Warehouse Robots are Here

Robots in the warehouse have been getting a lot of attention in recent years. This was evident at the 2023 Promat trade show I recently attended. I felt a little sorry for any exhibitor that didn't have a robot to show off; they were not attracting attendees. We're seeing a perfect storm for robots in the warehouse. There are numerous companies offering working robots for the warehouse; at the same time, we have been experiencing labor shortages in the warehouse. This means everybody is looking at robots. They may not be ready to write a check, but they are interested.

Just a few decades ago, I recall thinking warehouse jobs were pretty safe because the types of robots being used in manufacturing weren't anywhere near capable of doing the tasks that go on in the warehouse. While warehouse tasks were repetitive, they weren't THAAAAT repetitive.
And as robots became available that were designed for warehouse tasks, I always looked to see if any of them could go to a bin on a shelf with 3,000 screws in it and count out 100, put them in a bag, and then go to another shelf and pick up a 50-pound box containing an electric motor and deliver these to a pack station because that is what a robot would have to do to replace an order picker in a warehouse.

Well, I still haven't seen a robot that can do that. But they are getting closer.

I've embeded some videos in the article to show the operation of each robot.

It all started with Kiva and Amazon.

Kiva Robots first started getting attention in the 2000s with their squat little orange robots that were essentially a more sophisticated AGV. They used small shelving units that could be picked up by the robots and delivered to a picking station. I can't say that I know this for certain, but I think this was the first time this specific approach was used. So Kiva was selling systems and getting some attention in the trade magazines, but nothing substantial. Then Amazon started using them, and everybody was talking about them. Then Amazon purchased Kiva, and the industry went crazy. And when Amazon decided to stop selling Kiva Robots to other companies, it opened up a market for other warehouse automation companies to step in with their Kiva clones.

Collaborative Robots.

The term "collaborative robot" is typically used today to describe a small subset of warehouse robots that basically assist warehouse workers by transporting orders to locations that have something that needs to be picked. I'll cover these in more detail later. In reality, almost all of the popular warehouse robots today function as collaborative robots. Meaning, they need humans to complete part of the tasks, so the robots are collaborating with humans rather than simply replacing humans.

This changes the thought process behind warehouse robots. It's really hard to design a robot that can pick those 100 screws out of that bin and then grab a 50-pound motor, but what if the robots can do other things that make it easier for a human to pick the order? Order pickers can spend a lot of time traveling between locations to pick items for orders. What if we used robots to eliminate that part of the order picker's job?

And that is exactly what Kiva did. They didn't need to pick 100 screws from a bin; they just needed to deliver the bin to someone who could. And when you have as many people as Amazon has walking around their warehouses picking orders, that sounded like a great idea to them.

In the case of Kiva, they didn't create a robot that grabs a particular bin and delivers it to a pick station; they designed a robot to pick up the entire rack that contains the bin (along with other bins) and deliver the whole rack to an order picker at a fixed picking station. This is a simpler task than having a robot pick a bin off of a shelf, but it also provides some storage advantages when it comes to the racks.

So rather than figuring out what the ultimate order-picking robot would have to do, and then figuring out how to build that. They took another approach. Let's see what we can easily do with robots, then see how we can fit that into an order-picking process that reduces the amount of work humans have to do. They basically took the easy route by doing something they had a high degree of confidence they could make work.

Amazon using the Kiva robots and then buying the company (Kiva) got the ball rolling in a big way. This convinced the industry that robots in the warehouse is viable now, but it also convinced a lot of people that the Kiva-type robots were the future, simply because if Amazon is doing it, it must be right. And this is why we've seen numerous warehouse automation companies making their own version of the Kiva robot. When Amazon purchased Kiva and decided to stop selling the robots to other companies, of course, someone else would step in to make a similar robot to fill that demand. I don't know how they got around Amazon's patents, but I'm guessing they found a way.

The Kiva robots were not anyone's vision of the future of warehouse robots—certainly not mine. So having a bunch of companies manufacturing robots that basically do the same thing is not exactly moving us forward. But the Kiva robots were a necessary step in warehouse robotics. And fortunately, there are now a lot of companies doing more innovative things with warehouse robots.

Goods-to-Person Order Picking.

The Kiva system is a Goods-to-Person Order Picking approach. Check out my article on Goods-to-Person Order Picking.

Goods-to-Person is not new, but of course, since Amazon started using it, everyone thinks this is the new be-all-end-all way to go. And subsequently, you see numerous robotics businesses building on this approach. Well, there's that, and the fact that this leaves the tricker part of picking to humans. So let's start with robots designed for goods-to-person picking.

I've already covered Kiva. As I said, there are now numerous similar systems.

The Basic Kiva Process.

