Cold breaks a lot of hardware that runs fine in an ambient building, so Corvus One for Cold Chain meant re-architecting a system that already worked. Thermal management came first.
Cold Chain Robotics
Q&A with Michael McSpedon, CRO | Corvus Robotics
Tell us about yourself and your role with Corvus Robotics.
I'm Michael McSpedon, Chief Revenue Officer at Corvus Robotics. Sales and go-to-market sit with me, which in practice means most of my week is spent with warehouse and supply chain operators working through what's breaking in their buildings and whether we can help. Before Corvus I spent my career in supply chain technology, including project44 and Overhaul.
Working inside a freezer warehouse is dangerous and undesirable. The Global Cold Chain Alliance cites cold storage turnover at 32.6%, with some regions as high as 38%. This sounds like a perfect opportunity for robotics and automation to step in, correct?
Correct. A freezer is a hard place to work, and counting inventory is one of the worst jobs in it. Picture the setup. Freezer suit, hood, heavy gloves, then up 40 feet on a lift to read a label and key a count into a scanner. Cold takes your fine motor skills first, so the task that demands the most precision is the one you're least equipped to do well once you're up there. The cold makes everything around you more dangerous too, because stiff hands and bulky clothing don't mix well with working at height. Counting happens to be the piece of that work we can hand to a robot today.
What challenges did you have to overcome when building autonomous aerial systems that need to live and operate non-stop 24/7 in sub-zero environments?
Cold breaks a lot of hardware that runs fine in an ambient building, so Corvus One for Cold Chain meant re-architecting a system that already worked. Thermal management came first. Batteries lose capacity in the cold, and the electronics have to stay inside their operating window while the air around them sits at minus-20°F. Perception was next. Frost and condensation settle on labels, ice and wet concrete throw glare, and freezer aisles are often dim. The system uses adaptive focus and exposure control to keep reading labels the cold has gotten to. Flight stability turned out to be its own project. Freezers move a tremendous amount of air with blowers running constantly and dock doors opening and closing all day. The drone has to hold position and stay accurate through all of it instead of waiting for calm conditions. Then there's endurance. All of this has to keep working through months of continuous operation inside the box, at the accuracy that makes a count worth running. It took real creativity and a lot of testing to get there, and our team never wavered.
So Corvus One drones are fully autonomous? You can set up in a warehouse or freezer without any sort of localization beacons or stickers, or even a Wi-Fi network?
Right. No beacons, no stickers or reflective markers, and no Wi-Fi network to fly. Corvus One is a vision-only system. The cameras feed our CO2 World Model, a robotics foundation model we trained on millions of clips from real warehouse environments, and all of the perception and decision making happens onboard the drone. That's what lets it build its own understanding of a facility, know where it is in an aisle without GPS, avoid a forklift that wasn't there yesterday, and read a barcode 40 feet up with nobody flying it. It also means that if the customer needs to re-slot or move racking around, the system can adapt.
This seems like a perfect job for a robot to do — on its own, around the clock, with no operator and no person in the freezer. Tedious, dangerous and time-consuming work is exactly where robotics and automation fits best.
Agreed. Sending workers into the freezer to count is a job that's hard to fill and harder to keep filled, and the people who do stay are usually the ones you'd rather have running the operation. A robot does it better and safer, and it does it far more often than any manual program could sustain. The economics catch up to the safety case quickly. Set a subscription against what counting costs a site today in labor and lift time, and the payback shows up fast.
Manual inventory counting in sub-zero environments is miserable and error-prone work, and the safety of workers shifts must be short and efficient. How much of the inventory count can be done by Corvus drones, and what level of human support is required to set up and manage the Corvus Cold Chain drones?
The drones do the counting. Everything in the racking they're assigned to cover, at whatever frequency the operation wants. Corvus One for Cold Chain flies on its own from minus-20°F to ambient, during active shifts or lights-out, with no pilot and nobody swapping batteries in the cold. What people still do is the part that takes judgment. The system flags discrepancies and attaches photo evidence of the location, so your inventory control team spends its time resolving real exceptions in the WMS rather than riding a lift to confirm what the system already believed. On setup and management, our team handles the deployment. From there, every customer gets a dedicated customer success manager who owns performance of the system and works closely with the site to adjust coverage and cadence as the operation changes. Nobody at the customer inherits a fleet to administer.
What about maintenance? How do customers maintain and support the Corvus Cold Chain drones?
It's included. Under our Robotics-as-a-Service model, Corvus covers all maintenance on the drones, spare parts, and full drone replacement when and if it comes to that. Battery and device health are monitored continuously, so we generally see a problem developing before the site does. We also train associates on basic drone maintenance and best practices for the hardware, which covers most of what comes up day to day. If an associate can't fix a drone on site, we overnight them a new one at no cost.
The Corvus Cold Chain system is already deployed in live operations with several customers. What kind of impact are you seeing?
The clearest impact shows up in what the counts find. At one site with a national foodservice distributor, running both freezer and ambient racking, the drones surfaced roughly $2.1 million in inventory discrepancies across a six month window. Those are inventory locations the WMS had wrong, caught early enough for the site to do something about it instead of discovering the gap when an order came up short. Frequency drives the rest of it. That building was counting reserve inventory quarterly before we arrived. It now gets counted about every three weeks, faster in some parts of the building, so a discrepancy surfaces roughly four times sooner than it used to. That gap matters more in cold chain than it does in a dry warehouse because product with a shelf life doesn't wait around for the next quarterly count.
The content & opinions in this article are the author’s and do not necessarily represent the views of RoboticsTomorrow
Featured Product
