The Other Worlds Shrine

Your place for discussion about RPGs, gaming, music, movies, anime, computers, sports, and any other stuff we care to talk about... 

  • Interesting! Now let's see the others implement the same

  • Somehow, we still tolerate each other. Eventually this will be the only forum left.
Somehow, we still tolerate each other. Eventually this will be the only forum left.
 #114485  by Shellie
 Wed Dec 12, 2007 2:32 pm
It seems that sitting in the left lane, engine idling, waiting for oncoming traffic to clear so you can make a left-hand turn, is minutely wasteful — of time and peace of mind, for sure, but also of gas and therefore money. Not a ton of gas and money if we’re talking about just you and your Windstar, say, but immensely wasteful if we’re talking about more than 95,000 big square brown trucks delivering packages every day. And this realization — that when you operate a gigantic fleet of vehicles, tiny improvements in the efficiency of each one will translate to huge savings overall — is what led U.P.S. to limit further the number of left-hand turns its drivers make.

The company employs what it calls a “package flow” software program, which among other hyperefficient practices involving the packing and sorting of its cargo, maps out routes for every one of its drivers, drastically reducing the number of left-hand turns they make (taking into consideration, of course, those instances where not to make the left-hand turn would result in a ridiculously circuitous route).

Last year, according to Heather Robinson, a U.P.S. spokeswoman, the software helped the company shave 28.5 million miles off its delivery routes, which has resulted in savings of roughly three million gallons of gas and has reduced CO2 emissions by 31,000 metric tons. So what can Brown do for you? We can’t speak to how good or bad they are in the parcel-delivery world, but they won’t be clogging up the left-hand lane while they do their business.

http://www.nytimes.com/2007/12/09/magaz ... dturn.html

 #114486  by Lox
 Wed Dec 12, 2007 2:47 pm
Heck yeah! :)

The application I was mainly responsible for testing before making my move to development in October was how this info got onto the driver's handheld (his DIAD). So I am pretty familiar with the apps that generate the info and all that. It's all pretty interesting stuff. Within IS, we refer to the suite of apps as PFT (Package Flow Technology).

If you're interested in any details, let me know.

 #114487  by Flip
 Wed Dec 12, 2007 3:35 pm
Lox wrote:
If you're interested in any details, let me know.
Results = interesting. Details = not so much.

^_^

 #114488  by Lox
 Wed Dec 12, 2007 3:37 pm
Not you, sugar pants!

I meant our resident tech geeks (Sine, Seraphina, Tess, Kup, Andrew (I think!), etc). And that includes our resident tranvestite tech geeks too such as Ish. :)

 #114490  by Tessian
 Wed Dec 12, 2007 7:18 pm
I heard about it today too... pretty freakin cool stuff. I'm not sure if I want to have my mind explode trying to comprehend HOW it works... but I do at least assume it doesn't just make 3 rights for every left? I hope it will only avoid lefts if it wouldn't take longer to just sit and take the left.

The whole concept is awesome though... I'd be interested to see it's effects (mileage reduced, gas consumption, time spent in transit, etc)

 #114493  by Lox
 Wed Dec 12, 2007 8:26 pm
The funny thing is that we've been using the technology for years now though it's still being rolled out throughout the US. I guess someone outside just found out about it. :)

I don't think I could cause your mind to explode. While I understand it, I don't know everything about it since the main apps responsible aren't in my realm of expertise (my division deals with the DIAD and all of the stuff involving it). Though they are handled within my building.

 #114495  by Kupek
 Wed Dec 12, 2007 9:34 pm
I'd actually be interested in knowing more about the algorithm they used - the general case is <a href="http://en.wikipedia.org/wiki/Traveling_ ... oblem">the traveling salesman problem</a> exactly, which is <a href="http://en.wikipedia.org/wiki/NP-complet ... omplete</a>. If N is kept at a reasonable size (say, N < 20) it's not a problem, but for most interesting applications, that's not the case, and you'll have to resort to heuristic techniques. Particularly if you need to solve it in near real-time.

So. How'd they do it?

 #114498  by Lox
 Wed Dec 12, 2007 11:25 pm
Kupek wrote:I'd actually be interested in knowing more about the algorithm they used - the general case is <a href="http://en.wikipedia.org/wiki/Traveling_ ... oblem">the traveling salesman problem</a> exactly, which is <a href="http://en.wikipedia.org/wiki/NP-complet ... omplete</a>. If N is kept at a reasonable size (say, N < 20) it's not a problem, but for most interesting applications, that's not the case, and you'll have to resort to heuristic techniques. Particularly if you need to solve it in near real-time.

So. How'd they do it?
I can't speak to the algorithm used by the app as that is something that was probably developed by one of our teams that represents the field (drivers, package center employees, etc). Since I've never worked directly on these projects, I've never been privy to that info. Your question does make me curious though. I'll see if it's possible to dig up something a little more specific from someone who works on the team.

I do know that the data used to generate these routes is historical in nature so it's not real-time in the sense that it is being calculated as the driver is on road. As data is collected based on his previous stops, it is used to predict his path in the future and then that is factored in along with information about what is going to be delivered and picked up in the future (since we basically know where all of our packages are at any given time).

 #114511  by Ishamael
 Thu Dec 13, 2007 2:42 am
Lox wrote:Not you, sugar pants!

I meant our resident tech geeks (Sine, Seraphina, Tess, Kup, Andrew (I think!), etc). And that includes our resident tranvestite tech geeks too such as Ish. :)
Sorry, I was in the middle of polishing my heels!

Interesting problem. Data-wise, they have to be using NavTeq or TeleAtlas.