Items are stored in small racks, approximately 6 feet tall. The items may be in bins or stored directly on the shelves. An item may be in one or several locations in the racks.
The racks are then lifted by the robots and moved into a warehouse storage area, where they are essentially just dropped in front of other racks to fill the space up. The robots drive under the racks to lift them.
When an item is needed for a pick, the Kiva system sends a robot or several robots to dig the rack out. They may have to move several racks to get to the needed rack.
When a robot gets the right rack, it will deliver it to a pick station. There may already be several racks waiting at the pick station to be picked, so it may have to wait in a queue.
When the pick station has space available, the rack will be moved right up to the pick station so the item can get picked.
A light-directed system or computer screen is likely used to identify the correct bin to pick from, and a display will show the quantity to pick.
There are typically multiple orders being picked at a time at a pick station. A put-to-light system would likely show quantities above each tote representing the quantity of the item to put in the order. The picker would then push the button for that quantity in that tote. This is the same process that has been used for decades with horizontal carousel picking.
When all picks for a tote are complete, the put-to-light system will direct the picker to forward the tote, probably on a conveyor or via other robots.

Other comments on Kiva robots.

I just want to clarify that I include all the Kiva clones in this category. I'm sure the manufacturers of these don't like them being referred to as Kiva clones. And some of them certainly look like improved versions of Kiva. But that's not saying much since the original Kiva robots were not all that sophisticated and looked a bit clunky. However, they all copy the Kiva process of having squat little robots pick up racks and deliver them to pick stations.
Space utilization is tricky. The robots will stack the racks in pretty tight, so you don't really dedicate much space to aisles, but the racks are short. You would need to use a mezzanine to take advantage of tall ceilings, and mezzanines are expensive and complicate the processing of orders. You also tend to have a lot of empty space in those racks.
Ideally, you would put everything in the Kiva Racks. Then you don't need to incorporate zone picking or something else to deal with picks coming from somewhere else.

Autostore System

This is another goods-to-person system that has been gaining in popularity in recent years. Like Kiva, Autostore has been around for a while (at least ten years), but this is no copy of Kiva. Autostore works off a physical 3-dimensional grid. Totes are stored in the grid and accessed using robots that operate on top of the grid. When a tote needs to be picked, the Autostore system sends a robot to the access point above the totes. The robot lowers a carrier that grabs the tote and pulls it up to the top. Then it delivers it to a fixed picking station.

The basic Autostore process is similar to Kiva's in that there are pick stations that pick multiple orders at a time, and items are delivered by robots to the pick stations.
Even though the basic goods-to-person process is similar to Kiva's, the Autostore system has some distinct differences.
Since the robots are captive to the cube, the picking stations need to be attached to the cube, or another means is necessary to transport the goods to the picking stations.
Where the Kiva system may need to move other racks to get at the one needed, the Autostore system may need to remove other totes that are stacked above the one that is needed. So both systems spend a lot of time moving things out of the way.
The Autostore system delivers just a tote, not an entire rack. It is possible that multiple items are stored in that tote, though.
The Autostore system requires some physical infrastructure. Primarily, it requires a physical 3-D grid to be built. This makes moving it to another warehouse more difficult.

Other comments on Autostore.

Even though the Autostore system claims to store much more stuff in your warehouse (I think they claim to quadruple your storage), you have to do your own math on that. Clearly, there is an advantage of not needing aisles in the storage area, but there does need to be some space around it. And the Autostore system is based on storing everything in totes. If you have small items and basically dump them in a tote filling it up, you will get pretty good space utilization. But if you just put a physical carton of something in the tote, that may leave a lot of air space around the carton. They do have some different-sized totes, but unless your stuff fits perfectly in those totes, you will have wasted space. This is a consideration for whatever you use to store your inventory. You have to look at how you will use it to determine how much of it will be fully utilized.
Also, there are currently limitations on how high they will build the Autostore system (currently around 18 feet), so you may end up with a bunch of unusable warehouse space above the Autostore system.

Tote-retreiving Robots.

There are numerous companies that make robots that go into a storage area, grab a tote from the shelving, and deliver it to a fixed picking station where a human can pick from it. These perform similarly to a mini-load ASRS, but they do have some advantages over an ASRS.
With a typical ASRS, the crane is captive to an aisle; however, these robots can move from one aisle to another.

With an ASRS, the pick station is typically right at the end of the aisle; with robots, it can be located somewhere else because they can deliver it across greater distances.

You may be able to use these robots with your existing racking system. However, that is likely to limit what you ultimately get out of a system like this. I would assume you would be better off designing a storage system that best utilizes the minimum aisle widths required for the robots, and storage methods that fit the totes being handled.

On the downside, some of these robots are height-challenged and can't pick at the heights of an ASRS (though some models can go pretty high). Also, the ones I've observed seem to move slower than an ASRS, but that could just be the way the demo is set up.

