CarTel is a mobile sensor network data collection system.
The idea is we put a little bit of hardware in cars--
this hardware consists of a Linux box and some sensors--
and then we use this data that we've collected from the cars
to answer sort of interesting questions about the world.
So we synthesize data from a bunch of cars
together in order do things like give users
a picture of what traffic in the city currently looks like.
We have a range of sensors that are deployed in the cars.
All of the cars include a GPS, a Global Positioning System,
as well as a Wi-Fi radio that they can use for getting data
off of the cars.
Other sensors that we put on some of the cars
include accelerometers, which can do things like measure
the quality of the ride, or the surface of the road
as the car drives over it, so that, for example, you
can detect vibrations or bumps as the car hits them.
In the CarTel project, we have several different ways
of transmitting data off the cars.
And one of things we've--
one of the technologies we've been working on
is developing different means of data transmission.
And the most basic version of data transmission
is just to use a cell phone.
So this is somebody has a cellular data plan,
they connect the cell phone up to the car,
they connect a wireless data card up to the Linux boxes
in the car, and then use that to transmit information off.
And one of the things that we spend
a lot of time developing in the project
is technology to use Wi-Fi.
Part of the technology we've developed
is to-- allows cars as they drive by these wireless access
points to very quickly associate with the access points
and transmit little bits of information
every time they see one of these access points
that's open and available for access.
As the cars are driving along, they're
transmitting this data off.
And what happens is that they transmit it
to a server or a portal that's running on a machine at MIT.
And that portal is in charge of taking all this data
and synthesizing it together in order
to present a sort of global view of what's going on
in the world, or what these sensors in cars
are observing about the world.
Typically, this data gets put into a sort
of a visualization that gets overlaid on top of a map.
Users can then browse this visualization
in order to learn various things about what's
going on inside the world.
We have a range of different applications
that we've developed.
So the first application is-- and the simplest one--
is just one that allows users to browse their own driving
So you may be interested, for example,
in comparing the amount of time that it
takes for you to drive from home to work
on various different routes.
So you can compare your freeway drives
to your not freeway drives, if you
have several different routes from your house to work
or from your work to your house.
Other applications that do sort of more interesting things
that combine lots of data from different cars
together include one that we call the Pothole Patrol, which
is an application where we use these accelerometers on the car
that are measuring the quality of the ride to detect where
potholes are in the city of Boston,
and then categorize them in terms of their severity.
The third application that we spent a lot of time developing
is a traffic portal.
So the idea here is that we use the cars,
as they're driving along, we can learn things
about how fast they were able to drive on every road.
And then for each road segment, each little piece of road,
like, for example, Mass Ave in front of MIT,
we can compute a distribution of the delays
that your car has experienced when they were driving on that.
The first goal is to explore the use of cars or mobile nodes
as sensing devices.
So how can we use these mobile things
in order to measure the world?
And so it's about--
one goal is about just developing
the technology that's needed to get data off of these cars that
are moving around in the world, and to bring it back
to a server, and to synthesize it and store it together.
And some of the challenges there include things
like dealing with variable connectivity,
that the cell phones will cut in and out as you drive around
or there isn't continuous wireless conductivity available
So one goal is sort of a technology goal.
The other goal is an application goal,
which is to develop some of these-- use
these cars in order to develop applications that
are sort of new and different, and enable
things that will help people besides computer scientists.
This traffic portal that I mentioned
is an example of something that everybody could take advantage
Anybody would be interested in knowing what the best
route to take from A to B is.
Obviously, things like MapQuest and Google Maps
have huge uptake.
And adding traffic to those would just
increase their value.
So one of the things that we're doing in order
to bring the cost of this box down
is we're looking at solutions that
include smaller amounts of hardware inside of the car
because we think that hardware cost is a barrier to entry.
So that, we think, will allow us to produce a piece of hardware
that goes into the cars that maybe only costs $30 or $40
per car, which would be something that we think
an average person might be willing just
to put in their car in order to benefit from this technology.