Warehouse technology is getting better, faster, stronger — and smarter. Innovations like barcode scanners, microfulfillment centers and warehouse management systems have helped warehouse workers become drastically more efficient.
But with a labor crunch leaving many warehouse operators understaffed, they need technology that does more than support the work of people — it needs to do some of that work itself.
Capable of handling complex warehouse tasks on their own, AI-powered robots are beginning to permeate the space. Hundreds of businesses unveiled the latest in warehouse robotics and automation at last month’s Modex conference in Atlanta, a trade show where companies demonstrated robots that could pick, pack, sort and everything in between.
Walking around the convention center, a warehouse operator interested in deploying robots would have been spoiled for choice. But therein lies the problem: With such a wide range of warehouse robots available on the market today, how can operators pick the right one?
“I think we’ve gotten beyond the point that a lot of companies now know they’re going to have to invest in robotics and automation technology,” Jim Lawton, vice president and general manager of robotics automation for Zebra Technologies (NASDAQ: ZBRA), told Modern Shipper. “But they’re still unsure about, ‘Where do I start? How do I start? How do I make sure that this thing doesn’t fail?’”
Modern Shipper is here to break down what warehouse operators should — and shouldn’t — keep in mind when picking out a shiny new robot.
Function over form
Of course, there are still many functions that humans can carry out more effectively than robots — like piece picking, for example. But even within the realm of picking, robots can make things more efficient when deployed correctly.
For instance, case picking and putting away, which typically involves the movement of bulky or heavy items, can get a boost from robotic palletizers or depalletizers, machines that are designed to load and unload cases onto or off of a pallet. They typically handle materials that are difficult — or dangerous — for a person to carry.
In other cases, the robot can simply guide the picker. When picking e-commerce orders, warehouse workers can be assisted by robotic put walls or mobile shuttles that tell them what to pick using LED lights or some other indicator.
In the case of picking items that need to be sorted, a stationary robotic arm might work best. Some arms are designed to pick and sort goods of the same shape and size, while others are advanced enough to pick irregularly shaped objects out of a bin or tote.
Some robots are even capable of carrying out picking in several different scenarios, like those used by Target.
“Target is replenishing their stores out of their distribution center or warehouse, but at the same time, in some cases, they can be fulfilling e-commerce orders as well,” Lawton said.
But according to Lawton, function is just one of many considerations warehouse operators must make when choosing which robotic solution to deploy.
Location, location, location
Setting is an equally important factor when selecting a warehouse robot. Some facilities are massive, multimillion-square-foot behemoths; others occupy just a single room in the back of a store. Chances are that one robot won’t fit both.
“What’s good in a distribution center as an automation module, whether it’s to pick, pack, sort or transport — that’s not relevant for back of store,” Sridhar Solur, chief product officer and general manager of mobile robotics for Berkshire Grey, explained to Modern Shipper.
It wouldn’t make much sense to deploy a fleet of autonomous mobile robots — or AMRs — in the back of a store, where there isn’t much room to operate. In the same way, it wouldn’t be particularly efficient to use one or two AMRs to cover a massive distribution center.
In a microfulfillment center or the back of a store, a warehouse operator might use a stationary robotic put wall that sorts goods for the worker to pick. In larger spaces, where workers can spend up to 60% of their time walking, according to Lawton, operators might invest more in robots that can move materials to let workers focus more on picking.
The layout of a facility can also be important. In a facility where order volume and product categories remain stable over time, an automated guided vehicle (AGV), which travels along a set path rather than using AI- and sensor-based navigation, might suffice. For example, Lawton pointed to warehouse operators that are replacing traditional conveyor belts with guided robots.
“If you’ve got something that’s pretty structured, pretty well defined, where you want the robots to go to the same place every single time, then AGVs work out pretty well,” said Lawton, contrasting them with AMRs.
“[AMRs have] got sort of this mapping capability, navigation capability detection capability. And so as a result, they’re much more flexible in environments that have more sort of change going on.”
But setting is about more than the size and shape of the facility. It’s also about the kind of activity that takes place.
For example, a good candidate for an automated sortation and retrieval system (ASRS) is typically not a good candidate for a goods-to-person model.
An ASRS might be a good fit for a facility that moves high volumes of inventory but is struggling for space. Those systems are typically capable of shuttling horizontally and vertically across tightly spaced shelves where a person couldn’t fit, allowing operators to increase their storage density.
Meanwhile, a goods-to-person model, which is predicated on bringing the item to the picker, would be better suited for a fleet of AMRs that can allow pickers to remain in one section of the warehouse.
“If you start to get to the point where you’ve got, say, 20, 30, 40 pickers and a warehouse of some scale with a certain order volume, then you really want to move into more of the zone-based picking,” Lawton advised.
