Automate The Freight: Amazon’s Robotic Packaging Lines

In the “Automate the Freight” series, I’ve concentrated on stories that reflect my premise that the killer app for self-driving vehicles will not be private passenger cars, but will more likely be the mundane but necessary task of toting things from place to place. The economics of replacing thousands of salary-drawing and benefit-requiring humans in the logistics chain are greatly favored compared to the profits to be made by providing a convenient and safe commuting experience to individuals. Advances made in automating deliveries will eventually trickle down to the consumer market, but it’ll be the freight carriers that drive innovation.

While I’ve concentrated on self-driving freight vehicles, there are other aspects to automating the supply chain that I’ve touched on in this series, from UAV-delivered blood and medical supplies to the potential for automating the last hundred feet of home delivery with curb-to-door robots. But automation of the other end of the supply chain holds a lot of promise too, both for advancing technology and disrupting the entire logistics field. This time around: automated packaging lines, or how the stuff you buy online gets picked and wrapped for shipping without ever being touched by human hands.

Working the Line

Anyone who has ever worked in the shipping department of a manufacturer or distributor knows the irony of a job that’s both repetitious to the point of boredom and frustratingly unpredictable. Most of the time, goods coming down the conveyor toward your packing station are easily handled – pick the right size box from the selection of flattened cartons you have available, build up the box and tape the bottom flaps together, insert the goods, add packing material, seal the box and label it, and send it on down the line.

This can go on for hours, and then something can come down the line that causes all sorts of problems. It’s irregularly shaped, or too big for the biggest box you have. Or worse, it’s just over the size of one box but way too small for the next size up. Those packages can waste a lot of time while you improvise a solution, perhaps by cutting and pasting boxes into a makeshift carton. And all the while, the conveyor belt feeding your station is filling up with other, easier to pack orders.

Even when a normal packing line works smoothly, the waste of human capital on such tasks can be unconscionable. Yes, people are generally more dexterous and better at solving the puzzle of what fits where than machines are, and it’s true that manual jobs like these provide a living for a lot of people. But speaking from experience – I’ve done my share of temp work on shipping lines – it’s mind-numbing work that’s hard on both the psyche and the body.

In an attempt to address these and other problems, Amazon is beginning to roll out more automation to its already highly roboticized distribution centers. As Reuters somewhat indignantly reports, Amazon has been testing the automated system at two of its centers with plans to roll out dozens more, resulting in a loss of jobs. When you drill down a little further, though, you see that the net impact is perhaps 200 to 300 workers across all the distribution centers, each of which employs about 2,000 people. So while it’s certainly not good news for some Amazonians, it’s also nowhere near the robot job-pocalypse that Reuters implies with its headline – yet.

Cartons on Demand

Regardless of the hype, what Amazon is actually doing here is pretty interesting. Almost all of us have had the experience of getting an Amazon delivery in a box that makes absolutely no sense based on its contents. It’s incredibly wasteful, both in terms of the raw materials of the box and the packing, and in terms of the space that’s taken up by the outsized package in the planes, trains, and automobiles that will whisk it to your doorstep. But it’s also understandable; when a packer only has a few boxes to choose from, he or she does the best with what’s on hand.

The fix is to make the carton exactly fit the order, and that’s what the CartonWrap system that Amazon is testing does. Made by Italian company CMC SRL, the machine is capable of packing arbitrary orders into shipping cartons custom made on the fly. The machine accepts goods on an input conveyor where they are scanned and sized. At the other end of the line, a fanfold stack of plain cardboard is cut to the proper length, scored for folding, and die cut with a terrifying swinging guillotine blade. Various rollers and pneumatic actuators press and fold the precut flat sheet into a box around the ordered goods, sealing it and applying a label before sending it along the line.

The boxes formed by the machine are not quite the simple cartons we’re used to seeing from Amazon, but the company claims that the more complicated shape makes the box easier to open. You’ll also note that no padding material is added to the carton; the box is exactly sized to fit the order snugly. This would seem to preclude its use for some delicate items, but for books, small items, some foods, and soft items like clothing, it makes perfect sense.

