AI and robotics are transforming how we all get pretty much everything. One way, is via the world’s tallest autonomous robot, which can scan a million square feet of warehouse and 100,000 pallets in a day.
In this episode of TechFirst, we chat about AI and robotics with Andrei Danescu, the CEO of Dexory.
There’s an invisible but intricate network threading its way through all our cities, states, and countries, almost like a human circulatory system. It delivers pretty much everything we need to live, work, and play, and it’s something that AI and robotics is rapidly transforming.
One new piece: Dexory’s autonomous robot that scans a million of square feet or warehouse shelving daily, creating a digital twin of entire warehouses, and significantly improving efficiency and accuracy.
Watch: the world’s tallest autonomous robot
(Subscribe to my YouTube channel)
Skip to whatever section you’re most interested in:
- 00:00 AI and Robotics in Logistics
- 00:45 Warehouse Challenges
- 01:27 Dexory’s Technology
- 02:26 World’s Tallest Autonomous Robot
- 04:06 Technical Deep Dive: Robot Design and Functionality
- 07:04 Operational Efficiency and AI Integration
- 12:13 Subscription Model and ROI
- 23:21 Future of Autonomous Systems
- 24:43 Conclusion and Final Thoughts
Subscribe to the audio podcast:
Apple Podcasts
Spotify
Google Podcasts
Overcast
Other podcasting platforms
Transcript: the world’s tallest autonomous robot
This is AI-generated and not guaranteed to be full accurate
John Koetsier: How are AI and robotics transforming how we get everything? Hello and welcome to Tech First. My name is John Gutsier. There’s an invisible but intricate threading through all our cities, states, countries. It’s like a human circulatory system delivering pretty much everything we need to live, work, and play.
And it’s also something that AI and robotics is transforming. Today we’re chatting with CEO one of the companies leading that transformation. The company’s Dexory and their AI and robotic systems can scan a million square feet and a hundred thousand pallets in a day. The CEO is Andre Danescu.
Welcome, Andre.
Andrei Danescu: Thanks a lot for the invitation, and it’s great to be here.
Great to be part of this. Thank you, John.
John Koetsier: Great to have you. Let’s kick off right here. you think, Hey I put something in a warehouse. I know where it is. My system knows where it is. I take it out. I know I’ve taken it out. My database knows that and it knows how much I have left …
But it’s not that simple, right?
Andrei Danescu: So in the ideal world, it would be like that. But we all know that we don’t live in the ideal world. And there are pressures to get things done quickly. There are time constraints. There are errors that we all make growth human at the end of the day. So it’s not actually that simple. Indeed,
John Koetsier: I guess there’s also pilferage, there’s damage.
There’s just errors of where people put stuff, all that stuff. What does your system do?
Andrei Danescu: Exactly as you mentioned, there’s all kinds of errors that you have in a day-to-day operation of a warehouse in a day-to-day logistics operation, and the technology that we developed which is represented by Dexory Review, works to close and eliminate the visibility gap.
And the reason why we call it the visibility gap is ’cause you would like to know everything that’s going on in the warehouse, every single pallet movement, how much product you have in every single location, and also things like what are the conditions in which you’re storing the goods.
What’s the temperature, humidity? What how long were the goods? Therefore why hasn’t this location been turned over, or why haven’t the goods moved?
And that’s exactly what our technology does. Review. It’s a digital twin platform. It’s a global visibility platform, so you could deploy this across a hundred warehouses, and you have one central location where you have access to all the data.
And what’s really unique about it is that this information that we capture and collect on a daily basis is collected using autonomous robots. So we use this very tall world’s tallest autonomous robots to scan the warehouses, recreate the digital twin, process the information, and make this data available for 100% efficiency and eliminating all the operational errors.
How tall is this robot? So they’re 12 meters tall at the moment. Yeah. Which is about four, did you say
John Koetsier: 12 meters?
Andrei Danescu: Yes.
John Koetsier: Wow. 12 meters tall for our American friends. That’s 36 feet, 40 feet, somewhere around there. Did you say that’s the world’s tallest autonomous robot?
Andrei Danescu: It is indeed and deployed commercially in warehouses.
Yes.
John Koetsier: I’ll tell you a little fun fact about myself. When I was in university, I had a lot of different jobs. One of them was actually installing racking in warehouses. So you’d be up 12 meters handling all kinds of beams and everything, putting them together, welding them, all that stuff. It’s a little high.
