well, in the old world nothing is perpendicular, and all the roads have a different width and single direction without any clear pattern, so not getting lost is already a strong goal, there is very limited BW to optimize for time :)
This is anecdotal, but I've asked about twenty or so UPS drivers that I've met over the years whether they avoid turning left or not, and they all just laugh and say "No, that's dumb." Either they're messing with me, they just aren't following "the rules", this "UPS drivers don't turn left" thing is urban legend.
I'm kind of leaning toward the latter. Not so much that it never happened, but more that it was something that some data scientist said was a good idea, but never actually works I practice, and so UPS drivers don't actually follow this. Maybe it's a policy. I don't know. I don't have a list of UPS policies handy at the moment. But of the (admittedly small) sample size I've discussed it with, every single one says they do not follow that policy, or is it something they were instructed to do.
Which makes me kind of roll my eyes whenever I see these articles or podcasts or what-have-you. They pop up fairly frequently, but... it feels more like clever marketing by UPS than an actual policy.
(in the US) In very suburban streets in the middle of the day, nobody cares about turning left in an empty street. In more congested and wider streets it could get more delaying. But if the planning has been done by a software before the driver took the truck, the software might have avoided as many left turns as possible without telling the driver, and without having any need to ask the driver to micro-optimize themselves, they just follow the GPS (you probably can't do better), or the most sensical path ca't do better), or do a slight detour to get a snack (unforseeable).
Long story short it might be a software policy without being a driver policy, that would be enough to have a significant statistical result, I don't think drivers re-plan their shift.
I'd reckon that anyone on a bid route or other route that they are familiar with can do better than what ORION gives us. We aren't taking solely distance into account, that would be far too simple. We're considering other factors too, including hard time constraints like next day air commit times, regularly scheduled pickups, irregular pickups, business hours, and whether we can even physically access the packages [0, 1] that the system wants us to deliver.
>I don't think drivers re-plan their shift.
We most certainly do, we wouldn't finish the day if we blindly followed trace.
My understanding from someone who contributed to the development of ORION is that drivers are encouraged to try to “beat” ORION’s route guidance and those improvements are incorporated into the route planner. Any truth to that?
Simply put, there's no way this is the case. There's a route that I and everyone else I've spoken to all run a certain way and improvements have not been incorporated into the route planning. On this particular route if you run the apartments when ORION wants you to then you'll arrive there after Amazon and FedEx have filled the parcel lockers with their delivery volume, and you'll spend an hour delivering your volume to each individual apartment door. On this route you need to deliver certain dense and volume-heavy areas before your pickups so that you'll have enough room to fit the pickups into the vehicle. I would love to see ORION automatically learn about when streets and such are added, removed, or otherwise changed, but that's a manual process. I can ask the dispatcher about what you're talking about, but as a driver I have seen no such behavior.
Not turning left in leafy suburbs would mean driving a few miles to make a riskier U-turn just to make a right. So it definitely would be counter productive if it were a strict rule.
We were talking about routing choosing to avoid a left turn. What you are describing isn't a routing software choice, that's you missing a turn. Not equivalent
I'm not a UPS driver, but roughly follow this rule when driving and pick routes based on it. I've done so since before I heard about the whole UPS thing, as it makes sense to me, so when I heard that UPS did it too I was like, oh so it's not just me. But alas... maybe it truly is just me.
> This is anecdotal, but I've asked about twenty or so UPS drivers that I've met over the years whether they avoid turning left or not, and they all just laugh and say "No, that's dumb." Either they're messing with me, they just aren't following "the rules", this "UPS drivers don't turn left" thing is urban legend.
This article is more about the way routes are planned than the individual choices drivers make along those routes. There isn't a lot of room for agency in route choosing, drivers are given a stack of packages and deliver them in the way the route is built.
They must have designed the road system in New Jersey - almost no left turns are allowed. Instead, dog legs, roundabouts, and, when there is too much temptation, jersey walls where most states have yellow lines.
This article is a little off base. UPS is solving for "What is the best way to safely deliver packages to 300 addresses". There are a lot of ways to do cover that ground effectively without too much waste. The routes UPS drivers share little resemblance to the way normal people drive.
I'm solving for "How do I get to work" or "How to get to the grocery store". I'm taking the highway 80% of the distance, then my work is on the left... do I take 3 right turns to get there circling a potentially large city block? The UPS driver might well have stops on those side streets, I don't.
The idea behind this is sound, but it doesn't necessarily apply to the way individuals who aren't driving package routes should drive.
Right, you don’t want a delivery driver parking and crossing the street. You want them next to the address, walking less and carrying weight for less time. This is absolutely not applicable to anyone driving in their normal life.
“Never turn left” would be a ridiculous rule on its own, as it ignores the frequency of oncoming traffic, or the existence of left-turn bays (with left-turn signals) or 4-way stop signs.