It’s easy to see how CartonWrap could be integrated into a completely hands-off packing system. Rather than being fed by a worker, the input could easily be taken from one of the robotic pods that swarm about in some Amazon distribution centers. And similar systems could handle the outfeed from the packing line, collecting packed orders into pods that will be loaded into the backs of self-driving delivery trucks. With last-hundred yard automation, it could be that someday the first person to touch your order will be you when you pick it up from the doorstep.

Clearly, Amazon and other big shippers will not be able to completely replace human workers anytime soon. Some things are just too difficult to grasp mechanically or too delicate to be trusted to a package without padding. There will likely always be a need to staff a manual packing line alongside all the automation, but if shippers can even get to the point where 80% of the orders are handled by a lights-out warehousing system, the productivity gains will be huge.

[Featured images: Tucson Improvement and Beautification Org, Spokane Spokesman]

50 thoughts on “Automate The Freight: Amazon’s Robotic Packaging Lines

      1. Maybe, assuming they select the employees (like the author) that were wasted on the packing line.
        I predict some of this, as well:

        I hope their new programmers can manage to tell a robot what the humans should already be able to figure out (and still fail miserably). Like: ‘do not put a heavy sharp edged object loosely in a nearly empty large box with a fragile item’ or ‘thin cardboard retail box will not survive shipping, put the damn thing in another box.’
        (Three times they did the second one on the same order, before I gave up on calling them.)

        1. One time I bought a very large and heavy steel gear from Amazon then added a bluetooth CAN bus automotive code reader to bump it enough for free shipping. The gear came wrapped in bubble warp, inside a box, inside another box. That was tossed into a larger box along with the code reader and its software CD, in a plastic bag. There was no packing material of any kind.

          The well protected chunk of solid steel rattled around and broke the completely unprotected software CD.

          What was the gear for? To be bored out and pressed onto the back gear shaft of a massive LeBlond metal lathe from the 1920’s. The back gear shaft and its large and small gears were a single iron casting which was bored, faced, and had gear teeth cut. Several teeth on the small gear had been broken and fixed at least once, and some were broken again.

          I figured out the specifications and after feeling my wallet cringe while looking up prices on stock gears that size, I just on a whim plugged the specs into Amazon and up popped exactly what I needed, for just $35! Only 2 left! BUY! Mounted the back gear in my 1943 Monarch 12CK lathe to turn off the small gear, roughed out the center of the new gear on a mill with a large shell mill then finish bored it on the Monarch. Got the fit perfect so when pushed on with the 20 ton press there was no need for set screws or pins.

          1. Yes, Amazon’s packaging is haphazard at best, and generally poor. Your purchase rattling around in a large box with a couple of air pillows for company is the norm. If you’re lucky, it hasn’t escaped the box before it gets to you.

      2. Not everybody is suited for coding. Some are good for engineering, marketing, politics, teaching,general labor etc. There are people that you wouldn’t trust in your project as they would do more harm than good.

      3. Programming can be a good job but it’s not a bottomless pool of wealth. We certainly can run out of work just like any other industry. Why does it seem like EVERY time some profession becomes outdated programming is the first thing somebody suggest that they do? Are you trying to get us laid off or depress our wages?

        What we really need is more teachers, except that we need to be willing to pay for them first. How about redirecting some of the money that goes to bomb far away lands into schools and shuffling workers into that direction? We could stand to put some more money into basic research too. Lets discover the science that will enable tomorrow’s innovations. And if the displaced factory workers are not suitable to these tasks then they can take the places of people who are suitable thus allowing them to do so.

        1. “Are you trying to get us laid off or depress our wages?” yes and yes,

          It is also a foolish suggestion as programming is easily the most transportable job in the world and the market will shift to the lowest cost environment as it has been for decades. To the point that when I did my CS degree my professors were telling us that we should be considering a masters if we wanted job security, they were pushing people either further into academia (into the professor track) or into MBA’s and management.

          Dont get me wrong, in general people should be learning how programming and electronics actually works, if anything to help them make informed decisions on purchases and when to push back on what corporations say (Cough, Microsoft and its OSaaS, Cough), this goes triply true for anyone involved in politics.

