We're sunsetting PodQuest on 2025-07-28. Thank you for your support!
Export Podcast Subscriptions
cover of episode EP 569: Autonomous Drone Inventory Management and More with Corvus Robotics

EP 569: Autonomous Drone Inventory Management and More with Corvus Robotics

2025/3/10
logo of podcast The New Warehouse Podcast

The New Warehouse Podcast

AI Deep Dive AI Chapters Transcript
People
J
Jackie Wu
Topics
@Jackie Wu : 我是Corvus Robotics的创始人兼CEO,我们利用自主无人机技术来彻底改变仓库的库存管理方式。传统的库存管理依赖人工和条码扫描器,效率低下且成本高昂。我们的无人机系统利用先进的AI和计算机视觉技术,能够自主飞行并进行库存盘点,从而显著提高效率并降低运营成本。我创立公司的初衷是源于家族成员在仓库运营中遇到的库存管理难题,以及我对利用最新技术解决这一问题的热情。在研究生阶段,我走访了众多仓库,发现几乎所有仓库都采用相同且低效的库存追踪方式。这促使我与CTO Kabir合作,利用当时新兴的微型飞行器技术,开发出Corvus One,这是全球首个完全自主的无人机库存管理系统。Corvus One无需人工干预或额外基础设施,例如贴纸或信标,极大地降低了安装和维护成本。我们克服了其他公司在此领域面临的技术挑战,成功地将该系统应用于数十家大型公司的仓库中,每天进行全天候的库存盘点,并持续扩展。我们的系统不仅能够准确地识别库存差异,还能发现其他问题,例如破损的标签或散落的包装,从而进一步优化仓库运营。我们自主研发无人机硬件和软件,以便更好地控制数据收集和模型训练,并通过构建大型世界模型来实现无人机的自主性。这与大型语言模型类似,但应用于机器人视觉领域。我们收集的数据不仅用于改进库存盘点,还用于优化仓库布局和工作流程,提高效率并降低成本。我们提供免费的系统升级,以适应不同的仓库环境和意外情况,例如地面积水。 @Kevin Lawton : Jackie Wu分享了Corvus Robotics的自主无人机库存管理系统,该系统利用AI和计算机视觉技术实现完全自主的库存盘点,显著提高效率并降低成本。该系统无需额外基础设施,易于集成到现有的仓库管理系统中,并能发现库存差异及其他问题,例如破损标签或散落的包装。此外,该系统还能通过数据收集和分析,优化仓库布局和工作流程。 Jackie Wu: Corvus Robotics的无人机系统与市场上其他库存管理解决方案相比,最大的区别在于其完全自主性和无需额外基础设施。其他解决方案,例如使用地面机器人或信标的系统,各有优劣,但它们都需要额外的基础设施或人工干预。我们的系统更适合那些不需要实时监控所有库存的仓库,而其他解决方案则更适合库存周转率高或仓库规模大的客户。尽管市场上存在其他解决方案,但仓库和供应链自动化仍有巨大的市场空间。Corvus Robotics将继续发展壮大,并不断改进其无人机技术,以满足不断增长的市场需求。

Deep Dive

Shownotes Transcript

Translations:
中文

Want to learn about the benefits of transforming your propane or diesel fleet of forklifts over to purpose-built integrated lithium-ion equipment at ProMat? Go visit Big Joe Forklifts in booth N7157 at ProMat in Chicago, March 17th to 20th. With almost 75 years in business, Big Joe has established itself to be your trusted partner when electrifying your internal combustion fleet of forklifts.

Big Joe has knocked down the last remaining barriers to electrification with its purpose-built integrated lithium-ion forklift solutions with performance that rivals internal combustion equipment at just a fraction of the cost to maintain. Big Joe's broad line of counterbalance forklifts are built around an industrial-duty lithium-ion battery designed, manufactured, and supported all by Big Joe.