The algorithm to solve the left turn problem is up for grabs though. Seems like there'd be a lot of computation trying to calculate alternate routes and then figuring out which one of those alternates results in the fewest left turns.

 #114512  by Flip
 Thu Dec 13, 2007 3:01 am
The 28.5 million miles is the only reason why the article seems suspect... how can routing delivery that takes more right hand turns result in less miles? It seems like the program wants a result with more rights, not less miles...

 #114513  by Lox
 Thu Dec 13, 2007 7:37 am
Ishamael wrote:Interesting problem. Data-wise, they have to be using NavTeq or TeleAtlas.
Actually, the mapping technology is all proprietary. We bought another company years ago that does all that stuff.
sugar pants :) wrote:The 28.5 million miles is the only reason why the article seems suspect... how can routing delivery that takes more right hand turns result in less miles? It seems like the program wants a result with more rights, not less miles...
The focus is less redundancy. Don't criss-cross a driver's routes. Don't keep doubling back, etc. Less miles is achievable because prior to this the driver's route wasn't necessarily the most efficient in terms of mileage either. So mapping an order of stops for him while also reducing the number of left turns leads to less miles.

 #114519  by Flip
 Thu Dec 13, 2007 9:46 am
Lox wrote: The focus is less redundancy. Don't criss-cross a driver's routes. Don't keep doubling back, etc. Less miles is achievable because prior to this the driver's route wasn't necessarily the most efficient in terms of mileage either. So mapping an order of stops for him while also reducing the number of left turns leads to less miles.
Its hard for me to believe that such a large shipping company who deals with ground delivery more often than not didnt have any efficiency programs prior to this right hand turn heavy one... but what do i know. Its cool that they finally did something, but also stupid that it took them that long do do anything!

I'm slightly biased, though, i admit, because i'm not a huge UPS fan. It seems whenever i have something delivered by UPS, it comes from a sketchy guy in a Budget rental truck who has a UPS magnet on the side on the door. It bothers me that they contract out so much and dont provide uniforms or anything official besides a magnet.

 #114523  by Lox
 Thu Dec 13, 2007 10:30 am
Flip wrote:Its hard for me to believe that such a large shipping company who deals with ground delivery more often than not didnt have any efficiency programs prior to this right hand turn heavy one... but what do i know. Its cool that they finally did something, but also stupid that it took them that long do do anything!
They kind of did, but it was more based around training the driver to understand routes and such. The technology that this article talks about has been around for several years at least (before I ever joined the company). Prior to that, the driver would plan his day each morning by analyzing his car. But this technology also deals with minimizing the training required to be a package car loader because the turnover is so high in that job. There is a specific methodology to loading a car so that the driver can get all of the packages for a stop the 1st time he arrives there. Before computer software was able to handle these things, you had to trust that every driver, loader, etc did his job exactly the way he was trained. But that's not a realistic expectation.
Flip wrote:I'm slightly biased, though, i admit, because i'm not a huge UPS fan. It seems whenever i have something delivered by UPS, it comes from a sketchy guy in a Budget rental truck who has a UPS magnet on the side on the door. It bothers me that they contract out so much and dont provide uniforms or anything official besides a magnet.
What time of the year are you talking about? UPS does not contract out delivery drivers at all. They can't. The drivers all belong to the Teamsters union. The only possibility is that you're talking about deliveries in November or December (peak) or the center is badly managed and they can't handle their package load and those are helpers (who are employees of the company from other areas who wouldn't have uniforms) or center managers (who also wouldn't have a uniform).

In retrospect, FedEx, for example is almost all contractor-based.

 #114534  by Ishamael
 Thu Dec 13, 2007 1:44 pm
Flip wrote:The 28.5 million miles is the only reason why the article seems suspect... how can routing delivery that takes more right hand turns result in less miles? It seems like the program wants a result with more rights, not less miles...
Yeah, that's very suspect, but it depends on assumptions being made. I believe the spokesman was assuming that the drivers went from using NO mapping software, to using this new mapping software. So yeah, I can see how miles are saved in that scenario.

If however, you went from using mapping software where there was not left turn consideration, to one that went out of its way to avoid them, then it seems like you'd be driving MORE miles.

update:
Flip wrote:Its hard for me to believe that such a large shipping company who deals with ground delivery more often than not didnt have any efficiency programs prior to this right hand turn heavy one... but what do i know...
Missed this one. Yeah, it does seem odd. Lox says they had something before (which seems right), but if this new software is saving 28 million miles WHILE adding more right turns, the old stuff must have sucked pretty bad. :)

 #114537  by Lox
 Thu Dec 13, 2007 2:19 pm
Ishamael wrote:Missed this one. Yeah, it does seem odd. Lox says they had something before (which seems right), but if this new software is saving 28 million miles WHILE adding more right turns, the old stuff must have sucked pretty bad. :)
I said that there was planning but not using a software-based approach like we do now. The company has been examining routes and all that for years and years and years (heck the company is 100 years old). A lot of that leads to how we split our centers up and which center services which addresses, etc. But before Package Flow (which, again, has been around a lot longer than this article implies) there wasn't something to plan an order for the drivers. He literally plotted it himself using a report of today's stops and a map because most of the drivers knew their area very well. Once Package Flow started taking over, the routes were planned for the driver AND they were planned to minimize left turns and all that. So it wasn't just the "no left turns" that saved that much mileage.
article wrote:The company employs what it calls a “package flow” software program, which among other hyperefficient practices involving the packing and sorting of its cargo, maps out routes for every one of its drivers, drastically reducing the number of left-hand turns they make.
Note the parts I bolded. The article itself focuses on the "no left turns" but that's just a small part of what the technology does as a whole.

I just did a quick check. We've been using this stuff since 2001. :)