          We dont just need more teachers but we need teachers to be held to a higher standard than they currently are. Teaching to pass standardized tests is not teaching, its doing the bare minimum to fulfill job evaluation metrics. You wont see actual teaching and that is by design, it is the basic and ongoing historical struggle of the accumulation of power vs the distribution of power. Good teaching distributes the power (knowledge is power), most of today’s issues can easily be framed in that exact struggle, from taxes, to healthcare to education.

      4. “Turn into coders.”

        LOL, right, “learn to code.” That was the spew from “journalists” to unemployed coal miners, but when that became a clever retaliatory meme used against the droves of corporate media “journalists” being laid off, suddenly it was banned on social media sites.

    1. A long long time ago some primitive human ancestor poked a termite mound with a stick or threw a rock at some prey animal and technology was born. Suddenly one person could produce more for the tribe and so fewer hours of labor were needed. This process continued for many millennia bringing us our modern world with so many comforts and labor saving devices that few of us would ever wish to be without.

      The fact is that when more can be produced easier there SHOULD be more to go around. We just need to get better at distributing it.

    2. The author doesn’t understand the packers job, boxes are suggested, not picked by a packer. Overall the way he talks about the job it seems like looked up outdated descriptions of the Amazon process.

  1. You don´t need to “scan and size” the itens. If they are already in your warehouse/inventory, you already know their size. They can be sent to the packing line with the box number/size already specified.

    It is an interesting engineering problem to solve, yes.

    But I somehow agree with Reuters. The problem Amazon want to solve it its bottom line. If the trend catches, more and more simple, common jobs cease to exist. Then government has to increase taxes over businesses to help pay unemplyment help/benefits/dont-know-how-its-called.

    1. Until the supplier changes the packaging, or you have multiple sources for an identical item, or someone makes a mistake in data entry or stocking. Also corrects for small size/shape deviations due to tolerances.
      Though it looks like the machine leaves a fair gap for items to rattle around inside the box, so point taken.

      Ultimately I think this method prevents both damaging the goods and machine stoppage (and also damage) due to trying to put too small a box around an item due to bad data.

      Lastly, I’ll give them a pass on ‘dinamic’ since they’re Italian, but how many people watched that video before publishing it with “lenght” in the overlay? Seriously…

    1. LOL, so true!

      I once worked for a company that stocked tons of printer supplies (toner, ink, drums, etc) but when one of our printers went down, we had a service company tech come in to do the maintenance and guess where he ordered items from…. us. So a drum would ship from our warehouse to his corporate office an hour away where would pick it up the next day and bring it in. 2-3 days to do a fix that would take 10 minutes if he could have just walked out to our warehouse.

      1. He didn’t have the smarts to realize he was at a printer supplies company, and to think he should try asking someone if they had what he needed in stock?

        For that matter, why wouldn’t a printer supplies company have at least one employee able to service the printers used by the company?

      2. I’ve been in almost exactly the same situation as a service tech, and I can tell you that there is a 100% chance that even if he had known that he could get what was needed from the next room, he wouldn’t have been allowed to, thanks to contracts, bean counters, and lawyers.

  2. “Regardless of the hype, what Amazon is actually doing here is pretty interesting. Almost all of us have had the experience of getting an Amazon delivery in a box that makes absolutely no sense based on its contents. It’s incredibly wasteful, both in terms of the raw materials of the box and the packing, and in terms of the space that’s taken up by the outsized package in the planes, trains, and automobiles that will whisk it to your doorstep. But it’s also understandable; when a packer only has a few boxes to choose from, he or she does the best with what’s on hand.”

    And now the other logistics problem. Careless shippers and their chain.

    1. I was under the impression that amazon owns the whole shipping chain. If that is not the case, I got another wondering : A shipper company would prefer some standardized package sizes, to improvei their packing eficiency in trucks/containers/boxes/whatever. If amazon starts making packages with a myriad of different sizes, then wasted space in transport rises again .

      Having just some package sizes is not just for simplifying employee choices, but also to facilitate piling and boxing. And I can see that removing padding material will contribute to increase customer dissatisfaction and returning of goods. Is dealing with that their next automation project ?