Andrei Danescu: It is, and people have even higher warehouses. We are maybe secretly working in our deep tech labs on even higher robots, of course, but maybe secretly.
I wouldn’t know. I wouldn’t know. I’m not allowed in there. but exactly like you said, the robots were designed specifically for this application for acquiring data.
So obviously we made them as high as the racking, so you can scan the entire warehouse in one pass.
John Koetsier: Andre, tell me a little bit about the robot. How does it scan, what does it use, how many sensors does it have? All that stuff.
Andrei Danescu: So I’ll try and keep this relatively short ’cause we don’t have all day and I would love to speak about this technology all day.
John Koetsier: We have 500 hours. You can take as much as you want!
Andrei Danescu: That’s a good tease. I might actually carry on forever. So I, like I was mentioning, we designed the robots specifically for this application. So for data collection in warehouses. It’s designed to operate fully autonomously, which means you don’t have to make any kind of warehouse infrastructure changes.
You don’t have to put markers, you don’t have to change the space. None of the space interaction, but also very importantly, it’s designed to operate during normal working hours, during the night, during the day, whenever you want to be taking the scan.
Of course, on the bigger sites, like we were discussing a million square feet over a hundred thousand pallets … it does take, 22, 24 hours to scan and turn that around. So it is literally designed to operate during any type of conditions as a collaborative robot integrating with the existing for workforce.
John Koetsier: So how does it scan stuff? What sensors does it have? Obviously some visual sensors but I’m assuming more than one.
Andrei Danescu: It definitely has more than one. In order to scan everything at the speed. We use the autonomous systems on the robot to do the dig digital twin reconstruction. So the robots operate 3D Slam. We have our own development of 3D slam to be able to navigate the environment, and then we use additional lidars and we use additional cameras to reconstruct the digital twin.
When we say digital twin, obviously we’re focusing on the racking and the rack location side of the warehouse. So it’s not so much about reconstructing the building for the sake of having the building, but more about what is actually going on in every single one of these locations. So we have six lidars.
We have one camera for pretty much every single rack row.
So if you have 12 rows, you’d have over 12 cameras ’cause we have additional cameras to be able to perceive the environment better. And we also have other sensors like I was mentioning temperature, humidity, environmental conditions the amount of noise, the amount of light, the capabilities of the existing network infrastructure, because that’s another really big thing.
You would probably know this from building racking like you mentioned. But if you have a 12 meter, 40 feet high rack. And you go and scan something in location and then you are out of wifi coverage and then your handheld scanners don’t work … it is a problematic area for that warehouse operators to, to be able to do their jobs set.
And the technology is really all about eliminating these gaps and creating a very efficient work environment.
John Koetsier: Super interesting. So 12 cameras, is that per side? Can you scan both sides as you go down an aisle?
Andrei Danescu: No, we are focusing on one side at a time. So the scanning speed that that we can achieve, we’re scanning over 10,000 pallets an hour.
So from that point of view, it is incredibly fast. In that deep tech secret lab that I mentioned, I’m not allowed in, we are working of course, on, on higher scanning speeds, but the main reason why we do this. Is you have a large variation of warehouse layouts. So you could have very narrow aisles.
You can these DNAs, you can have wide aisles. It’s quite difficult to always be able to center yourself in the middle. And also, navigating down the middle can impact operations because you’re kinda eating up a lot of space. Whereas if we’re going very close to one face of the rack, there’s a lot of space around the robots for people to actually carry on doing their day-to-day activities.
John Koetsier: That makes sense.
Andrei Danescu: That’s the reason and the choice why we’re only scanning one side at a time.
John Koetsier: That makes sense. That’s interesting. Now, if you’re going 12 meters up, you can’t be just some skinny little robot at the very bottom, otherwise you’d be very tippy. So what are the dimensions on the ground floor?
Andrei Danescu: Currently we’re just under a meter wide and about 1.5, 1.6 meters long. As the overall footprint of the base, and the reason why we made it so narrow is it’s actually designed to fit very narrow or the narrowest VNAs. In VNAs is actually very important because it’s also one of the more difficult parts of the warehouse to actually acquire any kind of information because the narrow the lanes are narrow and you don’t even see.
With your with your eyes out all the way to the top.
John Koetsier: So correct me if I’m wrong, those are the types of warehouses and the racking that is completely automated, that there’s robotic pickers and other things like that going down very narrow lanes. Is that correct?
