|
Intelligent
driving
Dozens
of cars in the Boston area are testing the latest generation
of an MIT mobile-sensor network for traffic analysis that
could help drivers cut their commuting time, alert them to
potential engine problems and more.
In the CarTel project, Professor Hari Balakrishnan and Associate
Professor Samuel Madden of MIT’s Department of Electrical
Engineering and Computer Science use automobiles to monitor
their environment by sending data from an onboard computer
-- which is about the size of a cell phone -- to a web server
where the data can be visualized and browsed. They do so via
pre-existing WiFi networks passed during a trip.
The resulting data, accessible from the web or a cell phone,
not only helps a driver track conditions specific to their
own car, but when combined with everyone else’s can indicate
historical and real-time traffic conditions at different times
of the day. “Everybody’s data is contributing to
collective views of what congestion looks like,” Madden
said.
“Our goal,” Balakrishnan said, “is to make
the data behind CarTel available to help you plan and organize
your commute and drives. We want to minimize the amount of
time spent in your car.”
For example, the current system, deployed since January on
50 Boston-area cars -- including 40 taxis -- tracks traffic
by monitoring each vehicle’s speed at different points
during a trip. Unlike other route-planning systems, “CarTel
understands where traffic delays are and recommends routes
to avoid them,” Madden said.
The system has already cut Balakrishnan’s commute to
MIT by 25 percent. It recommended a new route that, although
a few miles longer than the approach suggested by some mapping
web sites, is considerably faster in practice.
CarTel is also linked to a vehicle’s onboard diagnostics
system (available in all cars sold since 1996), so a driver
can check various parameters key to maintenance and be alerted
to potential problems.
There are two principal research efforts behind the system.
First, Balakrishnan, Madden and Jacob Eriksson (now at the
University of Illinois, Chicago) developed a way to connect
to WiFi networks that is 35 times faster than other systems.
“It can take about 15 seconds to connect using a regular
system, so in a car you are already past the WiFi location
by the time you get the signal,” Madden explained. QuickWiFi
can connect in 360 milliseconds. “It’s the difference
between whether you can use WiFi with a car or not.”
The majority of the work, however, is focused on managing
the huge amounts of data key to the system. Depending on the
sensors in use, CarTel can receive more than 600 data points
a second. So the team has developed two generations of software
“to synthesize all that data into interesting uses,”
Madden said.
One such use is new algorithms for traffic-aware routing,
or obtaining directions between two locations that take historical
and current traffic conditions into account. Balakrishnan
and Madden have developed these algorithms with graduate student
Sejoon Lim and Professor Daniela Rus, both of the Department
of Electrical Engineering and Computer Science.
“CarTel makes it easy to collect, process, deliver and
visualize data from a collection of remote, mobile and intermittently
connected nodes,” the researchers concluded in one of
several technical articles and conference presentations on
the work. Most recently, they described the research at the
Association for Computing Machinery’s Conference on Mobile
Computing and Networking (MobiCom) in September 2008.
This work is funded by the National Science Foundation and
the T-Party Project, a joint research program between MIT
and Quanta Computer Inc.
|