      1. Why? A robot will just sort packages Tetris style so they fit in a larger standard cube size perfectly. Fitting a smaller product in a larger cube just so they stack in perfectly sub-dividable units is just as silly.

        You could apply the same system to container ships. Standard 55′ shipping containers revolutionized global trade. Imagine a container ship but instead of using one-size-fits-all lego bricks, they were replaced by a myriad of Tetris style boxes that a computer algorithmicly placed before the ship ever pulled into port.

        1. Hmm, no ? The robot cannot waif forever until the necessary sized items come its way. With a pre-selection of standard sizes, the chance of it achieving correct packing is greater. With random sizes, not so much.

          1. You would be surprised at how well a large assortment of items with ‘close enough’ sizes will stack together, especially taking aaronfish’s suggestion below, and noticing that one dimension of the box in the video can easily be extended to help standardize sizes. (or all 3 with some padding)
            Also note that most items Amazon sells are not overly large.

            I worked for a year in the distribution center for a large home improvement store building pallets of product to send by truck. Size, shape, and the order of items coming down the conveyor was both random and hugely variable. So was the weight and density, which was an even bigger challenge.
            We made it work without too much fuss, one pallet at a time. Packing many smaller, less variable items into a larger volume would be even easier. Especially if you have a big pile to choose from before you start building.

          2. The robot would have some foresight to what’s coming from upstream from various packing stations. It just need to make do with what’s available and doesn’t need to be a 100% efficiency in packing. This is a classic computer science knapsack problem.

      2. I had the exact same thoughts. However I imagine the boxbot could easily have pre-set templates so that it’s not literally making the smallest possible box, but the smallest possible box of an allowed set of box sizes.

        I would hope they would also tag certain items as needing more careful packaging so that they either bypass this process, or the box gets sized up accordingly to fit more padding and it just doesn’t get closed by the boxbot? Either way, it’s solvable and I *hope* that if two random people reading the article came up with the idea, the engineers devoting brainpower to solving it as a paid job are thinking of them :)

        1. Maybe the author didn´t understood the packing part and wrongly wrote to us that they wouldn´t be using it.

          Here in my job, we always were of the opinion that the cost of packing material ( bubble wrap, good boxes, brown paper, enough tape ) was far less than the cost of dealing with someone complaining their thing arrived damaged due to bad packaging by us.

          It seems kinda like good practices in programming : you do not assume your code will always receive correctly formatted parameters, same way you do not assume everybody will treat your package with the same care you think it deserves. One defends against the worst, or deals with system hangs/irate customers.

          1. That doesn’t work when you have a low cost product in the millions. Packaging can eat you alive and be a huge percentage of the overall cost. You can easily overdo it as much as under do it. In the lower cost realm, you may want to eat a very small percentage of damaged goods to save on the Ferrari packaging for everything.

  3. Who needs padding when you can have a form fitting carton that transfers the impact 100% to the goods? Amazon free in 2019 for me. It feels good not to support the race to the bottom and a guy who treats his wife and employees like trash.

  4. Fragile items don’t seem like they’d be a tremendous issue. Things like glassware or computer monitors aren’t normally rolling around loose in the warehouse. They’re already coming from the factory boxed up in specialized materials.

    If anything, this approach is probably superior to the existing one, which in my experience tends to amount to placing a fragile item like, oh, say, a replacement french press beaker, in a box several times too large, then throwing in a few desultory pieces of air pillow to rattle around alongside it. Most of the time, it would be safer to just slap a shipping label on the manufacturer’s package. At least that could be handled without internal movement.

    Speaking of, it’s interesting that Amazon doesn’t seem to be taking the approach of simply demanding some kind of standardized, label-ready brown cardboard packaging from their suppliers. I suppose either they haven’t been able to apply enough pressure to get that without added cost, or perhaps they *want* to pack it in their own material – for branding and inventory clarity, maybe.

    I wonder if or how they’re planning to incorporate multiple items for a single shipment. I could imagine a system that makes a custom sized box for small widgets to fit alongside the main item. Or maybe gluing on an extra box or padded mailer. Either way, there’s extra packing and coordination involved.