Andrei Danescu: No, actually not entirely. So you can have automatic storage and retrieval systems.
Which are usually very tall. They’re up to 30 meters and then 40, 50, 60 meters deep. But they’re they do have, these are the ones where you have a crane that puts the pallets and retrieves them. But that’s not what I’m talking about. Okay. I’m talking about standard rack, which is quite close to itself.
Yeah. And usually you have this narrow, forklifts Yes. Which collect the the pallets, but they’re still traditional forklifts with human operators.
John Koetsier: Okay. Interesting. Now, the base of the robot, do you weight that so that it’s heavy at the bottom so that you know you’ve got stability?
Andrei Danescu: Yeah, absolutely.
So by design it’s very bottom heavy because of course it needs to support exactly like you said, the entire height of the robot. So we are talking about 700 kilos and center of mass, which is below … 40 centimeters from the ground.
The overall product weight it is indeed very bottom focused.
John Koetsier: That is super impressive. And you mentioned you got this a hundred thousand square foot warehouse. It’s funny how we’re talking to meters for height, about a hundred thousand square foot for square feet for the warehouse, but hey, this is the real world. People use different dimensions.
And you mentioned scanning that in 22 hours, 24 hours. Can this robot change its own battery? What does it do when it’s outta battery?
Andrei Danescu: Just a quick one. ’cause you actually hinted a hundred thousand square meters or a million square feet.
John Koetsier: Whichever we can do. Whichever I misheard you. A hundred thousand square meters, that’s a different matter entirely.
Andrei Danescu: So a million square feet is what we were mentioning there.
John Koetsier: Excellent. Okay, so we’re gonna need battery power here. We’re gonna need to recharge. How do we do that?
Andrei Danescu: It’s completely autonomous in terms of looking after its state of charge and after its battery. Very similar to household robots.
So the Roomba we have, or any other devices we would have in our house would know when the battery is going getting low, and then find the dock and recharge itself. We have a very similar setup. Of course, the robot has a docking station as a permanent area where it’s actually charging and.
Usually kinda waiting until it’s it’s ready to do the mission. And when it’s actually getting low on battery, it will navigate itself back to the charging dock, back to the dock, wait for the batteries to recharge and then carry on from wherever it left off so it knows that every single point in time how much power it has available For scanning,
John Koetsier: is it a fast charge battery?
That would probably be critical if you’ve got a lot of work to be done. How quickly does it charge?
Andrei Danescu: I wouldn’t class it as a fast charge battery. In all honesty, it doesn’t it doesn’t charge at huge currents and it doesn’t charge in 15, 20 minutes. It takes about two hours for it to charge.
And it can operate around five, six hours on one battery charge.
John Koetsier: Okay. Okay. Interesting. And I guess if you needed more, then you buy more machines or actually you don’t buy them, it’s a subscription model. Or perhaps eventually in your super secret lab that the CEO is not allowed into. Maybe there’s hot swappable batteries being worked on.
Who knows?
Andrei Danescu: I don’t know. I wouldn’t be able to call, never been inside. I don’t even know the location.
John Koetsier: Exactly. This is Apple level security times 10. Wow. Excellent. So talk about some of the results of this. Obviously we’ve, I’ve focused on the robot. There’s obviously a software system here as well.
You mentioned digital. Digital twin. Recreating the physical warehouse in a virtual sense and giving you a sense of what’s all there. Probably integrating with the warehouse management system, all that stuff. When you do that, when you’re in operation, what do you fix? How do you improve stuff?
How much better is it?
Andrei Danescu: That’s everything you mentioned. Very good point. From point of view of the software platform is really what customers interact with. The robot, per se, is a means to end is the device, the machine we’ve actually built to acquire the data. It processes the information locally, and then the insights get sent over to review.
This software platform is where customers can log in, can have the complete overview of their operation, of their warehouse. They can also start seeing analytics. Once you run the technology for a couple of weeks, you start observing trends. You start having an understanding of how this space would evolve over time.
Now, I can tell you that we’ve started working in warehouses where the overall efficiency was even below 90%. Which is low for a warehouse especially for a big space if you think about it. And within one to two months, we’ve managed to take that efficiency to exceed 99.56%. Wow. So what’s really impactful is the fact that you can supercharge your existing teams.
