I always thought predicting traffic for a particular day and time would be something easily predicted from historic data with regression. Google Maps now has this feature:
It would be good to actually include season, holiday and similar information: the predictions would be better. I wonder if one can find this data easily, or if others have done this work before.
It might also be nice to retain information on residuals for generating routes; I imagine travel time across the bridges in particular has a substantial positive skew, and you might want to optimize the 95th percentile travel time rather than the mean. That would probably be substantially less easy, though.
The problem is, if these traffic predictions are taken seriously, they would be self-defeating. If everyone believes road A will be congested and road B free, road B will end up congested. It's an interesting question what the long-term equilibrium will be? Is the traffic information more than cheap talk?
It's Wednesday, 12:40 NY time, and the current traffic map looks pretty similar to that. (There are a couple of roadworks that make it worse). If I zoom in one stop, there's a lot more red and black appears though.
(I'm not especially familiar with NY traffic, so I'm guessing that you are saying that this seems like there isn't much traffic there.)
The Washington State Department of Transportation (WSDOT) has a first rate planning and research department which has done a lot of work analyzing and visualizing traffic. See, for instance, their <a>2010 congestion report. They provide an online calculator for <a> 95% reliable travel times.
PS – Every holiday weekend, the Washington DOT also publishes daily travel time charts for major routes showing peak congestion hours, such as: http://www.wsdot.wa.gov/Congestion/MemorialDay/20…
It's not NYC, but you can get traffic data for Sydney from this recently completed competition: http://www.kaggle.com/c/RTA
There's an entire field devoted to predicting traffic and designing highway and road and public transportation systems to handle it's anticipated growth: civil engineering! There's an extensive literature. They consider the effects of closing lanes, adding new freeways, raising bus fares, etc. etc.
Kaggle ran a traffic prediction contest, so the data must be available for at least some locations.
I'm not sure why you think it would be a simple problem. If we assume that all the data we need are being measured and measured accurately, then yes. But what you find in the literature is that data quality and availability is a big problem. Then you have a lot of the same issues as econometric modeling, the need to adjust for all kinds of things like which day holidays fall on, weather, special events, etc. And furthermore, there is the issue of climate v. weather. It's "easy" to predict the average traffic but how do you predict weather, accidents, and other important sources of congestion?
@Gamer: I imagine any equilibrium with more information is better on the whole than one with less, right? (And that any equilibrium is better for individuals with more information than for those with less.)
Just an intuition.
Such service's been available for 5 years alreasy in Russia. (http://maps.yandex.ru/-/CBQGzI6b)
Now, only weekdays are used for predictions. I guess, other factors wouldn't have much influence on them as Moscow is always jammed. ;)
actually, this video from Google introduces it to cellphone as well. http://www.youtube.com/watch?v=FW3ubvjG0PU
@gamer and @davidc: indeed, this sort of traffic info has very limited practical use, first as @gamer pointed out, there is a variant of what economist Anthony Downs calls the "triple convergence" problem, drivers do change their behavior based on the level of congestion; second is that knowing the average traffic conditions is not good enough; it's the variability or extreme values that cause angst, and these are due to non-systematic causes. I talk about these issues in my book (Numbers Rule Your World).