Come see their solutions and meet the Big Joe leadership team, including Martin Boyd, Jason Dunnigan, and Nick Malwecki, March 17th to 20th at ProMat in Chicago, booth N7157. That's booth N7157. The New Warehouse Podcast, hosted by Kevin Lawton, is your source for insights and ideas from the distribution, transportation, and logistics industries.

A new episode every Monday morning brings you the latest from industry experts and thought leaders. And now, here's Kevin.

Hey, it's Kevin with the New Warehouse Podcast bringing you a new episode today. And on today's episode, I am going to be joined by Jackie Wu, who is the founder and CEO at Corvus Robotics, and they are utilizing unmanned drones to do inventory management as well as some other tasks within the warehouse as well. And we're going to learn a little bit about that. We're going to hear a little bit about how

How their solution kind of came together as Jackie is the founder and what it's all about and also how they're using large world AI models. We're going to learn about what that is. First off, Jackie's going to teach us here today. And we're also going to hear why that matters when it comes to this inventory management solution. So, Jackie, welcome to the show. How are you?

Thanks so much. Hey, guys. I appreciate you inviting me and excited to talk about all things warehouses, AIMS, Autonomous Inventory Management Systems, and how AI will be changing inventory management for warehouses soon. All right. Awesome. I'm definitely interested to hear about this. As people know, they listen to the show. I'm a big inventory management control fan because that's kind of where my career started in the warehouse world. But definitely

Definitely interested to hear about Corvus Robotics and interested to hear too how this kind of came about. Tell us a little bit about it and how did you found the company?

I was in grad school for robotics in Chicago and now we're on the West coast where MountainV we build all of our vehicles, all of our robots here in the U S which is an important distinction, you know, especially with, you know, sort of everything that's, that's going on. I, I was in grad school and as a fellow warehouse nerd, uh,

I'd been to two dozen warehouses in 10 different countries, in Europe, in Asia, in the US, Canada, and Mexico. And I realized that all warehouses are tracking inventory the same way, like almost every single, like 90 something percent of warehouses, right? So it's people, barcode scanners, climbing scissor lifts. And when I was in grad school, there was this new technology coming out to allow robots, aerial robots to be fully autonomous using a vision-based system.

And so I wanted to solve this challenge with basically the latest technology coming out in hardware and software. And so I met our CTO Kabir, like the joke is we met online, like most couples do these days. We were on GitHub, which is this code repository website. And there's this one handle that kept answering everybody's hard questions.

And so I ping him thinking he's, you know, like a 40 year old man somewhere. And he said, you know, he's, he says he's going to be in Boston in the fall. And it turns out to be an incoming freshman at MIT. Oh, wow.

And so we built the first vehicle actually at Simmons, which is like his MIT dorm. I don't know if you know this, but most drones in the world are or were based off of essentially this autopilot stack. And he was one of the founding contributors of that autopilot stack back in the day, which is an interesting coincidence. You know, now these days there's so many DJI does this and that and

Of course, people repurposed dual-purpose drones for other reasons, but a lot of the core auto controller was kind of invented back in the day. Interesting there and definitely interesting too to learn about that. And I got to ask you because you said that you were a warehouse nerd, right? So where did that come from?

I have a family member who built this company that expanded to a lot of warehouses, essentially. And one of the core challenges that I would hear him say was that stuff has gotten it. Where is this shipment? Where is this going? Somebody else screwed up some order. And that was just the constant operational headache and the logistic sort of

problems that every, every, it seems like every company selling physical goods deals with on a daily basis. And, you know, I just, I thought that there had to be a better way with the computer vision and robotics that was just starting to become, you know, come into research. And so, you know, if it's coming into research, then a few years later, it should be commercializable. And, and if we were to seize that, then we would be on the cutting edge of this. And so that's what we've been doing for the past several years.

