While the skylines of every major American city differ from each other, there is one thing many of them have in common: frustrated drivers who may suffer road rage from traffic delays, accidents and congested roads.
The Boston area, home to MIT, holds the dubious distinction of ranking third among cities with the worst road rage problems, according to a May 2007 AutoAdvantage survey. Local commuters deal with a small, densely-populated city that was planned before automobiles were invented, a dearth of parking, winter damage to roads—and, (in)famously, other drivers who don’t consistently use their turn signals.
The factors that contribute to road rage have become an accidental point of interest for a group of researchers at MIT’s Computer Science and Artificial Intelligence Lab (CSAIL). The group, led by Associate Professor Sam Madden and Professor Hari Balakrishnan, studies sensor networking, where small devices placed at different locations are used to measure and collect information about remote environments, such as animal habitats, building interiors, or in their case, roadways.
The team has installed sensor devices in 27 hybrid taxi cabs operated by PlanetTran, a Cambridge-based livery service. These devices consist of a PC memory card, GPS, a wireless card, and an accelerometer. This small computer is capable or measuring a range of variables including: the speed of the car, the vibrations of potholes in the road, fuel consumption, on-board vehicle diagnostics, and its global positioning.
The sensors transmit this information when they find a wireless signal from a participating wireless access point (such as MIT’s network). A special collection of software called CarTel, captures data locally on the cars and then rapidly identifies and connects to network access points to transmit the data to a server. The data from PlanetTran’s cars is then processed and combined to distill aggregate data and produce a variety visualizations.
Sensors capture data at different rates, depending on their use. For example, the GPS sensor used for positioning is read once per second, whereas accelerometers, used for road surface mapping and other applications, capture data hundreds of times per second.
Staying connected and self-contained
One of the key features of CarTel is that it is designed to be both tolerant of long periods of no connectivity, and flexible with respect to exact nature of the wireless connection used to transmit data.
Currently the software uses a combination of available WiFi, Bluetooth, and cellular connections to transmit data from the car. CarTel continuously seeks WiFi connections as cabs are on the road. When it finds a network it’s authorized to use, CarTel connects to it and transmits data over it until the car moves out of range. CarTel’s modified collection of networking software allows it to establish such connections in just a few hundred milliseconds, and maintain connections despite the noisy and loss-prone environment presented by moving vehicles.
“There are so many wireless connections out there now that you could get almost continuous connectivity using our software,” Madden says, adding that cabs encounter a new wireless access point once every four seconds on average.
Sensing—and seeing—the information in travel conditions
One of CarTel’s applications is computing information about traffic delays and providing drivers with routes with a maximum probability of avoiding delays or getting them to their destination by a deadline.
To create these routes, CarTel uses the cabs’ GPS data. Each cab covers about 150 miles of roadways a day, producing tens of thousands of GPS points each day. Three algorithms, developed by graduate student Sejoon Lim along with Professors Madden and Balakrishnan, and Professor Daniela Rus (also in CSAIL), process this data.
The first is a matching algorithm that matches the raw longitude and latitude coordinates from the GPS to the road segments on a map. The second algorithm computes how fast the car was able to travel on that road segment at different times of day. In the final phase, the software uses the travel statistics as input to a route planning algorithm, which gives travel times that reflect historical and current traffic conditions.
CarTel differs from other route planning applications because it provides travel times based on current road conditions, not travel estimates based solely on the distance and rate of speed permitted on the route, like MapQuest and GoogleMaps do. CarTel also saves all the statistics it collects, which can provide drivers with historical information about which route has proved to be the most reliable in getting them to their destination on time, even if it’s not the fastest.
Telling drivers how to go
In time, CarTel could help drivers avoid accidents, gridlock and traffic jams, find an empty parking spot, and know when their car may be in need of a tune-up. It could allow cities and towns to better maintain their road systems and plan more reliable routes with accurate travel time information. And also importantly, reducing the road rage stemming from being stuck in traffic.
But traffic management wasn’t CarTel’s initial focus of research. The project originally meant to explore how to make the components of the sensor devices mobile. However, the CSAIL team quickly realized the many practical applications the devices could be used for and the impact it would have in people’s lives.
“Cars are interesting because everyone has one and everyone has an opinion of how our roadways could work better, or the best way to get from one part of town to another,” Madden says. “Once a system like CarTel has captured and synthesized information about traffic and road conditions, it creates an amazing opportunity to build software that improves the lives of commuters.”
Such improvements, undoubtedly, would be welcomed by Boston drivers.