Science & Engineering | News

Researchers Tackle How Cars Can Share Data via WiFi

As automakers continue working on the best approach for adding communications technology to their cars, a project involving researchers from MIT, Georgetown University, and the National University of Singapore will help ensure that WiFi has a fighting chance against other pricier offerings. The team has developed an algorithm that exploits car data in an ad hoc way.

Advanced WiFi technology in cars will help drivers identify and respond to potential dangers. In theory, cars would send messages about where they are and where they're going, and intelligent vehicle sensor technology would notify drivers as they entered risky situations or take action on their own without driver intervention.

But driving in traffic isn't a lone endeavor; it's a team activity. So cars need to be able to share data with each other. The problem is that exchanging information requires having it uploaded to the Internet from one car and then down again to another, and that would most likely involve a pricey paid service such as a 3G cell network or satellite connection.

Shortly, three students will present a paper that examines the use of ad hoc networks to allow "free" WiFi-connected cars to share their Internet connections. "Aggregation in Dynamic Networks," by MIT graduate student Alejandro Cornejo, Georgetown's Calvin Newport, and NUS's Seth Gilbert, puts forth an approach in which data from a fleet of hundreds of cars--that ad hoc network--is aggregated and then uploaded online to help any individual car identify and respond to traffic problems.

   Researchers Tackle How Cars Can Share Data via WiFi  

 

As Cornejo, lead author, described the problem, the layout of a network of cars is constantly changing in unpredictable ways. But as they move, and with enough data aggregated, traffic patterns begin to surface. The patterns would work to determine which cars are coming into contact with the largest number of other cars, and therefore should act as the aggregation nodes. The problem is, they can't be identified in advance.

The team began by examining the scenario in which every car within a grouping would reliably come into contact with some fraction of the rest of the group during a fixed period. As two cars from that grouping come into range of each other, they would "flip a coin" to decide which one will convey data to the network, Cornejo said. Over time, however, "we bias the coin toss," he noted. "Cars that have already aggregated a lot will start 'winning' more and more, and you get this chain reaction. The more people you meet, the more likely it is that people will feed their data to you."

He estimated that a thousand cars could see their data aggregated by only about five, even if some of those cars drop out of action, for example, by heading off to somewhere outside of the network.

In their paper, the researchers were able to show that if the network of cars can be envisioned as a series of dense clusters with only sparse connections between them, the algorithm will still work well. Oddly, Cornejo added, if the clusters are well connected, the algorithm doesn't work as well. "We can show that it's impossible to aggregate. It's not only our algorithm that fails; you can't do it."

The paper will be presented at the ACM SIGACT-SIGOPS Symposium on Principles of Distributed Computing, taking place during July in Portugal.

About the Author

Dian Schaffhauser is a writer who covers technology and business for a number of publications. Contact her at dian@dischaffhauser.com.

comments powered by Disqus