Like the big problem in logistics that people have at the moment is that there’s a massive labor shortage. There’s an increasing pressure from customers to have everything delivered yesterday. And of course there’s pressure in terms of margins, in terms of operational efficiency.
And of course when you bring all this together, it’s really hard. You can’t just keep aspiring to hire more people because. Your problems are not going away magically. Bringing this technology closing the visibility gap means that your existing workforce can actually do more in their time.
You can, first of all, you eliminate a lot of the errors almost completely remotely because you can look through the platform, see the picture, see the quantities of stock you have, make all the changes without even having to set foot in the warehouse. And if you still have a couple of locations that need physical manipulation, or you actually want to change.
Where those goods are located in the warehouse, it’s much, much easier to be very targeted and know exactly, okay, you need to go in this place. This is where the pallets are located. I want you to move these goods. I want you to distribute them. They need to be closer to the picking phase, and so on and so forth.
So it, it literally supercharges your existing workforce to do 10 times more than they could do.
John Koetsier: It makes a ton of sense, right? Because another job I had while I was in university was working in a warehouse. It’s amazing. Built them and worked in them. Who knew students needed money. And yeah, we waste a lot of time trying to find stuff.
It’s in this bay. No, it’s not in that bait. Maybe it’s in the other bay. Who knows? I didn’t even know if there was a warehouse management system then back in the dim, miss of time passed. But who knows? Anyways, I can totally see that. How do you two questions. First one, how do you measure warehouse efficiency?
What does 90% mean? What does 99.5% mean?
Andrei Danescu: So that’s definitely an interesting question because it does vary a lot because different warehouses have different KPIs and they’re targeting different areas of improvement or different areas where they’re measuring. So if you take one example, a pharma warehouse where the product that you’re keeping in there needs to.
In very well controlled conditions or well monitored conditions, maybe they’re not controlled, but they need to be very well monitored. The inventory needs to be in the exact location where you think it is, because otherwise you might be storing the type of drug in a completely inappropriate. Overall environmental condition and so on.
Inventory accuracy product location accuracy, occupancies, so product volume in every single area of the warehouse become some of the driving KPIs because if you have a pallet of drugs, for example, and they’re very expensive and very rare drugs, and they’re being like misplaced and then they get lost in the warehouse.
It literally impacts a whole it, it triggers a whole chain of events and impacts a lot of people further down the line. So being able to go in index or review and say, okay, I’m looking for this pallet. You will type the pallet number and it shows you exactly where it was detected, whether it’s in the right location or not.
It will find it for you in the warehouse. That drives a huge amount of impact. And then some of the other warehouses where they maybe have different velocities of. Different volumes, much higher volumes. For example, they want to be able to store as much product in the same space as they can.
So for them, utilization becomes really important. So one of the big drivers from KPIs and performance drivers from Dexter review is how well are you actually utilizing how well are you storing and occupying your existing space, your existing racks? There’s many more makes we only have 500 hours.
John Koetsier: Exactly. Makes a lot of sense. Absolutely.
Now what kind of AI are you using? What kind of vision are you using technology there to know what you’re seeing? You mentioned a pallet number. Okay. You can see a pallet number that makes sense. Are you scanning a barcode? Are you looking at an object and saying, this is what I think it is, or this it’s of this size, or other things like that?
What are you doing?
Andrei Danescu: So we’re using a combination of techniques there. It’s not, I don’t think it’s like a one size fits all. You only have to use one specific thing, and that’s the magic formula you can apply to everything. It’s it’s a combination of machine vision technologies for barcode scanning, QR code scanning for character recognition, which then gets.
Coupled and sensor fused with other types of perception to understand I’m looking at five boxes or six boxes, or I’m looking at this specific type of product. It’s a big problem because in a warehouse you have a log boxes. So you don’t necessarily have an easy time knowing what’s inside them, which is why you have to have a combination of perception systems.
Algorithms, like I said, a lot of machine vision technologies to do that recognition. We then have different types of perception to understand, okay, I have this product, I have these boxes. Then I look at my cloud points, I look at the 3D reconstruction to say, this is the volume that they occupy. This is what I’m scanning.
This is how much product I have on the shelves. And this is where the more advanced AI technologies come into play, allowing us to really go deep inside the data sets that we’re gathering. So you can scratch the surface and extract the superficial stuff around, reading barcodes and position of those barcodes in space.
