Google Maps, New England style

[cos]

I just got the Google Maps version of "You can't get theah from heah!"

I'm planning a road trip for this summer, and using Google Maps to help me figure out distances on unfamiliar variants of cross country routes, and what stops make sense between one place and another. I just tried to get a route from Tempe, Arizona (new home of listgirl) to Hillsdale, NY (approximate site of Falcon Ridge Folk Festival).

This was Google's answer:


"We could not calculate driving directions between Tempe, AZ and Hillsdale, NY."

Comments

[tisiphone.livejournal.com]

That's what they get for using Navteq data.

possibly slightly out of the way, but:

[scromp.livejournal.com]

Ever been down southern Utah's highway 12? Particularly the bits between Capitol Reef and Escalante are fantastic. Highly recommended if you like geology. It starts up around ~5000 feet on the edge of the Colorado Plateau and leads you gradually down through billions of years of rock (all the way to the oldest visible rock on the planet if you go to the bottom of the Grand Canyon.) You do need to get out of the car at certain points. Let me know if you want more details. :)

It drops you out within spitting distance of Page, AZ as I recall.

[cos]

This is the first time I've ever seen this particular error, and I suspect it's not the date, it's the route-finding algorithm taking too long. When I tried the search again a few minutes later, it had a route for me.

But since you know loads about this, do tell me, what is your complaint about the Navteq data? What do you think they should've used?

Re: possibly slightly out of the way, but:

[cos]

I've been through Utah a bunch of times. The first was part of a long summer camping trip with my family when I was 10, when we visited most of the national parks (Arches hadn't been "discovered" by most tourists at that point, and had very few people, which I hear is not the case anymore). My favorite interstate drive anywhere in the country is I-70 from the beginning in Utah, to Denver. I don't think I'll have any chance to linger in Utah on this trip. DemocracyFest is in San Diego on July 14-16, and I intend to be at Falcon Ridge by the time it starts on the 20th. And on the way out west, after Colorado I'm probably going to Seattle, then downcoast from there.

[tisiphone.livejournal.com]

Heh, I was mostly just being snarky. The biggest problem is that Google Maps uses a combination of Navteq and Tele Atlas data, but AFAIK they don't sew the data together; which data you get, if they've got dual coverage, depends on which server you hit; they're both about the same quality overall, but I think TANA ranks a bit higher on the interstate connectivity scale - NTEQ is better for in-city travel, frankly.
(The problem would indeed be in the routing algorithm, probably it was disjointed at a state border, and had to find a way around.)

... but bad directions

[cos]

This is another one of the many, many cases where Google and MapQuest give different directions, and MapQuest's make a lot more sense. The pattern I think I've been seeing may have something to do with how they weight the relative value of using better roads vs. travelling a longer distance, but there may be multiple factors at play.

[struct.livejournal.com]

GM definitely won't give you the "as the crow flies" route every time, and a slight change in your starting point can generate completely different routes, as I learned the hard way when plotting a trip from Austin to Taos, NM.

I'm guessing that's because the route-plotting algorithm is chaotic (in the mathematical sense that is, i.e., an iterative function is used such that a slight change in initial conditions will completely alter the final outcome) and also I think adjusts the output route based on probable travel time, which in turn is based on road type and average traffic conditions.

Now if GM really wanted to get fancy, they could let you specify prefs like type of driver (conservative, average, aggressive), type of vehicle (fuel-efficient, avereage, gas-guzzler), etc., add to that factors like predicted weather and traffic, and adjust route-plotting accordingly.

At any rate, even though I do think Google Maps is the coolest (true dat, double true!) I prefer Rand McNally's TripMaker for anything over a day's journey.

Re: possibly slightly out of the way, but:

[whitebird.livejournal.com]

Highway 1 in part of Marin county is currently closed due to rain damage, no idea when it's going to be open, the same with the Devil's Slide area a bit south of San Francisco, that's definitely closed for multiple months.

Re: possibly slightly out of the way, but:

[cos]

I'm mostly going to take the 5, or possibly 101 for portions, to maximize amount of time spent in places where I have friends (Seattle, Portland, SF/Berkeley/Silly Valley/Scruz, Santa Barbara, LA) and minimize the time spent getting between them.

