How Does Google's Predictive Traffic Maps Work?

Oct 27, 2010 • 8:07 am | comments (4) by twitter Google+ | Filed Under Other Google Topics
 

We know how Google Maps plots much of the real time traffic. In short, they watch people via their mobile phones and see how fast they are driving in real time. But how do they go about the predictive traffic?

If you click on the "Traffic" button and then select "change" from the real time and select the "traffic at time and day" you will be able to have Google predict traffic conditions on a certain day and time.

Google Traffic

So how does this work?

Dave from Google explained a bit on how that works in a Google Maps Help thread. He said:

Yes you are correct it is not an attempt to predict travel time under "current" traffic conditions. It is simply a measure of how quickly (slowly) traffic is moving along those roads at times when it tends to move its slowest. There is no specified time interval.

It's not like we say "rush hour tends to be at 5:00, so let's measure the speed of traffic at 5:00 every day for a month and say that's the time in traffic". That obviously wouldn't work, because rush hour times are a little different everywhere. Think of it more as if we said, "when traffic was its worst today, how slowly was it moving" and answered that question every day for a month. By coincidence, the time when that happened would tend to be rush hour, but that time could be different depending on the city, road, and direction of traffic. It's more sophisticated than that, because as I said we don't want to include extraordinary events such as accidents, but that gets the general idea. The point is, that time is intended to be the worst case driving time barring extraordinary events.

So what I understand, Google pulls out outlier data caused by accidents or other issues and does some basic statistics to formulate a good prediction based on the live traffic data they collect on a daily basis.

Forum discussion at Google Maps Help.

Update: Dave from Google commented:

Hi, this is the same Dave that answered the question in the help forum. I want to clarify that question concerned our "time in traffic" prediction associated with some driving directions results*. It didn't (at least as I understood it) have anything to do with the "traffic at day and time" layer that you reference in your post.

*For example, if you get directions from Seattle to Portland (http://maps.google.com/maps?f=q&saddr=seattle&daddr=portland) the first suggested route will give estimated driving time under typical conditions and estimated time "in traffic". The "in traffic" time is what the poster to our forums was asking about.

Thanks for the clarification.

Previous story: Optimizing Your Site For TV & Not Creating Duplicate Content
 

Comments:

Jason Crabtree

10/27/2010 01:16 pm

It seems like Google are in the best position to organize a highway patrol service, plus an ambulance and fire engine service for traffic accidents, and to provide general traffic news to motorists - the information they're now gathering is truly awesome.

Dave Barth

10/27/2010 10:31 pm

Hi, this is the same Dave that answered the question in the help forum. I want to clarify that question concerned our "time in traffic" prediction associated with some driving directions results*. It didn't (at least as I understood it) have anything to do with the "traffic at day and time" layer that you reference in your post. *For example, if you get directions from Seattle to Portland (http://maps.google.com/maps?f=q&saddr=seattle&daddr=portland) the first suggested route will give estimated driving time under typical conditions and estimated time "in traffic". The "in traffic" time is what the poster to our forums was asking about.

Barry Schwartz

10/27/2010 11:36 pm

Thanks Dave. Very good clarification. Appreciate that.

jimbeetle

10/28/2010 01:52 pm

Hey Barry, This post got me to thinking how Google might differentiate between a vehicle and pedestrian in congested areas. Dave was good enough to reply that the "answer is not simple" and he couldn't "go into detail." Wonder what would happen if x number of joggers were crowdsourced going twice the pace of vehicles on, say Manhattan's 49th St? Sounds like it could be fun.

blog comments powered by Disqus