MIT Algorithm Tackles Network Data Congestion

• By Dian Schaffhauser
• 06/14/18

A research project at MIT has developed an algorithm in which a constantly updating network — of sensors, drones or data-sharing vehicles — minimizes how much new information is received at any moment to avoid data congestion, while still keeping the most important data "as fresh as possible." Right now, the approach works for "simple" networks; but eventually, the team expects to tackle complex ones as well.

The project is important in an era of smart cars, manufacturing and delivery. Sensors are designed to feed fresh data to a wireless network quickly and continuously, to allow the network to maintain a current status on all variables in the system, as an MIT article explained. Yet, a wireless channel can only transmit so much data without overwhelming the network. The algorithm addresses that problem by tackling data exchange from the perspective of the recipient.

"If you are exchanging congestion information, you would want that information to be as fresh as possible," explained Eytan Modiano, a professor of aeronautics and astronautics and a member of MIT's Laboratory for Information & Decision Systems, in charge of the research project. "If it's dated, you might make the wrong decision. That's why the age of information is important."

Modiano said he first got interested in the problem while thinking about how unmanned aerial vehicles need to exchange information with each other to avoid collisions. "If they don't exchange this information often enough, they might collide," he noted. "So, we stepped back and started looking at the fundamental problem of how to minimize age of information in wireless networks."

In a paper titled "Optimizing Age of Information in Wireless Networks with Throughput Constraints," lead author, MIT Ph.D. candidate Igor Kadota, and co-author Abhishek Sinha described how a smart car uses multiple sensors to transmit time-sensitive information to a monitor "over unreliable wireless channels." Each sensor is sampling information from some part of the car — the tire pressure, fuel amount, proximity to obstacles, engine performance — and transmits the data to the monitor, which is continuously receiving new information about all of those components. However, the wireless channel poses limitations in how much data it can process at any given instant. The system has to manage the available channel resources in the most optimal way.

The approach modeled a single-hop network, with a single data receiver and multiple nodes. The question the researchers wanted to answer: "Which node should transmit data at which time, to ensure that the network receives the freshest possible data, on average, from all nodes?"

The solution involves calculating an "index" for each node, based on age, freshness of the data being transmitted, the reliability of the channel doing the communicating and the "overall priority" of the node in the entire system. That index can change instantly. The algorithm directs the node that has the highest index at any given moment to feed its data to the receiver. The result: the nearest possible fresh data in a simple wireless network setup.

While the team is testing its index framework on single-hop networks, the group is also planning to address ever-more complex networks as well. Modiano said future work would look at a network with multiple base stations, which would emulate more real-life scenarios in the age of information.

The paper won a "best paper" award during the IEEE International Conference on Computer Communications (Infocom) in April. The research was supported by the National Science Foundation and the Army Research Office.