Imagine a novel location-aware information system where objects such as cars, humans, or cell-phones, equipped with sophisticated sensors, collect information about their physical environment. They either report this information in response to queries or spontaneously disseminate it around. Examples of such data that can be collected include traffic conditions (e.g., travel times) as measured by cars, allergen levels (e.g., pollen levels) as measured by sensors on human beings, or spectrum quality (e.g., available bandwidth) as measured by cell phones. Travel times, measured by one car, are of interest to other cars that are likely to take that route. Information about presence of allergic material in a region is of interest to others who are sensitive to the toxins and are intending to pass by that region. Signal quality maps are of interest to the cell phone users who want to know the best locations for making a call.

In such a system, each mobile sensor contributes a small piece of information to the "overall picture", which is aggregated from multiple such individual reports. We call such an environment Grassroots¹ to signify the "bottom up" information gathering and dissemination involved. Grassroots consists of a massive number of individual agents, called dataflies, that move around, and collect, summarize, and classify information about their immediate physical environment. Grassroots is different from the usual sensory environment since it relies on mobile sensors (dataflies) rather than on a fixed predefined infrastructure. Dataflies can visit areas that are not instrumented by sensors. Also, dataflies can offer multiple reports of the same physical space (i.e., cars traveling through the same road segment, or cell phones reporting signal strength in the same area). This highlights two important characteristics of Grassroots environment: redundancy, which imparts robustness, and dynamic nature of coverage, which changes with the location of its dataflies. Grassroots architecture has the potential to create a highly scalable and robust information acquisition system. 

¹ Grassroots effort is often used to describe a political campaign that relies on small contributions or large scale individual efforts. This is similar to the way an overall picture of the environment is obtained from data collected from individual mobile data sources.

We believe that transportation is one of the domains that is ideally suited for Grassroots architecture. "Traffic Congestion has become part of daily life in 
many places as traffic continues to increase on a relatively unchanging highway network. Highway congestion is not just a problem of recurring "rush hour" delay in major cities. More than half of all congestion is non-recurring, caused by crashes, disabled vehicles, adverse weather, work zones, special events and other temporary disruptions to the highway transportation system. Unless we manage highway congestion, our nation will continue to incur economic costs in forgone productivity, wasted fuel, and a reduced quality of life." According to the Texas Transportation Institute (TTI), the total congestion "bill" for 75 urban areas in the United States in 2000 came to $67.5 billion, which was the value of 3.6 billion hours of delay and 5.7 billion gallons of excess fuel consumed. One possible way of controlling the extent of congestion is by disseminating traffic information. This gives an opportunity to the drivers to bypass congested routes thus preventing congestion from building up. Today a number of commercial systems exist for collecting and disseminating traffic information (e.g., Traffic.com, Metrocommute, EtakTraffic). However, these systems tend to cover select highways while leaving out a major fraction of roadways, thereby creating a "digital divide". In case of congestion, if any of alternate routes falls outside the coverage zone, drivers are left to guesswork, past experience, and their instincts in deciding on a faster route to their destination. The main factor that prevents these systems from covering the entire road network of US is the cost involved. Each of these systems requires an infrastructure to be deployed (e.g., helicopters; static cameras at busy intersections and entrances to tunnels/bridges; traffic flow sensors along major highways). This represents a huge amount of money in one-time deployment cost, and a significant recurring cost in maintenance. In order to address the problem of traffic congestion, there is an urgent need for a solution that can be deployed on a large scale at a low cost. 

In this work, we present our vision and architecture of a novel information system for collecting and disseminating traffic information that enables all vehicles to be completely aware of the traffic conditions (travel time) relevant to them. We would like the system to be highly scalable, able to cover the entire road network of US. Most importantly, we would like the system to require zero or minimal additional infrastructure. This requirement ensures that cost does not become an issue preventing widespread deployment. We believe that such a system has the best chance of being deployed. Our key idea is to turn vehicles themselves into sensors that collect and disseminate traffic information. 

References:

• Samir Goel, Tomasz Imielinski, Kaan Ozbay, and Badri Nath, Grassroots: A Scalable and Robust Information Architecture. Technical Report DCS-TR-523, Department of Computer Science, Rutgers University, June 2003.[PS][PDF]