The need for fast and accurate order fulfillment has led to numerous automated applications. Pickr.AI has taken a novel approach to piece picking with its PickrMate robot, which performs item recognition, grasping, transportation, and delivery along a wide area.
The Stavanger, Norway-based company said its “3D picking concept” also utilizes vertical space while requiring less warehouse footprint than conventional robotic conveyors, reducing risk and saving money. Pickr.AI promises to provide “friction-free access” to robotics with rack, goods-to-person, vertical lift module, and automated storage and retrieval system (ASRS) integration.
Pickr.AI also aims for “high predictability, continuous operation, and reduced labor dependency” with zone picking and standalone order-picking cells. The company is looking for partners, engineers, and capital.
Mobile Robot Guide spoke with Roald Valen, CEO of Pickr.AI, about the company’s offerings and approach to small and midsize enterprises (SMEs).
Building a career in robotics and AI
How did you get into robotics and AI? How did your prior experience in R&D and drilling inform your current position as CEO of Pickr.ai?
Valen: Growing up next to Trallfa — now ABB — when it made the world’s first painting robot, that got my attention quite early, being a tech nerd.
I started at ABB in 1994 and developed new control systems. I also had the chance to go to Japan to be a representative for our products and competence for the Asian market. Once, I sat at a table facing Toyota and telling them that the reason for hiccups in the system was due to radiation from space.
At the time, the memory chip we used didn’t have auto-refreshing to see if any bit was changing without any commands being given. At the same time, everyone started to release chips that fix that automatically.
Explaining that in a reasonable way to Toyota and how it would never happen again was a good experience regarding how quality matters for a big customer.
I later jumped over to the oil business at drilling equipment provider National Oilwell Varco, or NOV, which delivers drill floors for oil rigs worldwide. That was a completely different ballgame. Then I was given the opportunity to join a startup that was going to make a seabed rig for exploration with robots.
How did you move from marine systems to supply chain robotics?
Valen: Industrial robots are marvelous at repeating what they’ve been told to do, but we couldn’t reach a drilling rig on the seabed with people or rely on remote control, so we worked with Energid Technologies using its axis-control system. Energid is now part of Universal Robots.
The work also took me to Stanford University to make a hand to be used for the subsea robots, and to NASA JPL, to meet the Mars rover team about autonomous robots. We needed robots that could slow down rather than go into a critical stop, like the Toyota case. We also had BostonDynamics‘ BigDog as an inspiration because it could handle unexpected situations like being pushed and stay in action.
From 2014 to 2020, I was in different business-development positions in various companies, including building up an intelligent automation department in Capgemini, a consultancy. It focused more on robotics process automation, AI, chatbots, and big data to get autonomous IT systems.
Then, early investors in Pickr.AI reached out to me and asked me to help lead it because its co-founder and former CEO had left the building. There has been a lot of change, and for me, it’s back to building physical robots again. This is an opportunity to build a tech business that could be an international success.
I’m still a newbie to warehouse automation, but as I say to warehouse customers, I don’t sell robots; I sell improved processes. Selling a robot is like I’m selling you a CPU that you can’t use without a motherboard, a screen, and a keyboard. It can’t help your business unless you have all those things.
Register today to save 40% on conference passes!
Pickr.AI takes on piece-picking challenge
Why has piece picking been so difficult for robotics?
Valen: Amazon said picking was a “holy grail” of automation, but something happened when the latest wave of AI came to play. Suddenly, the algorithms got better, the CPUs were more powerful, and the data was more available.
Computer vision is central to seeing the products to be picked, but the complexity comes with actually gripping something.
The connection to the object or SKU is still a challenge, and there are a lot of different griping mechanisms. The suction cup is currently the most efficient, but if it picks up two items at once or takes more time, that’s not cost-effective. We and all the other robot providers are working on improving these types of solutions.
How is PickrMate different?
Valen: If you look at our picking robot, it’s made so it can reach bins from shelves on both the front and back side. Our robot can also move bins, which means you do not need additional automation to move bins around — whether it’s a conveyor, an ASRS, or an AMR [autonomous mobile robot].
Our robot can of course pick from conveyors and cube storage like traditional picking robots, but it also does zone picking and vertical lift modules. And since our robot has an extreme working range, it does not need a complex bin-transportation system like conveyors on the order-bin side. That’s something traditional picking robots need, adding a lot of cost and complexity.
In a traditional bin-picking system, if I order 10 items, the bin needs to come around 10 times, which is complicated and takes a lot of space for AMRs or snaking conveyors. But we can handle bins with one arm in a gravity-fed rack 25 to 30 sq. m [269 sq. ft.] versus 250 sq. m [2,690 sq. ft.] in space with, for example, 500 bins.
My racking costs about $30,000 U.S., while a conveyor setup could be up to $625,000. That’s ideal for businesses with a few thousand orders per day.
Pickr.AI builds its own tools
Many robot providers have shifted from trying handle pose estimation by programming CAD designs for each SKU to saying that they’re not trying to replace the human hand but rather how objects get from one place to another. How do you see the competition?