A robotic system that can actually pick an order without a human.

The Brightpick Autopicker is the first robot I've seen that can actually pick an order. This robot is similar to the tote-retrieving robots previously covered, but it carries a second tote that contains the order being picked and has a robotic arm on it that picks from one bin and drops it in the other.
So the basic process is the robot goes to the location and grabs the tote that contains the needed item, then uses the robotic arm to pick the item from the bin and place it in the bin that contains the order. Then returns the tote to the rack and moves on to the next pick.

I'm guessing there are many limitations to exactly what this thing can pick. I think it probably can't pick the earlier example of 100 screws out of a bin of 3,000. But, if you have the right item and order characteristics, this thing looks like it can actually do the job of picking orders.

I recently viewed another video by the same company that showed the system without the robotic arm. Instead, it brought both totes to a pick station with a human that would pick from one tote into the other, which was a disappointing and incredibly slow option.

"Collaborative Robots" Part II

The term "collaborative robots" or Cobots isn't new, but Locus Robotics is the first company I encountered using the term "Collaborative Robots" to describe their robots, and they've taken a somewhat unique approach towards order picking with them. They are not a Goods-to-Person system like almost all other warehouse robot systems. Instead, both the robots and the pickers are moving around in the warehouse, but not together. The robot goes to a location where its next pick is located and then waits for a human to come by and pick it for them; then, it moves on to the next pick, which will be picked by whatever human is closest.

These robots are picking multiple orders at the same time, using put-to-light to direct the picker to the correct tote to place the item in. They have a screen on the robot that tells the picker what to pick.
When all orders are complete, they can deliver them to a pack station.

This doesn't eliminate all of the travel time of a human picker, but it should be able to increase the number of picks each person is processing.

Locus is not the only robotics company using this process.

Bipedal Humanoid Robots for Warehouse Work?

These are robots that are truly intended to replace humans. They have legs that walk, and arms with hands that grab stuff. At this point, I haven't seen any of these that quite perform to human standards, but they are getting better.

While the development of a bipedal humanoid robot certainly demonstrates the capabilities of a robotic company, I think this concept is seriously flawed. This concept assumes that a human is the best form factor for warehouse work and then tries to mimic it. As someone who has done a fair amount of warehouse work, I don't agree. Why limit your robot to the limitations of humans? We've had millions of years of evolution that was focused on hunting, gathering, and avoiding being eaten by predators. While some of this helps us with warehouse work, we are not the perfect form factor for it.

If I wanted to be the perfect robotic order picker, I definitely wouldn't have legs. Warehouses have flat concrete floors; wheels work better for traveling through them.
Then, why would I limit myself to two hands? Having picked a lot of orders back in my day, I can say I absolutely could have used at least one extra hand. In fact, I would give an order-picking robot as many arms/hands as I could fit on it.
And I can only reach up to about 7 feet. Why would I limit a robot to that? I want one that is taller or maybe can fly.

So I'm saying you need to design a robot for the specific task. I really don't expect that in the future, you'll walk into a warehouse and see a bunch of C3POs running around.

But then again, the concept of an order-picking robot may be flawed itself. Remember, almost all the other robotic systems for the warehouse use humans to execute the part of the tasks the humans do better than a robot. I'm sure robots will be more capable in the future, but maybe there will always be some advantages to humans.

Pallet Storage Systems.

There are numerous racking companies that provide shuttle systems to move pallets within high-density storage racking. These systems vary quite a bit in functionality but they basically attempt to make high-density storage more selective, kind of the best of both worlds.

Here is a Google Search.to give you an idea.

And here is a video of a fairly sophisticated system from Swisslog

Other types of robots.

Squid Robot.

Do you need a robot that can climb your rack, pick a carton off a shelf, then deliver it to a station at the other end of the warehouse? Well, the Squid robot does all of that. Now, it does require installing some tracks on your existing pallet racking, and it looks like it limits the spacing of the load beams to coincide with the robot, which is not insignificant, but damn, these things are cool. The applications for these are somewhat limited, but damn, these things are cool. Did I mention these things are cool?

Trailer loading/unloading.

Order picking ain't the only thing going on in the warehouse. Unloading floor-loaded ocean containers is very labor-intensive and is also very common these days. The Pickle Robot can unload cartons from an ocean container and place them on a conveyor for further processing, either by humans or another robot.

There are numerous companies that have solutions for trailer unloading (and trailer loading), and they are not all doing it the same way.

This is one of those applications where it doesn't take a whole lot of effort or money to implement and start using right away.

Sorting robots.

There are robots that are simply designed to sort or consolidate orders either by moving totes or moving the items contained in the totes. The ones that pick items out of one tote and sort them into other totes essentially work like a human using a put wall. These robots are built around a robotic arm with suction devices or grabbers in place of a hand. Others can sort totes or cartons, essentially replacing a sortation conveyor.

