Google Maps

[asterroc]

This xkcd isn't too far from my experience with Google Maps during my recent road trip. It saw into the future (I-840 in NC didn't yet exist, and "Future I-840" ended at Exit 19 where Google Maps told me to get off at Exit 21), and it didn't see into construction (I-95 didn't have an exit 351C in Florida). I guess this is why they recommend a sanity check at the bottom in fine print on all their directions.

Comments

[jrtom.livejournal.com]

Better in what sense?

(Yes, I work at Google (although not on Maps), but I'm not being snarky, it's an honest question; I personally much prefer Google Maps, and have done so since long before I worked here. :) )

[kadath.livejournal.com]

MapQuest never sends me down "streets" that are basically unpaved cow trails, which Google has done repeatedly. There's one in our neighborhood that we have to warn people about when they're visiting for the first time, because they always use the Google Maps directions.

Google's pathfinding algorithm is also apparently allergic to not moving on a vector that points as closely toward the destination as possible, even when backtracking briefly would be the best way to get where you're going (which happens a lot in New England.)

Google Maps has some really cool functionality, but I'll always go to MapQuest for driving directions.

[jrtom.livejournal.com]

Fair enough. I haven't had either experience, but I don't do a lot of driving in areas where these things might come up, either.

When you say "would be the best way", what do you mean? Not using unpaved or otherwise smaller/slower roads?

Do you mind if I pass these comments along to the Maps team?

[kadath.livejournal.com]

Do you mind if I pass these comments along to the Maps team?

No, go ahead.

When you say "would be the best way", what do you mean? Not using unpaved or otherwise smaller/slower roads?

Things like taking highway exits that are "before" your destination, even though passing it on the highway and going "backwards" on surface streets would be faster and/or less involved.

It's also really bad at route-finding on the city level. Like picking weird side streets and throwing in inexplicable U-turns.

Here's an example of both, from the MIT area to Worcester Polytechnic Institute:

Google Maps is weird.

MapQuest goes the way I would.

[jrtom.livejournal.com]

Thanks, I'll pass this along.

Looking at that particular route, I think that I understand the decision to go via Salisbury rather than Highland: the time estimate is the same but the Salisbury route (even with the inexplicable U-turn) is a tenth of a mile shorter. There may be things about each route (in terms of ease of navigability, e.g.) that aren't apparent to people that haven't driven there, though.

I don't understand the U-turn at all, as I don't see any reason (and if you drag the route just right, neither does Google Maps) that you can't turn left from Salisbury onto Lancaster.

Anyway, I'll pass this feedback along. Thanks!

[kadath.livejournal.com]

There may be things about each route (in terms of ease of navigability, e.g.) that aren't apparent to people that haven't driven there, though.

That's the part that MapQuest usually seems to win on, really. And I have no idea how you quantify that algorithmically, but MapQuest has figured out something like it, apparently.

Speaking from local knowledge, in the routes I used as an example, MapQuest only uses two streets, one of which is an easy-to-follow numbered route and the other is a major cross street, and the all the turns they want you to take are assisted by lights. Google does...um, I'm not sure what. Both routes suck during rush hour, but that's unavoidable. (Neither uses the super ninja back way that avoids left turns and the usual congestion spots, but I don't expect that out of general-use mapping software.)