Tracking cell phones to Track traffic

"...the Missouri Department of Transportation is finalizing a contract to monitor thousands of cell phones, using their movements to map real-time traffic conditions statewide on all 5,500 miles of major roads."

Full article here.

Comments

[Daniel]

Interesting, but you don't have to think very hard before you realize what's wrong with the concept. You're only tracking people who own and use cell phones (51 percent this year), and that distribution is certainly not uniform across race, class, gender and other demographics.

So while you will see patterns, imagine the engineering you'd have in a complex hydraulic system if you only saw half of the liquid you were trying to move. Talk about your sensitivity to initial conditions! If you've only got 50 percent on anything, your modeling is going to be pretty useless.

[Janet Edens]

Uh, like, what about the places where it's illegal to use cell phones in the car?

Sounds like Big Brother wants to know where you're going, who you are talking to and what you're printing.

Canada, anyone?

[DeweyS]

Actually, I think it will work well for their current application. Demographics don't really matter -- they're just measuring the speed at which individual cell phones are moving. You don't have to see the whole liquid -- you can presume that the surrounding liquid is moving at roughly the same rate as the piece you're observing.

They're not yet trying to track the full population.

Also, the cell phones don't have to be "in use" to track -- any cell phone that's on is in low level communication with a cell tower -- that's how they know how to ring *your* phone when the call comes in.

[Daniel]

Yeah, I've got that part. But if all I can track are people with cell phones, and I'm trying to solve traffic problems in a complex grid, there's a lot of information I'm not getting about why that fluid is moving at that particular speed.

I suppose it depends in part with what I want to do with my data. If you're trying to build models, I question whether they're going to work in a complex grid. If you're just trying to monitor speeds and you're not concerned about the data's predictive power, then have at it.

[DeweyS]

Hmmmm. I don't think having only 50% trackable is a problem. We analyze lots of things measuring much less than 50% of the population -- we just have to figure out if the portion we can measure cheaply is statistically a good representative of the general population.

You're quite right that it doesn't give them any information on *why* traffic is behaving. On the other hand, since they're tracking cell phones maybe they could call someone and ask :)