Valen: Berkshire Grey, RightHand Robotics, and others have tried, but first there’s the risk of implementing robots in your e-commerce process.
Covariant, Plus One Robotics, XYZ Robotics, and Kindred have all made robots that fit into larger systems. They’re like making CPUs for IBM or Apple — they can’t move around.
If you install a robot to do all the picking, if it stops, production stops. The remedy is to have a parallel manual station, but that can also become quite complicated in terms of expense and space if a stop diverts items to manual processing.
How does PickrMate address these challenges?
Valen: Pickr.AI has three different solutions based on our core concept, and we’re actually delivering on the first one. We start picking from storage bins in traditional vertical lift modules, or VLMs, to order bins in a passive racking system.
It enables customers to get orders picked while they are home sleeping or enjoying the weekend. And when the workers return to the warehouse, they can command the robot to move away and use the same port to replenish the VLMs.
This is very space-optimized, and passive racks are cheap, so if you want reconfigure them, it’s easy to do. It also works with any other automation, such as cube storage or AGVs [automated guided vehicles]. We offer our customers the option of capitalizing on existing storage investments by expanding order picking outside normal working hours.
It’s like a pick wall but with a robot on a rail. Also, for risk mitigation, you can move the robot away at any time and do manual picking, which you can’t do with a stationary robot. You can also use this for nighttime and weekends, and keep manual picking during the day, helping SMEs with labor shortages.
What are some of your technology partners?
Valen: We had a grant from Eureka, a Eurostars program, and then got involved with the University of Eindhoven, the University of Reykjavik, and the University of Stavanger. That led to our pilot installation, which demonstrated a fully automated 24/7 order-picking station.
The pilot led us to design the newly released piece-picking robot, PickrMate. First, we tried using standard robotic arms, but we soon realized that it was not a fit for our SME customers, so we made our own arm. This includes a very compact picking tool, as we are limited in space compared with all the standard picking tools on the market. This was all of for the sake of customer benefits.
For order buffering with a robot in the middle of two flow racks, it needed to be very compact and posed a real-time coordination challenge.
PickrMate to work with wider range of WMS, SKUs
How much integration is necessary for “Robot @ Goods” to work with other processes and warehouse management systems (WMS)?
Valen: WMS integration is a challenge, and we’ve already integrated two types. There’s work to be done, but it isn’t the biggest challenge because we’re developing generic functions, so we can focus on the APIs [application programming interfaces] afterwards.
Since the robot is acting like a human, we’re mainly working with customers on the splitting of responsibilities between their IT systems and our robotics. We’re trying to reduce the need to make changes in the customers’ IT systems.
What is the range of items that Pickr.ai’s systems can pick? PickrMate was updated in December — were the upgrades in response to customer requests or part of your roadmap?
Valen: Today, we can do up to 2 kg [4.4 lb.] and 25 cm [9.8 in.], and we’re already moving empty bins weighing 3 kg [6.6 lb.]. We know that bigger items means that you have fewer items in storage, and it doesn’t make sense to have only one or two SKUs in a bin.
Our solution isn’t made for sorting multi-SKU bins but to fetch the same types of items. The robots can pick from shelf to shelf — what we call 3D picking — rather than 2D robots picking from the floor.
These 2D picking robots are replacing workers in pick stations designed for humans. This means the humans have limitations in reach and height, which together with ergonomic requirements dictate the design and processes built around today’s pick stations. Swapping the humans with 2D picking robots only swaps the uncertainty of the humans, keeping the same process.
We started with the power of robotics and designed the picking station around it. Given our large working range on a small footprint, we ended up with a 3D picking solution. In fact, we removed the whole picking station and put the robot in the storage and order-buffer space instead, resulting in the “Robot@Goods” concept.
Global demand still growing
Which markets do you get the most demand from, and where would you like to expand?
Valen: We’re focusing on the Nordics first — Norway, Sweden, Finland, and Denmark. Plus, we have some interest from Germany and the Netherlands.
We’re receiving interest from industries like wholesale/distributors, or B2B [business-to-business]; retailers for B2C [business-to-consumer]; and 3PLs [third-party logistics providers], ranging from grocery, health and beauty, pharmacy, electronics, and meal-kitting markets.
Our first delivery of a system will be to Veso Apotek, a veterinary pharmaceutical company. With this installation, we will show [our capabilities to] other customers and future integration.
What areas are you still working on, and what would you like prospective customers to know about your solutions?
Valen: We’ll have our first reference installations up and running in Q3, so companies can be convinced that our system works.
Some providers pitch ASRS first, but smaller companies should start with picking and packing processes to figure out their capacity, and then they should find the automation that facilitates them.
Pickr.AI has the elegance of a robot on a rail with picking arms that can extend the capacity of human pickers. I think we have something that’s a bit different, and it can actually help businesses get started with robotic picking, which doesn’t need to be complex or scary.