The "prisoner's dilemma" stuff at the end of the article makes no sense. The article starts by explaining why UPS does this - it benefits UPS. UPS doesn't do this to be a "good citizen", it does it because it is more efficient for UPS. UPS doesn't even do it because it's somehow better for UPS overall even though it's worse for individual UPS drivers. If an individual driver could do better by turning left according to their route planning software then it would have them turn left.
If this works in general for typical routes driven by individuals (rather than only for delivery routes like a UPS driver is making with multiple stops that don't have to happen in any particular order) then it would make sense for individuals to do it and they could benefit by doing it without needing anyone else to do it as well. There is no prisoner's dilemma here.
I always wondered if there’s a dedicated team at UPS of just mechanical engineers (for some reason most process optimization professionals have a MechE degree) and mathematicians just figuring out these sort of problems.
I'm sure there are. There's a team at my company (retail, not UPS) whose sole responsibility is figuring out how to optimize getting from Point A to Point B. At least 6 people on the team, probably more involved that I don't realize. At the scale my company and UPS operate in, even minor efficiency gains can add up to gazillions of dollars in savings.
I talked to an Uber employee once about a "Zoolander mode" project for drivers so that it only turns right and cannot turn left. The code was pretty simple if you've ever written a maze-solving algorithm.
(Hint: What happens if you follow either side of wall for the entire maze?)
In theory, this would be useful for truck drivers or prioritizing passenger pickup locations.
But now, I use it as an example of overengineering and overoptimization.
In practice, the main concern should be reducing cognitive load on the driver. Additional instructions, rules that can be potentially broken, inconsistencies in logic, all of it poses a substantial risk.
33 comments
[ 2.5 ms ] story [ 143 ms ] threadI'm kind of leaning toward the latter. Not so much that it never happened, but more that it was something that some data scientist said was a good idea, but never actually works I practice, and so UPS drivers don't actually follow this. Maybe it's a policy. I don't know. I don't have a list of UPS policies handy at the moment. But of the (admittedly small) sample size I've discussed it with, every single one says they do not follow that policy, or is it something they were instructed to do.
Which makes me kind of roll my eyes whenever I see these articles or podcasts or what-have-you. They pop up fairly frequently, but... it feels more like clever marketing by UPS than an actual policy.
https://www.ups.com/us/en/services/knowledge-center/article....
(This link claims they had 55,000 drivers using the system in 2016.)
(in the US) In very suburban streets in the middle of the day, nobody cares about turning left in an empty street. In more congested and wider streets it could get more delaying. But if the planning has been done by a software before the driver took the truck, the software might have avoided as many left turns as possible without telling the driver, and without having any need to ask the driver to micro-optimize themselves, they just follow the GPS (you probably can't do better), or the most sensical path ca't do better), or do a slight detour to get a snack (unforseeable).
Long story short it might be a software policy without being a driver policy, that would be enough to have a significant statistical result, I don't think drivers re-plan their shift.
The UPS route-planning system is called ORION.
>they just follow the GPS
I'd reckon that anyone on a bid route or other route that they are familiar with can do better than what ORION gives us. We aren't taking solely distance into account, that would be far too simple. We're considering other factors too, including hard time constraints like next day air commit times, regularly scheduled pickups, irregular pickups, business hours, and whether we can even physically access the packages [0, 1] that the system wants us to deliver.
>I don't think drivers re-plan their shift.
We most certainly do, we wouldn't finish the day if we blindly followed trace.
[0] https://youtu.be/fMeBkPd8ZJk
[1] https://youtu.be/Oi4DW04_D5o
Umm, I’ll do a u turn.
Used to happen all the time.
This leads to it picking absurd routes.
So it’s a tangent on the topic of routing and left turns.
In any case, a left turn in this situation is preferable in suburban and semi rural areas.
This article is more about the way routes are planned than the individual choices drivers make along those routes. There isn't a lot of room for agency in route choosing, drivers are given a stack of packages and deliver them in the way the route is built.
I'm solving for "How do I get to work" or "How to get to the grocery store". I'm taking the highway 80% of the distance, then my work is on the left... do I take 3 right turns to get there circling a potentially large city block? The UPS driver might well have stops on those side streets, I don't.
The idea behind this is sound, but it doesn't necessarily apply to the way individuals who aren't driving package routes should drive.
If this works in general for typical routes driven by individuals (rather than only for delivery routes like a UPS driver is making with multiple stops that don't have to happen in any particular order) then it would make sense for individuals to do it and they could benefit by doing it without needing anyone else to do it as well. There is no prisoner's dilemma here.
In theory, this would be useful for truck drivers or prioritizing passenger pickup locations.
But now, I use it as an example of overengineering and overoptimization.
In practice, the main concern should be reducing cognitive load on the driver. Additional instructions, rules that can be potentially broken, inconsistencies in logic, all of it poses a substantial risk.