It is interesting/surprising to me that there isn't the suggestion of auto-loading/unloading the trucks themselves. Is this one of those problems that seems pretty tractable from the outside looking in but it turns out there are some fundamental difficulties that aren't obvious at first but make it much more difficult than first imagined?
I could imagine the beds of the trailers being conveyer belts themselves that would "plug in" to the system at the sorting facility. From there the automation could take over. It would be something like training the robots to play a combination of Jenga and Tetris.
The loading of the trucks requires much more dexterity than a machine would be able to deliver. Also the packages themselves would be difficult for a machine as you get these huge pieces that come down the belt weighing 150 pounds and the weight isn't evenly distributed at all. In these cases we have to stop the belt and two loaders load the item quickly, and the belt is restarted. What happens when a package comes down a slide and busts open and gets over everything? This actually happens quite a bit.. one time a goats head came out of the package and blood went everywhere. Or better yet, what happens when a package accidentally has two labels instructing the package to go to two different trucks, perhaps even two separate parts of the facility? We call this a double label and it usually requires a human to sort it out.
There is a lot of idiosyncrasies like this with many of UPS' jobs. Automation will come not to take the paycheck from the low wage worker, but rather from the high wage worker such as driver dispatching, human resources, etc. Those people won't be fired though, most will be reassigned to over see some drivers I would imagine. That is a tough job and those people need all the help they can get.
Spent about a year loading and unloading at the Trabue UPS hub in Columbus back in the 90's. Everything you say is true. Unfucking the belt is something that is likely to be strictly a human endeavor for the next 20+ years.
There's no way automate it with the current hub architecture and today's technology. You'd literally have to have a stadium-sized package 'cache' where all package could be unloaded to, some machine vision doohickey to characterize every package, then have a solver identify the idealized load configuration, then picker bots go out and start streaming packages in order so that some high end robot could load the trailer with 90+ percent packing efficiency (which i believe we achieved as loaders getting paid $8/hr).
Then you'd still have to leave the last 8' of trailer space to handle the goats heads and exhaust pipes and all the other crazy shit that people just slap a label on and make someone else's problem. Maybe you could find a way to use air bladders or some kind of encapsulating foam to handle that, but mostly you'd just have to have a floater running from trailer to trailer to load that stuff.
> UPS is negotiating with the International Brotherhood of Teamsters to renew a five-year contract, which expires July 31. Representing 260,000 UPS drivers, sorters and other workers, the union wants UPS to hire more full-time workers to help handle the surge in packages. It has opposed technology such as autonomous vehicles and drones and is wary of projects that do work with fewer employees.
> “The problem with technology is that it does ultimately streamline jobs,” says Sean O’Brien, a Teamsters leader in Boston. “It does eliminate jobs. And once they’re replaced, it’s pretty tough to get them back.”
It will be a sad day for a lot of people when UPS moves to automation.
Theyre one of the only companies left that still supports the 20th-century idioms of corporate loyalty-- with only a high school degree, you can start as a truck sorter and work your way up to a vaunted driver or stationmaster position over the course of your career.
Robots might improve a few things though, like safety. If they can not only account for spatial positioning but also weight, automatons might ensure trucks arent over- or unevenly-loaded.
Speaking awhile ago with an expert in this kind of automation, I was told that conveyor belts are much less reliable than I imagined.
Your idea also could be implemented using different tech: Think of the Amazon warehouse bots that lift entire shelving units and bring them to the people who pick items from them. Preloaded racks could be delivered into their places on the truck.
> Is this one of those problems ... ?
It also could be one of those problems that isn't a bottleneck and isn't worth automating.
Worked at UPS from ‘03 - ‘05. I can tell by the labels and codes on the packages I get that not much has changed since then. Didn’t know anything about programming at the time, but I still spent plenty of shifts daydreaming about how the job could be automated or at least done by trained monkeys. 15 years later I can look back and pretty confidently say that it’s a much more difficult problem than I thought at the time.
Loading semis is basically 3D Tetris where you have to consider not only shape (boxes, tubes, cylinders, and everything in between) and size (fits in the palm of your hand up to 2- 3m long), but weight (basically 0-75 kg) and rigidity. Humans pretty easily develop and intuition for these things. I certainly unloaded plenty of semis that were for all intents packed to capacity. In theory you could probably have some kind of fully integrated system that let you know ahead of time the details of every package so you could model the entire load and figure out how to fit it... but doing what humans do loading on the fly would be damned hard.