Very interesting there. Yeah. Yeah. Cause yeah, you definitely surprised me there when you said you were, you were a warehouse nerd in grad school. I'm like, Oh, I gotta, gotta hear about robotics. Right. But that intersection is very small. Yeah, definitely. Yeah. Interesting. So, so interesting to hear the story there and obviously, you know, very interesting to how you met your, your co-founder and,

and how you're able to develop the solution utilizing drones and that unique kind of, I guess, positioning from a family perspective that you're able to see this challenge that people were dealing with. So tell us a little bit about, we know it's a drone definitely, but tell us a little bit about, you know, how the solution works and how it addresses some of these challenges like missing inventory. Yeah.

So when we first looked at the problem, we tried a variety of things, right? We thought naively, what if we put cameras all over the facility? We quickly discarded that because, you know, as you know, these

facilities go up 40 feet 70 feet in the air uh cameras can't see all that much plus the maintenance for all the cameras for a million square foot facility like that's that's not going to work yeah we thought about building a ground robot sort of like the ones that you see in grocery and now there's a very excellent company out in europe doing that dexterity but for us that wasn't the you know

When I think about segmenting solutions, there's sort of orders of magnitude, right? And the technology that was coming out in grad school was micro aero vehicles, was drones. And these are small drones that fit on the palm of your hand.

And they're very cheap to make if you can master the manufacturing process, which we have. And we can we figure that we can get these very small vehicles that can easily go up 50 feet, 70 feet. There's there's no there's no high limit. Right. And do this fully autonomously. And so when you sort of segment out solutions for inventory tracking for autonomous inventory management systems, there's sort of your auto stores.

you know, that are, you know, 10 million, a hundred million, right. Then there's sort of ASRS system, like sort of more traditional systems that are, you know, you know, one or 2 million X million, right.

Then there's sort of an order of magnitude. The ground robots would always be substantially more expensive than a, frankly, a micro aero vehicle. And so we kind of segment that. Of course, the more you pay in capital, the more the functionality. But inventory control is something that doesn't necessarily have to be done 24-7. You don't need eyes on every single location at every moment because it

It's not moving most of the time, right? Yeah. Depending on the type of facility and so on. And so the core technology that we wanted to capitalize on was we've built Corvus One, which is our product. It's the world's first, and as far as we can tell, only fully autonomous device.

drone inventory drone system, meaning that it flies and does its mission day after day, week after week, month after month with no human pilot, nobody moving it from aisle to aisle, no stickers, thousands of stickers that you have to put on for location or beacons and reflectors that you have to put on.

And it's a very novel technology because it means setup costs and maintenance costs are so much lower. If your warehouse racks ever change, if you sort of move the racks around, then you don't have to go up and put up another 10,000 stickers or ceiling mounted reflectors or anything like that.

Looking at the history of the space, I mean, we're not the first company to attempt this. We're just the first company to have succeeded. There's another dozen or so companies that have raised tens of millions of dollars, $30 million in order to do this, that have attempted and then shut down over the years, either

using an off-the-shelf drone, which then they have to move from aisle to aisle. There's been a dozen or so companies that have done that. Or they're trying to build their own vehicle, but then they weren't able to have the technical expertise to be able to do that. But we have. It's through a lot of hard work. There's been no silver bullet. It's been lots and lots of lead bullets at all of the technical issues, working with customers to understand each of the pain points and each of the different environments in every warehouse that we go into to be able to solve this.

So I'm proud of what we've done. I think our team is, and it feels weird to say this, but our team is the best in the world at what they do, at building the fully autonomous warehouse inventory trends. We're helping several dozen very large companies, like global 2,000 size companies, track their inventory every single day. Like they are flying every day, every night, weekends, and so on. And they're expanding. So we're quite happy to be able to help our

you know, a lot of the customers that frankly comprise, um,

major parts of the US economy in tracking their inventory and sort of optimizing that. It's a key function that everybody does, even though it's not as sexy, it's not as cool, right? Inventory control is not sexy. I think it is. I'll say it. I think it is. Thank you, Kevin. You and me. You and me. I'm there with you, definitely. But yeah, so I mean, I think it's very interesting. And very interesting too, you mentioned there that, you know, there's not much

infrastructure change that needs to happen, right? No reflectors, like you said, or labels that need to go everywhere, which makes it easy to expand, right, if you're going to change your racking layout or put in more racking or whatever the case may be. So I think that's a really interesting point there. And you've said, too, a couple times that it's autonomous, right, which means...

