so we can very accurately position objects in 3D space.
In its simplest, easiest to explain form, we detect all kinds of objects in front of the sensor. We give it the distance to the object and also the location in a coordinate system.
FWM: So this is really meant for a robot that moves, like an AMR, in an environment that changes. With some cobots, you drag them through all of the turns and twists that they’ll need to do, and you more or less show them their universe, and that universe doesn’t change. But in a very dynamic environment like we have today, where you may even move a robot from one task to another or have it move itself from one task to another, this becomes hugely important.
Sandven: You’re right. I would say both cases or application use cases are important. In the static installations where you have a robot arm, for instance, doing its operations, you need to detect if people are approaching, and then the robot arm needs to slow down or stop. It doesn’t need to be a robot arm; it can be a welding machine. It can be a copy machine for, for that matter. For any dangerous machinery where you need to detect if people are approaching, this technology is really ideal. That’s also true for moving machinery—autonomous robots, machinery, any vehicles moving at slow speed. Ultrasonic is not good at high speed, but at slow speed, up to five or six meters per second, it’s fine.
It is really good for object detection. And the advantage we have is that acoustic signals at the frequencies we are using, make all objects appear more or less equal. So if you’re looking at a soft object, a glass object, a metal object, they all look the same in in the ultrasonic signature. And that’s an advantage when the only thing you want to do is to detect if there’s something there. The disadvantage is that it’s difficult to use our signals or the signals from the sensor to understand what you’re looking at. You will see that there’s something there, but it’s difficult to understand what it is.
FWM: But you will be able to see if it’s moving or static.
Sandven: That’s right, we do. And we can measure the the velocity. So yeah, we see the movement of all objects very well.
FWM: Maybe this is getting ahead of the game and I’m sorry if I am, but can it calculate a trajectory for something moving?
Sandven: Yes it can. We haven’t implemented that yet, so I’ll not highlight that too much, but it certainly can. Yes.
FWM: Okay, good. Every worthwhile innovation comes from problems. What did you see as a problem, and how did that become a solution to use ultrasound?
Sandven: To be honest, we started off a little differently than where we are today. What we set out to do in the beginning was to make what we called room sonar. Place a sensor in a room and detect people and their movement. For instance, in a meeting room where we’re sitting now you could quite easily see where are people sitting, how many people are in the room, or in a store what’s the flow of people? Where are people moving? How many people are there? It replaced cameras for privacy reasons. That was our initial idea. We started doing that and we quickly developed a sensor that was really good at detecting people.
And then a little by coincidence, we came into or were moved toward robotics. The team we have at Sonair, we have developed sensors for safety applications previously, as I stated. When we learned what robots are using currently for detecting people safely, so reliably in all circumstances, that’s where we really understood where the big opportunity was because we believe we can do that so much better than what’s on the market today.
FWM: They are both huge markets. One is kind of consumer driven, and you have manufacturing. I fully expect by the time I retire at the age of 90 that I won’t recognize manufacturing because there will be so many robots and so many pieces of automation that don’t just do jobs, but are looked at holistically. In order to look at it holistically, we still need people in there. So all of that has to come into the equation, and it’s a perfect canvas on which to paint your solution because of the constrictions, but also all the different levels of needs from each piece of manufacturing. I think that’s a great place to be.
Sandven: It is. The team is used to working with industrial applications. We know that space and that’s also important making things robust and reliable, fit for industrial environments. But I certainly agree with you very much. Robotics is now coming at full speed and it’s changing, right? Because it’s now coming from a place where, really, the environment or the warehouse is designed for the robot. You’ve done lots of adjustments to make the robots work, but that’s changing. It’s expected now that robots move anywhere and can work with people not trained for working with robots.
Then you need the safety precautions in place. And and working with safety can be quite complicated. It’s a rigorous system. Our theory is, make safety simple—safe, but simple.
FWM: That really comes across on your website when you land on the website. Your marketing team deserves a “way to go, guys.”
Sandven: Yeah, which is great! We’re Norwegian. We are known for being quite open and honest. What you see on the website is what you get. That’s what we have.
FWM: It seems very much that way. Also it reminds me that still, probably the number one most important concern of people is safety. You know, can we be safe with these metal objects moving around and throwing their arms around?
Sandven: It’s a big discussion in the robotics forums these days, how to make robots safe. It’s certainly on the agenda. We launched our product as an evaluation kit a year ago. And the interest we got from all over, really, I’ve never seen anything like it. We had difficulties handling that interest with a small team. But it really confirms the need for new solutions in this field. We feel that we kind of hit the nail on the head and have something the market needs and wants.
FWM: Absolutely. I would agree with you 100%. Is most of the technical work for 1.0, so to speak, done now?
Sandven: Yes, we are now shipping product versions. That means we have sensors ready now for integration into robots. We started shipping those in July and have some customers already having designed their sensors, so integrated, fully flush mounted into their robots. That’s a big achievement for us. I can’t give any names yet, but it’s really cool to see our sensors in robots.
