# Tracking Traffic Via Cell Phones

### from the *see-me-feel-me-follow-me* dept

There are plenty of companies around that want to send traffic data to mobile phones so people can avoid jams. But some are also working to get traffic data from cell phones, by monitoring their movements along roads and highways to get an idea of how well traffic is flowing. The idea’s been around for a while and was tested in Finland a few years ago, and now the state of Missouri has awarded a contract for a company to track phones across the entire state. While the Department of Transportation promises all the information will be anonymous and won’t be used for any other purpose, privacy advocates are concerned that the next step will be to track speeders or to monitor people’s movements surreptitiously. But the system seems like it would suffer from many of the same pitfalls as the earlier Finnish one, and inaccuracy could cause it more problems than invasions of privacy. For instance, how can it differentiate between 30 cars with just a driver and a bus with 30 passengers, or be able to distinguish between a number of people near a road, but not moving and stopped traffic? There are a lot of unanswered questions about how the system will work, and indeed, how well it will work. And for less than $3 million a year to monitor 5,500 miles of roads, it almost sounds too good to be true.

## Comments on “Tracking Traffic Via Cell Phones”

## Dirichlet Distribution

Actually, the concern of buses carrying 30 cell phones is relatively easy to solve, given the proper statistical technique. We can assume that buses constitute a certain proportion of traffic, and they will have a higher mean of phones-per-vehicle, but with higher variance as well. Assuming the number of buses follow a Poisson distribution, we can perform Poisson regression, partial F-tests, Newton-Raphson algorithms, or any number of other statistical techniques to derive the overall vehicle density. We can additionally assume the software can be “trained” to calibrate the proper parameters.

## Re: Dirichlet Distribution

ok………interesting…..now can you repeat that say..in english….coz i is foreigner…

## Re: Re: Dirichlet Distribution

If 30 people are travelling at 30 mph in a bus, aren’t you just establishing that the stretch of road they’re on is moving at 30mph? Doesn’t matter how many people are in the bus. Gay.

## Re: Re: Re: Dirichlet Distribution

Yeh if it is moving at 30mph it may show 30 people travelling etc…however the problem may lie in 30 people stopped (bus stop with a lot of people getting on, or nobody letting the bus out etc) on a 30 mile an hour road this may suggest a traffic jam with 30 people. I believe the resolution of most current GPS apparatus to be about 1-2m which is not small enough to distinguish between 30 people on a bus and 30 people on say a single decker bus. Not being familiar with some of the statistical equations you mention I can’t comment on them – however I hope I have highlighted their problem.

There is soon to be 10cm resolution GPS systems but these will be expensive and not financially viable in this instance.

## Re: Re: Dirichlet Distribution

What I’m saying is that the mathematical modeling for this situation is relatively straightforward. Any first-year graduate student in statistics should be able to do it. The field of multiple linear regression is all about this.