you know it's doing this on its own so i mean is there is there any level of human involvement at all like what what does that human drone interaction look like if there's anything yeah not much i mean typically you know of course you're gonna have to tell it to fly at some point you can kind of set a schedule you know kind of like google calendar or ical right so you can have it set

Every Tuesday, I want it to fly these particular aisles on these particular times, Wednesday, these other aisles, or have it go on a routine. It's important for us to save that labor component such that every week, maybe you're spending a few minutes looking at this, as opposed to staffing a full-time person to track this or manage the drones, which doesn't make a lot of sense. Yeah.

Autonomy is really important. I mean, before this, you've seen the self-driving car space, right? Like the autonomous car space, they've invested hundreds of billions of dollars and there's maybe Waymo and SF and LA, right? And Arizona. And then Tesla works in certain parts of the West Coast. But autonomy is a very hard problem. And

And it takes a lot to be able to build that. So part of our secret to success, and you alluded to this before, is we've just collected a lot of data on this and all the different types of warehouses, they look very different.

different conditions and built this large world model that robots are robots operate. And so the way, you know, an analogy would be 2023, 2024, you've heard of large language models. It's like chat, GBT, open AI, all this stuff that people are playing with, right? You're typing in there instead of asking Google about, you know, something, right? 2025 and 2026 are going to be the advent of these large world models or large vision models. It's beyond just text.

It'll be the things that robots and sensors see and how they can act in the environment. And it's a very interesting space. I'm super glad to be working on this stuff right now. I mean, you know, key to doing this is like we build our own drones here in California and that lets us control the hardware stack such that we can collect the data that we want and train the way that we want and run the models on board the way that we want. And all of that is key to being able to deliver this type of solution to our customers.

Interesting. Yeah. Yeah. And I think it's interesting too, that you're building the hardware yourself, which you said, it gives you like a little more, more control on that. So, you know, obviously we've talked about the hardware here and you talked a little bit about kind of the, the data capture itself and how that's working, but what, what does it look like on the backend? Because you're going out there, you're capturing the data, which, you know, a human would normally do with, you know, whether it's,

uh, hopefully not paper. Uh, you guys are using out there who's listening, but you know, or like a, a scan gun or something like that. I mean, you know, so what happens then, you know, captures this data and now, you know, there's a discrepancy, like what happens on the backend? Oh,

I'll give an example. Well, for the warehouse themselves, of course, we connect to the WMS. It's a pretty easy integration. There's just, I mean, it's inventory control. It's like, what is where when, right? And then there's some pictures and fancy stuff that we do, but the basics of it is very easy. So we can integrate to whatever Manhattan, SAP, Oracle, proprietary, AS400 things. But they, and the customers get a

snapshot of their inventory, especially of discrepancies and exceptions. And so we'll note some of the things that we see that maybe they can touch up on. Here's some barcodes that are ripped.

Here's the loose shrink wrap, here's whatever. And then we work with the team to add that as a part of their core inventory control process. You get the discrepancy report, here are the locations, swipe, swipe, see what the issue is. And truly if there is an issue, then we make an adjustment. WMS says there's something, here's a picture taken 10 minutes ago showing an empty location, then we got to make a change.

Then the other piece of it is for whatever type of new environment that we encounter at a site. So an example is actually out in your neck of the woods. There were some storms recently and there was some rain in some of the facilities and leakage. And so there were water puddles on the ground.

