Students & Turtles Mesh

Building on the theoretical study of discontinuous networks, computer scientists at the University of Massachusetts in Amherst have come up with a WiFi network with a very practical application: helping students catch buses in a timely fashion.

Associate professor Brian Levine heads the UMassDieselNet project, a wide-area network on Amherst's 150-square-mile bus system that gives passengers limited Internet access onboard, and will eventually provide riders with up-to-the-minute information on buses' locations and arrival times at particular stops.

The DieselNet project, Levine says, arose out of work on a larger problem in wireless networking: the absence, in many environments, of what he calls "unpartitioned networks."

"There are lots of places that have partitions," explains Levine. "For instance, a region like New Orleans that's been hit with a natural disaster, when all the infrastructure has gone down. The power's out, cell towers are down, how can you maintain a network?

"Or areas where no infrastructure exists in the first place -- like India in particular."

Trying to achieve some level of connectivity and information flow over partitioned, or discontinuous, networks, Levine and his colleague Mark Corner came up with a model that is more "disruption-tolerant" -- such as a bus moving through a town where WiFi hotzones appear and fade as the bus follows its scheduled route.

The result is some level of connectivity whenever possible, so that information -- on the bus's location, flowing away from the bus, or flowing to the bus from Web pages requested by passengers' computers onboard -- can be updated on a regular, if sporadic, basis. Each of the 40 UMass Transit buses carries a Linux-based computer with a 40-Gig hard drive, a GPS transmitter, an onboard WiFi access point for the passengers, and a second 802.11b USB-based card that constantly scans for APs on other buses passing by or any open AP that the bus encounters.

Like a panhandler moving through a crowded city square, the bus receives data from any openhanded source it passes, whether the source is another bus or a coffeehouse or a public WiFi hotspot (there's one on the UMass campus and one at the main bus terminal). By its nature, the system is imperfect (and it may one day be superseded by a ubiquitous high-speed network blanketing the city, such as WiMax) -- but it's designed to operate in imperfect environments.

"It's a little like the Pony Express," comments Levine. "You have a very delayed network, but it's necessary when there's no other way to get communications through."

The cost (supported by research funding) has been relatively minimal: around $1,000 per bus for the basic equipment plus $30 to $40 in monthly cellular-service fees per bus. Levine intends to shift the system to WiMax, which would require only a one-time connection fee rather than monthly charges, as soon as that technology becomes available.

Ultimately, Levine and his counterparts at UMass Transit hope to build out the system so that students hustling to make an early-morning class (or a late-night rave) can call up arrival times for specific stops on specific routes. Web-based systems like this already exist, most prominently NextBus, which covers major routes in cities in 14 U.S. states; but the goal of DieselNet is to create a city-specific network that will send arrival and departure information to a variety of mobile devices including SMS-equipped cellphones.

The first version of this real-time system will go into beta mode this summer -- a time, Levine notes, when demands on the transit system are at their lowest.

Meanwhile the discontinuous-networks model has found other applications -- such as tracking a threatened population of turtles in the wild near Amherst. Levine's colleague Corner has fashioned a mini-AP that can be strapped to a turtle's shell. Weighing just 50 grams, each unit includes a GPS tracker, a solar cell and battery, a mini-computer to store information, and a wireless transmitter. Each backpack-toting amphibian acts as a node in a "mesh" that transmits information back to the laboratory computer when one turtle wanders in range of a base station.

Ultimately, Levine says, DieselNet and the turtle-tracking project are testbeds for building real-world networks in situations where ideal, continuous-networking conditions will never prevail.

"It's a great opportunity to have a positive impact on the community by deploying a real network," he adds. "Mathematicians use paper, chemists use labs, and wireless professors need enormous networks that are up and running to do their research."

— Richard Martin, Senior Editor, Unstrung