We haven’t finished the safety certification yet. But we are so close now we are through with with all the milestones; we have the final assessment left. We have booked time for that; it’s coming up soon. And then it’s paperwork from the assessor side that’s remaining.
FWM: Here in this country that would be people like UL, right, Underwriters Laboratories.
Sandven: Yeah. We are working with a German safety assessor called EDA. They are part of a group called TÜV.
FWM: Yes, good. Will this roll out where you don’t have to sell it, will it be sold as something that’s embedded in the robots? So it’ll be an OEM type of arrangement?
Sandven: For the robots, it’s an OEM type of product. For machine safety, it’s more an add-on possibility. The machine safety is more mounting on the wall type of installations.
FWM: Okay. And is that something you would have integrators do, or would an end-user be able to do that?
Sandven: For machine safety, we will typically have integrators. It’s usually a part of a safety assessment of the whole machine installation. The integrators will work with different types of sensors like safety safety sensors, what sensor fits a certain purpose, and do a full assessment of the whole installation.
FWM: I could see a whole family of products.
Sandven: Yes.
FWM: Also, you have some very mature integrator groups. Like I know Universal Robots <and others> has bunches of integrators, and they have to qualify by doing the integration work. There are a few ready audiences for you, so that you don’t have to develop them yourself.
Sandven: Yes, we started talking to some of them. We are looking for some global partners now. Of course, we will also do direct sales, especially to robotics, to the OEMs. We are setting up that in Europe for now, but but also having very good dialogues with several US companies both on the west coast and in the Boston area.
FWM: Okay, and we don’t want to give away all the secrets. And in fact, I’m sure you’re patented in all the right places, right?
Sandven: Yes, and we’re still working on getting all the patents in. It’s a long process.
FWM: Well, yes it is. Can you, again without giving away your company secrets, can you just give us a kind of technical explanation that is like a morning in the life of a robot, and here’s what it’s doing. Here’s what’s happening with our hardware and software to help it.
Sandven: Seen from the robot.
FWM: Seen from the robot, yes.
Sandven: Okay. That’s a good challenge. If I was to design the sensor package for a robot, I would give it a camera and an ADAR sensor. It would have an ultrasonic 3D safety sensor, and it’ll get a 3D camera. It will get a point cloud from the 3D camera, and it’ll get a point cloud from the sensor. Each point has a distance and location in a coordinate system. And the point cloud from the ADAR sensor, it has a full 3D view. It gives an overview 180 degrees up and down and to both sides in front of the sensor.
It’s a full hemisphere you see in front of the sensor. Then the robot will set up what’s called its safety zone, and that’s the area where you don’t want objects to be present, or especially not people or a body part. If you have something entering into that safety zone, there will be a trigger in the robot control system that will trigger a stop. It will go directly to the brakes of the robot, and the robot will stop immediately.
That’s really how it is, it’s a very good combination with a 3D camera, because as I said, the sensor will not tell you what you see. If we get the point cloud from a person, we will give you all the points, but we will not tell you it’s a person. And then you can use the the image to understand if it’s a person or a box you’re looking at.
FWM: Okay. Yeah. You’re doing spectacularly well, so go right ahead.
Sandven: We could talk about this for a long time! The robot also needs to to navigate, so to get in input for navigation, currently very often the point cloud is from the lidars. It’s also used for navigation. We very much believe the cameras will take over that function and that’s that’s also our mission to really replace the functionality of the lidars. So the lidars are not needed in the mix anymore.
FWM: Can we pause right there? Because lidar is used everywhere. It’s used in so many places for doing things like this. This is a great opportunity for you to say, this is why our combination of things is better than lidar and maybe cheaper and use less electricity than lidar as well. I don’t know.
Sandven: What’s being used for safety on the robots are the 2D lidars. They see only a horizontal plane in front of the robots. It’s like swiping a laser pointer, that’s what the lidar is. That’s normally good enough. But there are also objects that laser pointer will not see. That’s where we have the biggest advantage using ADAR, that we have a full 3D view. It’s safety in 3D compared to 2D. Also lidar has known weaknesses of not seeing glass, for instance, nor shiny objects. That is also no problem for ADAR. When you go from lidar navigation to camera-based navigation, ADAR is really a better match to the camera view than the lidar is.
Then you have the the 3D lidars, really cool technology that can give you a dense 3D point cloud. They’re currently not safety rated, so you will need a safety sensor in addition. They also need to be mounted high up on a robot. On an AMR for instance it’s not the form factor that is suited for a 3D lidar. For an AGV for instance, having a 3D lidar on top, it gives you very good good input, but still you need a safety sensor. There’s there won’t be one solution, but I think you have a quite competitive one with ADAR and cameras in combination.
FWM: That’s great. Wonderful. thank you for that description too. You did a perfect job on that, <laugh>.