At the same time, a person-to-goods model, the traditional method of picking that necessitates the picker traveling to the item, might not lend itself to robots at all.
Automate at your own pace
It goes without saying that not every warehouse operator is in a place where they can spend thousands or millions on robots. Luckily, there are ways for less established companies to dip their toes into automation without breaking the bank.
One option is to deploy a robots-as-a-service (RaaS) model, which is essentially a subscription service for warehouse technology. Warehouse operators and 3PLs can ink contracts with warehouse technology companies that allow them to use their robots for a set period of time, rather than buying them outright.
“Many of the people that are buying these solutions are 3PLs, and 3PLs are trying to match their investment in technology with the contract duration of their customer,” reasoned Lawton.
More often than not, facilities that are just starting out also lack the complex software of a more established warehouse. But there are ways to get around that too — Lawton recommends deploying robots with workflows that can be triggered by the operator, such as with a button.
“You can have a physical button, like the Staples Easy Button. And when you push the button, the robot comes, takes the trash and then replaces it,” he explained. “You could also have it triggered via a barcode, so if you’ve got a barcode scanner, you scan the barcode and that basically says, ‘Hey, send a robot to this location because you need to take something away.’”
But as a warehouse grows, the robots need to grow with it. At a higher scale, when companies begin integrating more advanced softwares into their facilities, it makes sense to deploy robots that are smarter.
For example, if an operator’s software (and budget) allows it, it might be prudent to utilize AMRs capable of detecting and navigating their environment rather than AGVs. Because they’re intelligent enough to link to a warehouse management system or other software, AMRs can continuously improve their performance by feeding data into the system, which collects it and transmits it back to the robot over time.
Businesses at a higher scale might also benefit from robotic solutions that can be deployed in different settings, such as in a distribution center as well as in the back of a store. Those settings are wildly different, but operators can connect robots across facilities using what Solur calls transfer learning.
“You can’t have one vendor for here, one vendor for there, and then a spectacular array of interfaces that have to work together. That’s really, really hard,” he explained. “So if you can have these meaningful Lego blocks, these bots that can work in both places and scale as needed under a common orchestration layer — bingo.”
Watch: Warehouses of the future depend on teamwork
Transfer learning works as a sort of hive mind, allowing a business’ robots to communicate new information to any other robot in their network. Over time, that leads to yet another benefit: the ability to stand up new facilities without having to reprogram or reconfigure the robots.
“We call it the Benjamin Button effect — our robots actually get younger and younger over a period of time because they understand the winning strategies,” Solur added.
Typically, a scaling business will also contend with facilities that are constantly in flux. As new verticals and functions are added to a warehouse over time, having a flexible infrastructure becomes essential. Lawton pointed to the example of static conveyor belts, which are gradually being phased out in favor of mobile conveyor robots.
“If you want to be making continuous improvements that involve changing the workflows around, it’s very hard to do when you’ve got traditional heavy metal infrastructure in place,” he said. “So I think in general, people are looking for more flexibility, the ability to do more with the investments that they’ve already made.”
Testing the waters
Lawton and Solur each highlighted some ways that warehouse operators can get an idea of how a robot will fit into their facility before they actually deploy it. In Lawton’s view, it could be as simple as looking at what similar warehouses are doing.
“If you’re looking at implementations and videos and testimonials in warehouses that look a lot like yours and your 3PL’s, for example, chances are you’re going to be able to get the same kinds of gains out of them,” he said.
Lawton also suggested that businesses start by automating a single facility, using it as a sort of pilot. But there are ways to get an even clearer picture.
“We have our own mock warehouse facility inside Zebra,” Lawton said. “There’s the ability to build ROI models and payback models leveraging some standard assumptions that have been verified and validated with other companies doing similar things.”
The ability to create lifelike simulations of the conditions within a facility is something Solur refers to as digital twinning. By collecting data from existing facilities and systems, certain softwares can construct a digital twin of a real-world warehouse, accounting for everything from the shape and size of the building to the type of dust lining its surfaces.
“I would say a high-fidelity, bespoke digital twin is absolutely needed before you do anything,” Solur advised. “But to have the ability to do that, you have to make sure that the dichotomy between the physical world and the digital world is not too huge.”
Simulations like digital twins can be a useful tool for warehouse operators looking to see where robots could unlock new efficiencies in their facilities. Of course, they shouldn’t expect robots to fit in everywhere. But with order volumes climbing and warehouse staff declining, a little automation can go a long way.
“Automation will serve a subset of the needs within a warehouse or a subset of the needs within the factory or the manufacturing environment. And that’s OK,” Lawton explained, “because you don’t have to automate everything to get a whole bunch of value.”