About 1.3 million vehicles are involved in weather-related accidents each year. Cars are increasingly equipped with sensors that could trigger early warning systems about dangerous conditions such as icy roads with valuable real-time data, but analyzing and distributing that data currently requires costly communication between cars and cloud computers. Researchers at Rensselaer think there’s a better way.
“We want cars to communicate with one another—sharing and analyzing information—as well as with the cloud, so that they can learn from each other, reducing the expensive bandwidth that comes with contacting the cloud,” said Stacy Patterson, the Clare Boothe Luce Assistant Professor of Computer Science. She is developing the approach with Montana State University researcher Mike P. Wittie, in a new project supported by the National Science Foundation (NSF).
The project, “A Hybrid Vehicle-Cloud Solution for Robust, Cost-Efficient Road Monitoring,” is supported with a two-year $250,000 NSF grant. The research combines the best aspects of existing approaches to big data analytics, which generally fall into two categories: centralized approaches where data is sent to a single location for analysis; and distributed approaches where data analytics is performed by passing messages between participating entities. Centralized approaches are often simpler, but difficult to scale. Distributed approaches, while scalable, are difficult to analyze and may be disrupted by changes in the network such as loss of network connectivity.
“Our project combines the best of both approaches to get scalability and guaranteed performance,” said Patterson. “This is a hybrid architecture, a new way of performing data analytics that could have applications beyond vehicle networks, such as smart grid design and Internet of Things applications.”
Conceptually, the project leverages sensors and networking capabilities already present in new vehicles, and uses connectivity to the cloud to provide a stable backbone when necessary. The proposed system will be designed from the ground up specifically to support data analytics at a low communication cost. The result will be a scalable, agile, cost-efficient solution for real-time, widespread sensing and estimation in vehicle networks.
We want cars to communicate with one another—sharing and analyzing information—as well as with the cloud, so that they can learn from each other, reducing the expensive bandwidth that comes with contacting the cloud.”—Stacy Patterson
Patterson said the first step in the process is to conduct research confirming that they can use select sensor data to estimate road conditions. To do that, they will equip a small fleet of cars with a custom suite of sensors, gather data, and determine whether their application is capable of accurately estimating actual conditions on the ground using only a portion of that data.
“I know we can determine this literal ‘ground truth’ using massive amounts of information. But can we build an application that can estimate this ground truth without measuring every segment of the road and communicate that information to the cloud?” Patterson said. “The benefit of that is that you would need fewer measurements and less bandwidth to make a valuable determination.”
After gathering data, the team will build a simulator of their approach, in which they are able to feed in real-world data and simulate cars moving around, sampling the data, and using the algorithm and network to generate information on current road conditions.
“You can think of the goal of this project as a prototype, a system where we work out what information needs to be sent to what place and when, and how accurately it’s able to solve the problem,” said Patterson.
Patterson is an expert in dynamic networks, distributed algorithms, and cooperative control whose research creates new utility by linking the sensors that pepper modern consumer goods. Her current research is aimed at enhancing the utility of sensors embedded in automobiles, by creating real-time networks that allow automobiles to pool their individual information into a larger shared picture of driving conditions in the area.