Sandven: Thank you. <Laugh>.
FWM: Robots last a long time. It would be nice to allow your company to get into the already installed base of robots, because there are certainly over a million, I’m not sure how many, but a few million. And probably tens of millions coming, right? People want to extend the life of capital equipment, and that would be a nice opportunity. Any plans to retrofit like that?
Sandven: Yeah, we can do that. The sensor can just be attached with screws to a robot, and connected with an ethernet cable and you’re up and running. That’s absolutely possible. It’s not where we see the the most requests. We’ve had some of those, but where we have the strong interest is for integration into new models of robots. The turnaround time on, or the cadence on new model is quite fast. The big companies come out with new improved models continuously. They’re changing their sensor package and improving all time.
FWM: Yes. Is there a particularly good target for you on robots versus cobots? Or do you think all of it’s pretty much set?
Sandven: The AMRs are certainly good, that’s a use case I believe in very much. It’s the category with the most robots out there. They have kind of crossed the maturity threshold if you can say that. It’s a scaling application. They are out in numbers now. They’re now being used in all kinds of environments, industrial environments not really designed for AMRs. They need to be more robust in their safety sensing. The AMRs’ form factor is a good fit to the ADAR. It’s a very small sensor. If you’ve seen how the lidars are built into the AMRs today, the AMR is more or less built around the lidars, right? You have one in each horizontal corner, and I can imagine that that can be done much simpler using ADAR.
FWM: Yeah. And even on a retrofit situation, I saw a form factor that was about the size of a portable camera for the ADAR. And it adds almost nothing to the footprint of an AMR.
Sandven: That’s true. The built-in version can be even smaller. You only need the acoustic elements to be exposed and then everything else is electronics. You can kind of move back into the robot. It can be made almost invisible on the robot.
FWM: You know, that, that’s interesting. I thought maybe that the AMR market would be your core market but there are, there are many AMR companies that don’t have a line of regular robotics or a cobot. Forward X Robotics is one such company. But even old companies, hundred-year-old companies like Yale who makes, forklifts and things like that, they’re coming out with some self-tracking and safe warehouse delivery products. They’re more about material delivery, picking and things like that. But still, they can navigate a warehouse. That could be another great opportunity for you. I’ve seen many driverless forklifts lately.
Sandven: And and we even have request for manual forklifts, driver assistance.
FWM: Really?
Sandven: Oh, yes. If you’re backing up a forklift, it can be difficult to see if there’s something behind you, and to get automatic stop if you are about to back into someone. That’s also an interesting application.
FWM: That’s right. It’s just like what you mentioned before, on the bumper of your car.
Sandven: What we have is a very advanced parking sensor. An automotive application is also interesting, but we need to maintain our focus. We need to get our product out in volume as soon as possible.
FWM: You cannot do it all at the same time, right?
Sandven: No, you can’t. At least not from little Norway <laugh>.
FWM: It’s interesting, too, looking at your site you have I guess around 20 or so people. Most of them look like they are—compared to someone like me—quite young and full of ideas and energy. You can just sense it from the people that you have. Do you agree with that assessment?
Sandven: Yes. I think we’re 22 people now. And quite consciously I built a team with a senior management team. I’m 50 years old, and we have a layer of people who have been out there before and really developed very cool technology products before. Then we have recruited some people directly from university. We have a bunch of quite young people—extremely bright, hardworking, real doers. That’s a good combination to have people grow together with more and more senior people.
FWM: It seems like to a young engineer, innovation comes easily. It’s second nature. It’s just what I do in that role. It’s a great place to be.
You have a lot to roll out if you want to capture the different pieces of these markets. That’s a lot of work. Do you think you’ll pursue everything we talked about, or focus on geting these into the robots, and then we’ll talk about these other markets?
Sandven: We have had one focus, that was to get our product into robots. And my goal is to have it into all robots. That’s the potential, and that’s what we are pursuing. Then in the static applications, machine safety has come up and we just have to do that as well because it’s such a good fit for the technology. At least we’ll handle those two use cases or applications, and we’ll see how much more we can do in parallel.
We we are growing the commercial team now. We have some experience working with global partners, so these, you know, big central companies and industrial companies. And if you get that role that’s of course very powerful, but it takes some time. It’s a two-punch approach.
FWM: Yes. I can really sense your excitement.
Sandven: We have the ambition, so let’s talk again in a year and see how it’s going. Right now we’re in a very good place, with lots of interest and lots of testers getting the results we hope for from using our sensors. And at the beginning of next year, we will already see robots with our sensors on the market being sold.
FWM: Great.
Sandven: So these are very good times. We are in a positive drive.
FWM: Well, all of that is terrific. Anything else you would like to add before we wrap up?
Sandven: No, I think this has been a very nice conversation, David. So we’d love to to talk with you again when we have been out on the market for a while, to see if things progress. It was a pleasure talking to you today.
FWM: It was very nice talking to you. Thank you so much.