But then when you start building these big data sets and you start using more advanced sensor fusion techniques, bring multiple sources of information together. It really allows you to go deep into that data set and bring more value to customers.
John Koetsier: Interesting. And that data somehow has to make its way into the warehouse management system, I assume, which I assume also that you are not the warehouse management system, although you feed data into it.
Is there some sort of like progressive ETL or something like that to build up that, digital twin in data in the warehouse management system so it’s continuously aware of what it has, where and all that sort of complexity?
Andrei Danescu: Yeah, that’s a very good point to touch on.
Of course, the warehouse management system is an fundamentally open-ended system. It relies on data being pushed into it. And that’s one of the reasons why we decided to build this. Robotic technologies really acquired the data, the ground truth was actually happening on the work warehouse floor.
We can integrate with any warehouse management system. Most of the systems have APIs. We obviously our system, DEXA Review has its own set of APIs. So you can pull data or you can push data from WMS systems. And what’s really interesting for DEXA review is the fact that it allows us to bring data from multiple types of warehouse management systems so we can create the single source of truth for the entire warehouse.
But exactly like you said, it’s very important not only to be able to pull data, but also to data exchange. Pretty much most WMS systems in the market.
John Koetsier: Makes sense. You don’t sell your system … you sell a subscription to it. Correct? How much would it cost?
Andrei Danescu: Oh, asking the the secret questions, all the secret questions.
Is that also in the deep secret lab? No, that’s in the secret folder. That doesn’t live in the lab. It does really depends on the type of warehouse primarily on the feature set they, you select, right? So the business model, exactly like you said is a pure subscription to the platform.
Similar to most modern ways of utilizing software we will then look after the hardware, make sure that the robot’s always operating to feed, to provide that data stream into the platform.
But in terms of the pricing model, it does depend on what we’re looking in terms of the feature set while we’re looking in terms of advanced functionalities, if there are customizations that are specific to customer, to any location or any type of customer that what I would say is the value that the system brings greatly exceeds 10, 11 times the price that you pay. So the ROI is really good.
John Koetsier: Anybody with a warehouse would be a fool not to have it?
Andrei Danescu: You said it. I didn’t, but I couldn’t actually agree more. No, but that’s, I think, it’s a good point.
And we can build just a little bit on that and say, at the end of the day, in phase world, you have to be able to bring new technology into the market and you have to give people the confidence that this technology brings the ROI they need. Because it’s not about buying something, it’s about bringing that service that really I.
Drives the value for these locations.
John Koetsier: And we chatted about this a little bit before we started recording as well, right? I’ve been doing a lot of stories about drone delivery. And we know that Google’s Wing is doing that. We know Amazon is doing that. There’s a bunch of others there.
There’s many other companies. Manna Arrow is doing that. It’s happening in the states, it’s happening in Europe. It’s happening other places as well. Asia, I’m sure. And as you enter that world you’re entering a world in which the amount of time that a product. Is at rest versus the amount of time it’s in motion that’s decreasing.
It’ll stay less time, it’ll, you’ll have this continual flow, right? And something may be in the warehouse for seven minutes, not seven days. And there it goes. So you’ve got to be really sure about what you have, where it is, how to get it, how to access, how to get it out, how much you have left, all that stuff as that pace increases, correct.
Andrei Danescu: Connectivity is one of the most, most important elements of that puzzle as we’re building that future of fully autonomous deliveries and this global visibility across our supply chain.
John Koetsier: Interesting. Autonomous delivery. Autonomous warehouses. Autonomous production, autonomous everything. Excellent.
Andre, thank you so much for this time. I really do appreciate it.
Andrei Danescu: It’s been a pleasure and hopefully we speak again soon and I can tell you some of the secrets. I’m gonna try and find that secret lab.
John Koetsier: I’ll parachute in at night. James Bond style or we’ll see.
Andrei Danescu: You don’t know the location though?
John Koetsier: I have my ways.
TechFirst is about smart matter … drones, AI, robots, and other cutting-edge tech
Made it all the way down here? Wow!
The TechFirst with John Koetsier podcast is about tech that is changing the world, including wearable tech, and innovators who are shaping the future. Guests include former Apple CEO John Scully. The head of Facebook gaming. Amazon’s head of robotics. GitHub’s CTO. Twitter’s chief information security officer, and much more. Scientists inventing smart contact lenses. Startup entrepreneurs. Google executives. Former Microsoft CTO Nathan Myhrvold. And much, much more.