Yeah. And so our drones hadn't seen that before because you're in a warehouse. Why are you what do you expect that? And so we collected that data and we trained on that data of seeing puddles on the ground. And then we updated all of our customers drones fleet wide with a new

new information on what to do when there are, well, water puddles in the ground, which there shouldn't be, right? There shouldn't ever be. But there are. And so that's, you know, an example of sort of the

the AI model that we do. And we provide all this to our customers at no additional charge that the free upgrades as a part of our robot as a service offering. Interesting. Yeah. Yeah. And I think obviously, you know, one of the benefits to AI there or,

or in any sense, right, is that, you know, the more data that it captures, and in this case, the more data it sees, I guess, in a sense, right, the more that it learns and improves over time, too, as well. And you mentioned in there, you know, talking about not just inventory discrepancies, right, like not just the traditional, they're supposed to be 10,000,

and there's only eight, right? You talked about like a ripped label, right? Or loose plastic wrap in the location. So tell us a little bit more about that and kind of some of, you know, obviously the core focus I think is this inventory control, but then it sounds like there's these additional benefits surrounding that that just naturally happen because you're already taking a picture or visiting that location. So tell us a little bit about some of those additional benefits and how you

Some of your customers have been able to leverage that too. We'll be back after a quick break. You don't want your employees to be idle, so why would you want your robots to be idle? Zebra Technologies is addressing idle robots by bringing their latest solution to the market, which requires 30% fewer robots.

resulting in increased asset utilization and a cost-effective entry to automation for your fulfillment operations. The Zebra Symmetry Fulfillment Ecosystem is bringing you balanced utilization, more cubic capacity, and optimized pick paths so you can operate at peak performance. Want to take your fulfillment operation to the next level? Learn more at zebra.com or click the link in the show notes.

Yeah. Inventory control is mostly a cost center, except for some 3PLs who there are some ways that they can figure out ways to make more money. But for the majority of retailers, distributors, manufacturers, inventory control is frankly a cost center. And you want to spend as little as you can in order to get as good inventory as you can. You don't want to spend a million dollars a year, hopefully, on just inventory for a warehouse.

or even half a million. And so we're able to provide, because of the technology that we built, right, with these autonomous drones, we're able to provide this solution that's quite cheap and collect that inventory and collect the data for the inventory in a way that's cost effective. And so with that, in addition to the direct improvements to inventory discrepancies, there's the things that we can do to improve how that warehouse is laid out and some more.

like workflow opportunities. And it depends for that facility, but you know, essentially it's improving utilization, improving routing such that you can get the most out of your labor. Cause we realized that a lot of these, you know, for, for folks, these are cost centers and if we can improve that, you know, how they, how they operate, how they store that's, that's all, you know, bottom line that the company gets to save. Interesting. Yeah. Yeah. I think that's an interesting benefit is cause I mean, I'll say, um,

you know as a previous warehouse manager i mean you know i would sometimes send people just just out to look for those things right you know look for like is there a location that has a just an empty pallet in it right or is there a location that's got trash in it that somebody didn't clean up when they were done picking from it or something and you know how do we clean that up so you know on the other side when people are coming to put in new inventory or

move inventory into those locations, whatever the case may be, you know, there's not a slowdown for them. But, you know, I think that's great that it's constantly looking for those types of things too, because that keeps the operation flowing as well, which I think is super important. So interesting there. And I think, you know, even a great example too is the water as well, right? Like, oh, you know, now I see water and, you know, I create some kind of alert or do something differently in that sense. But, you know,

So now, you know, if we look at Corvus and we look at kind of the landscape of inventory management and control and trying to automate that, there's a couple players in the industry. Right. So so tell us a little bit about, you know, what some of the differentiators are for for Corvus and what makes you a little different than maybe some of the other ones that people have heard of.

I have a lot of respect for Dexory and for Verity. The first company is, they're both European companies, and the first company has innovated using this basically like a mechanical giraffe. It's like a very tall ground robot that goes up very high and this goes around.

And the second company essentially puts up, they came out of doing drone light shows. And so they put up a bunch of these sort of wireless beacons, just like for drone light shows for the drones to know where they are. And we are in a unique space. Corvus Robotics is the only one that is building drones, but they are fully autonomous, not using any sort of beacons or stickers or anything like that.