    1. Amazon has a certification called “Frustration-Free Packaging” where items are already boxed in a ready-to-ship format from the manufacturer. They recently mentioned that companies who did not comply with this policy for certain goods would eventually be charged an extra Prep fee. Apparently the standard has been around for a while, but the push for adoption is fairly recent.

      1. Are they extending that to larger items?

        Obviously the dynamic is similiar (getting suppliers to change their packaging), but I’d always assumed the “frustration-free” label implied “easy to open” rather than “ready to ship”. I’m not sure I’ve ever seen it attached to anything but small items like USB cables, which then come in a simple plastic baggie or something (rather than a theft-and-tactical-warhead-resistant heat-sealed plastic clamshell). I’m all for it, but shipping-wise, they’ve always still arrived in a larger box or padded mailer.

  5. Would love to see that machine handle multiple items of irregular shape. Seems to be replacing one wasteful aspect with another – individually packaging each and every item.

    As for the guy in the second video claiming that jobs replaced by robots are simply created in the factory creating the robots… can he not see past the nose on his face? What happens when the factory that makes his robots gets automated by robots? Yea, there might be jobs for the installation of the assembly line, but after that, it’s just a few to monitor the works. Where does it end?

    With all the hype of AI/Machine learning, I would LOVE to see someone come up with an AI CEO. What does the CEO of a company really do beside analyze industries, spot trends and choose paths. How interesting would it be to see if an AI could be trained to be an effective CEO. Also how rewarding would it be to see corporate leadership positions lost to the march of technology. Without human greed, maybe AIs wouldn’t overlook the need for a financially healthy population in a society dominated by consumerism and would actually choose NOT to automate away every possible job. Wouldn’t that be a kicker!

    1. “Where does it end?”

      Well, ideally, with no one having jobs.

      In theory, if I were an Amazon box packer, and I could, personally, invent a simulacrum robot that went to work did my job everyday with no one the wiser, everyone would be better off. I’ve automated myself out of a job, and yet the only upshot is I can stay home and play video games, or work on my self-published anthology of bad poetry or whatever. And Amazon still gets their boxes packed.

      But that’s because in that case I would have managed to automate just the *work*, not the income from the productivity being performed.

      The real problem isn’t so much that automation is replacing “jobs”, it’s that we don’t have any mechanism for directing the benefits of automation to the workers who have been, theoretically, liberated from their labor. We’re still stuck in a mindset where “job” doesn’t just mean an activity that needs performing, it also means income/personal value/place in society/etc. Or maybe more precisely, that not having a job implies not being entitled to income, or a place in society.

      That’s an increasingly incoherent mindset in a world where machines can do more and more of the day to day stuff for us. Dunno if it’s going to turn out well or not, but something’s gonna have to give one way or another.

      1. It is bad enough that the politicians are so far removed from the average Joes that they do not comprehend the day to day struggle for the people he/she is supposed to serve.

        May be the AI decide that human is bad for the Earth or that they are only good for fuel? Good luck in the Matrix or Terminator world.

    2. The CEO is the fall guy who takes the blame when everything goes horribly wrong.

      Being the boss is a tough spot. If you do your job really well everyone is self motivated/directed and doesn’t need you.

      1. Well, seen from france, either the CEO is doing a good or a bad job, he just take the money and leave with a golden parachute.
        Sometimes, he stays in place until the compagny goes bankrupt (IE everybody is the fall guy).

  6. I bought a pair of binoculars from Amazon. The manufacturer’s box was not designed for shipping, with no padding. It came in a larger box with no filler so that it was free to bounce around. Luckily, it wasn’t damaged. Maybe a robot could be programmed to be smart enough to dump in a little padding. The Amazon employee wasn’t.

  7. The mathematical model was 52% to 48% for. I am not sure how the math can be any clearer, but as Erdogan has shown do-overs are now in vogue if the elites don’t get their way. Or in America, if the voters won’t vote for you, replace the voters (through illegal immigration over multiple generations).

  8. i work for a business owner who hates labor. he will do anything to prevent labor costs because he doesn’t have anything he can reclaim the costs. assets are another thing, you can put your hands on them, he also doesn’t have any friends

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.