[cos]

All of the online directions thingies wieght by road class (they don't, as far as I know, know anything about travel conditions and speed limits specifically, they just classify roads on a scale of several discrete classes). And they're all (or at least the ones I've tried) more or less deterministic. They change over time, probably due to new data or tweaks to the variables, but I've never done the same search with the same site and gotten different results within, say, the same week. And of course I don't want "as the crow flies" routes, these are driving directions. Road classes do matter. I wasn't expecting Google Maps to find the shortest route. I do expect MapQuest to find the most sensible route, because it usually does. Someday, maybe I'll have that expectation of Google Maps, too.

P.S. Many of the most egregious examples of MapQuest finding the sensible route and Google Maps not, are local trips.

Re: possibly slightly out of the way, but:

[cos]

Also - I understand that portions of highway 1 fall apart and get closed every winter, but California makes great effort to fix all of it by summer, because the tourist revenue is greater than the cost. This road trip is going to be in July, so if I do take highway 1 I think I'll just expect it to be there for me by then :)

Re: possibly slightly out of the way, but:

[yix.livejournal.com]

Seattle?

Seattle

[cos]

Yes! *bounce*

[struct.livejournal.com]

I figured online route planners wouldn't use fuzzy logic, and it'd be bad (but amusing) if they did.

My experience with the big three is that they weight road classes too heavily. For instance, getting back to my Austin-to-Taos example, GM, MQ and YM all plotted an Austin-Wichita Falls-Amarillo-Taos route for me, probably because 35 is an Interstate. The straight shot down Highway 84 through Abilene, Lubbock and Clovis is the better way to go however, as I found out through personal experience.

The other problem with road classes (and this might explain in part the local route problem) is that a given road class can be a help or a hindrance depending on the time of day. For example (and here again I speak from personal experience) the Kennedy and the Dan Ryan are the best way to navigate Chicago northwest-southeast, except between the hours of 7:30-9:30a and 3:30-6:30p, in which case you need to avoid them like the plague because they turn into parking lots.

[superfinemind.livejournal.com]

Is this a googlemapwhack?

googlemapwhack

[cos]

Oooh, it is! And just like with normal googlewhacks, once you've found it, it's done, and nobody else can use it. You don't even have to post it. Google seems to figure out a route after it has failed someone. Someone else I was talking to yesterday found another "could not calculate driving directions" and a few minutes later it worked too.

Re: googlemapwhack

[superfinemind.livejournal.com]

Well, you should post it, a screenshot like you've got, so you can report it to all of your friends, since there's no indicator for "I couldn't calculate this, once upon a time."

^_^

[xthread.livejournal.com]

That is excellent!

[xthread.livejournal.com]

What's a navteq?

[tisiphone.livejournal.com]

Navteq and Tele Atlas NA are the two main players in the auto navigation GIS field (as well as most other GIS applications, actually) in North America. I can't really get you specific information on them (I'm at work, and that might raise eyebrows) but their data gathering methods are markedly different from TANA's; sometimes, in a good way, sometimes in a bad way.

Re: googlemapwhack

[warren8472.livejournal.com]

Sort of like what happened with the NYC Subway hack of Google Maps, where directions from Astoria (in Queens) to Roosevelt University (in Manhattan) suggested that you take the N train to Queenborough Plaza, walk east until you run out of dry land, swim across the Hudson River, dodge traffic on the FDR Expressway, and finally tresspass through a hospital and climb the fence to get to the corner of York Avenue and East 68th Street. Brilliant. That got fixed in a jiffy.

[georgejas.livejournal.com]

i've gotten this when i don't put the colons and spaces in exactly the right spot.

from: city, st to: city, st

is the only thing that consistently works for me

Re: googlemapwhack

[ron_newman]

but Roosevelt University is in Chicago.

Re: googlemapwhack

[warren8472.livejournal.com]

Rockefeller. My bad.

Re: googlemapwhack

[warren8472.livejournal.com]

I was thinking of Roosevelt Island, which sits in the East River (not Hudson, duh), which I forgot to mention in the already crowded obstacle course I mentioned above (map of the area here (http://www.mapquest.com/maps/map.adp?formtype=address&country=US&popflag=0&address=york+ave+%26+east+65th+st.&city=new+york&state=ny)).