And they are, I would say that the ground robots sit in a marginally different sort of market profile than the drones. If you are using the ground robots to scan just everything constantly, you have a higher need. Maybe you have a much faster inventory turn or a much bigger facility. And that's a fundamentally different approach than the vast majority of warehouses, which like, do you really need to see all of your inventory, you know, every single hour and

And the answer is, and pay the money to do so. And the answer is, frankly, usually not. So these folks are quite good. We sit in a slightly different vertical from one of them. And the lack of external infrastructure of having to put up

hundreds of beacons or come back and do that if the racks change or something like that is a key difference. What's interesting though is that despite this, there's plenty of space for everybody. It's a very large market. I mean, warehousing logistics is one of the biggest industries in the world. The biggest competition, like the 98% of the competition is the existing way of doing things, right? Which is...

guy with barcode scanner and a scissor list or something like that. And so until we are automating, having autonomous inventory management systems at the majority of warehouses, I mean, it's a, it's a blue ocean market. It's a very large market for, for every, every, every company to go after. I mean, even for full automation solutions, like the auto stores, the

you know, the locus geek plus of the world. I mean, they have the sum of all of those companies is like a couple percent or something of the market share rate. So yeah, there's a lot of opportunities for warehousing and for supply chain if we both understand the industry and also understand the technology.

Interesting there. Yeah. Yeah. I was interested to hear your thoughts on that because it's certainly something that I think is, you know, like you mentioned, there's a couple of different solutions, a couple of different form factors to even the way people are approaching it. And, you know, I think it is interesting there that not necessary to make those infrastructure changes is certainly a big differentiator that I see for sure. Because, you know, as you said, is that,

or that the operation evolves, whether it's, you know, taking down rack, putting up more rack, changing the way the rack looks, adding more levels to the racking, whatever the case may be, you know, having that to easily adopt to what this is

you know, environment looks like now, I think is a definitely a big plus and value add there for sure. So very interesting to hear about this and kind of learn a little bit about how you guys are approaching this problem and some of the additional benefits on the side too, as well, and how you're building out that hardware and keeping control of that a little bit. So, you know, as we look to the future here, what does the future of Corvus look like?

We're growing quickly and we'll continue to do so in 2025, especially with a lot of the supply chain disruptions and challenges that are coming up. Supply chain is not going to get easier. You know, people, companies, customers' expectations are going to get higher and we have to continuously innovate and adopt technology and innovations in order to keep up with them.

For us, we will keep developing our technology. There's a lot of sort of improvements that we can continue making using the data that we've collected for our vehicles. And it's all toward that vision of the autonomous inventory management system.

Autonomous Inventory Management System, AIMS, right? Is what we're calling it. AIMS. Gotcha. All right. Definitely. And we'll definitely be looking towards the future there and, you know, seeing what you guys continue to do with that evolution. And I think what makes it, you know, most interesting is we talked about here is that continuous data capture and

you know kind of leads to like what additional things can this do also while it's accomplishing this this inventory management goal so it's very interesting to talk to you here Jackie and and learn about Corvus Robotics and what you guys are all about how you were founded and and what you're doing there in the space so if people are interested in learning more about Corvus maybe you know getting in touch and seeing if this is a good solution for them what's the best way to do that

Of course, go on our website, Google in Corvus Robotics or corvusrobotics.com. And always feel free to send us an email, sales at or info at corvusrobotics.com. All right, great. And we'll definitely put all that information at thenewwarehouse.com as well. We'll also put it in the show notes here too so people can easily find it. So Jackie, thank you once again for your time on the show today.

You've been listening to the New Warehouse Podcast with Kevin Lawton. Subscribe and check us out online at thenewwarehouse.com. Enjoyed this episode? Make sure you are subscribed to the podcast and for more content from The New Warehouse, find us on LinkedIn and YouTube. Links to subscribe can be found in the show notes and for everything The New Warehouse, head to thenewwarehouse.com.

We're sunsetting PodQuest on 2025-07-28. Thank you for your support!

Export Podcast Subscriptions