These are nice applications for robots because they are highly repetitive tasks and can be set up as standalone workstations.

Transporting Robots.

Then there are basic robots simply used for moving something from point A to point B (or C or D). Does this mean we don't need conveyor systems anymore? I hope so. Conveyer systems have served us well over the years, but they're also a pain in the ass. They take up space, restrict travel, and have limited capacity. This is an area where robots can do better. But can they compete on cost???? When you think about it, the Kiva system is just robots that transport objects.

Robotized Lift Trucks.

I'm guessing the correct term would be Autonomous Lift Trucks. I decided to dedicated a separate article to these. Go to Robotized Lift Trucks

Learning from other industries.

The automotive industry adopted robotics decades ago. But if you look in an automotive manufacturing plant, they still have a lot of human workers. So the lesson here is you have to be realistic about your application of robotics. Some applications are better than others. Welding has become a slam dunk for robots. They do it faster and better than humans, and nobody gets burn holes in their clothes. In fact, many other industries have started using welding robots due to their success in automotive manufacturing. Painting has also been a good application of robots. Both welding and painting are good choices because a robot can execute these tasks with greater precision than a human, and both are somewhat hazardous for humans. But when it comes to assembly, there are varying degrees of robotic use. In some cases, a robot can execute the assembly task; in others, a robot can assist a human, making the human more efficient; and in others, a human does the whole task.

I would like to note that the painting and welding robots are purpose-built robots, not humanoid robots using the same welding equipment or painting equipment a human would use. In fact, if you observe the robots used in automobile manufacturng, most of them just look like other machines in the plant. I'm not really sure what the difference is between a CNC machine and a Robot. I just read an article that attempted to explain the difference, but it didn't help me.

Early adopters pay.

Robots in the warehouse are still emerging technologies. And with emerging technologies, comes a high price. In fact, the greatest downside to warehouse robots is the cost. When you look at these robotic systems, most of them do what they were designed to do, but the big question is do they pay for themselves? Multi-million dollar price tags for some of these systems are tough to swallow, especially for something that could be obsolete in a few years. The robotic system suppliers are trying to get around this by offering robots as a service (RAAS). So you don't have to spend millions on buying a system; you can simply rent them. Every robot supplier I talked to at Promat was providing rental programs. I don't know the specifics of each contract, though I'm guessing currently, they are probably trying to make it easy to rent a robot.
There are a lot of advantages to renting robots. First, sometimes it's easier to get internal approval for renting versus buying. Next, depending on the contract, renting may take care of that "obsolescence" issue. In addition, renting may allow you to increase the number of robots you use during peak seasons. I wouldn't bank on that one because most businesses have peak seasons in the last quarter, so there may not be enough extra robots to rent at the time you need them.

Someone needs to babysit the robots.

Eventually, they will work the bugs out, but for now, you need to have someone around that is capable of doing things to keep the robots operating. That may mean picking something up that is in their path, unjamming them, helping them get back up. adjusting a load, stopping them from murdering everybody, etc.

The next generation.

Amazon is already working on its next generation of robots. I haven't seen anything particularly innovative from Amazon, but maybe they will surprise me. Over the next ten to twenty years, I would expect to see significant innovation with warehouse robots. I don't know what the warehouse robot of the future will be, but I'll be watching closely.

I'm very impressed with the out-of-box thinking that went into robots like the Squid or the picking process used by Locus. I don't think these are the ultimate solutions, but I'm impressed that they aren't just copying Amazon's Kivas.

Unfortunately, I would also expect to see many failures. Who remembers the Airtrax omnidirectional forklift? Twenty years ago, this thing was getting a lot of press. Since then, it has fallen into obscurity--I don't think they are being made anymore. My thought when I originally became aware of the Airtrax forklift was that it was pretty cool, looked difficult to drive, and didn't really solve any problem I had experienced in the warehouse. As an interesting related note, I noticed those same omnidirectional wheels on one of the robots from Locus.

Tips on ROI for automation

These are tips on automation in general, not specifically robots.

ROI will be tough if you can't keep the automation running near capacity for at least a full shift. Multiple shifts are better, but if you are a same-day shipper that picks all your orders in a 5-hour window, the automation needs to be dramatically better to pay for itself.
Sometimes automation gives you capabilities that allow you to meet your business needs. Think about my previous comment on picking orders in a 5-hour window; if the automation helps you to accomplish that, there is value there that may not show up in pure labor savings.
Streamline your existing processes first. One of my pet peeves is companies saying they absolutely have to do something when their human warehouse workers have to make it happen, but when they look at automation, they are more than willing to make compromises on their requirements.