Loading package cars (the brown delivery trucks) drops the need to maximize volume carried and adds constraints about the order and location of packages so the driver can find them. Again though, it’s a fairly loose system. In theory the load is pretty well distributed across the truck’s numbered shelves by the planning system... except all the days that it isn’t. A loader ends up making a lot of on the fly calls to overflow shelves and rearrange things so they fit on the truck and the driver has a reasonable chance to find them.
tl;dr - Anything is possible, but automating that part of the process is an enormous challenge IMO.
Formally, this is the bin-packing problem, which is NP-hard. There are lots of heuristic and approximate methods for coming up with configurations, but hitting an optimal configuration is unlikely to be done in short order.
I imagine load distribution, box fullness, rigidity etc can be included as dimensions to optimise. It'd be necessary, really, but doesn't make things any easier.
This is, however, what I believe Amazon are trying to achieve. As you point out, they have the advantage of knowing in advance what the different items being packaged are. They can presumably optimise within a single box, over a single order and across multiple orders and shipments. At their scale it would make sense to devote a large amount of compute to squeezing a few percent extra out of each cargo flight and truck delivery.
The loading of a UPS semi might not look like this - at all... In addition to the criteria already mentioned by your parent-post, that need to be considered while loading, stability of the load is another important factor. The system in the video just stacks same-sized packages on top of each other, which is not the best approach. The reality is much more complicated as this. You might also be surprised, that the loading-speed shown is not that much faster than a human can do it.
Public Logistics Network suggested by Professor Michael Kay: http://www4.ncsu.edu/~kay/pln/
"A public logistics network is proposed as an alternative to private logistics networks for the ground transport of parcels. Using the analogy between the packages transported in the network and the packets transmitted through the Internet, a package in a public logistics network could, for example, be sent from a retail store and then routed through a sequence of public distribution centers (DCs) located throughout the metropolitan area and then delivered to a customer’s home in a matter of hours, making a car trip to the store to get the package unnecessary. The DCs in the network, functioning like the routers in the Internet, could also be located at major highway interchanges for longer distance transport."
One aspect of it is powered platforms that move boxes around without human intervention: "Modular storage design: In order to be cost-effective, the loading/unloading, sortation, and storage activities at each DC in the public logistics network must be highly automated since each load might visit a dozen or more DCs while it’s in transit, likely traveling on a different truck between each DC. Since existing automation technologies do not provide the flexibility needed to allow any size load to move to any location at anytime, a new DC design has been developed that would allow packages of varying size to be automatically unloaded at a DC, sorted, stored, and then loaded onto an outbound truck. Such a design would result in diseconomies of scale because it is cheaper to ship a single package compared to a larger consolidated load. Integral to the design is use of arrays of small square modules with orthogonal pop-up powered wheels."
That's a pretty good idea! Though my inclination was to think of a truck that dispenses (most) packages to the driver via a external hatch. Small win but possibly save a lot of bending, searching, and lifting.
>It is interesting/surprising to me that there isn't the suggestion of auto-loading/unloading the trucks themselves.
Surely UPS has already solved the logistics planning of putting the right packages with the right drivers, so what's left is prestaging the next truck load for a bot to pick and shift as soon as the truck returns. Although you could argue this means fewer jobs loading it could also potentially mean more jobs driving and frankly if a bot is doing the loading potentially less package damage from careless handling.
I worked loading and unloading at a FedEx ground sorting facility for a couple weeks 15 years ago; some of the trailers had rollers down the middle to speed loading and unloading. For the others we had rollers to put into the truck. But you wouldn't want the whole bed to be a conveyor, the sorting really needs one package at a time in the line, otherwise it's hard to scan and direct packages, and packages will get jammed at the turns.
Conveyor belts are designed carry packages, trucks carry pallets with 20-80 packages on them. Packages are loaded and unloaded onto transfer trucks in pallets. Conveyor belt will be useless in a truck bed. They could do one of those floors that cargo airplanes have where entire floor is tons of balls that enables you to slide anything across the area but then again special flat bottom pallets would be needed and those metal balls will make truck a lot heavier.
Non-UPS trucks the same size as UPS's carry much heavier loads. If the balls made sense in other respects, their weight would be much less of an issue for the truck than for an airplane.
I could imagine the beds of the trailers being conveyer belts themselves that would "plug in" to the system at the sorting facility. From there the automation could take over. It would be something like training the robots to play a combination of Jenga and Tetris.