This post is also available in: heעברית (Hebrew)

Sensors can provide huge amounts of information about an environment. Video from unmanned aerial vehicles (UAVs) and social media posts could provide valuable information about a road network during flooding. However, too much video footage and too many tweets could overwhelm the computing power of an automated system that is trying to rapidly determine the best routes for ambulances that are responding to the disaster.

Worcester Polytechnic Institute (WPI) researcher Raghvendra Cowlagi has been awarded $530,029 by the National Science Foundation (NSF) to determine strategies that could sift through large amounts of sensor data to improve the operations of autonomous systems. The work could have applications for traffic management, aerial parcel delivery, deployment of scarce resources during emergencies, and responses to natural disasters or industrial accidents.

The three-year project will use mathematics to address the problem of identifying the most useful information in large data sets for automated decision-making. “This is about trying to access the most useful information that may be contained in massive sets of data collected by everything from traffic cameras to social media accounts,” Cowlagi said. “We want to find the most relevant data, fuse together data from diverse sources, and enable automated systems to reach decisions with fewer, but more meaningful, pieces of information.”

Cowlagi will develop mathematical models to configure sensors according to an identified objective and then test them using publicly available Air Force Research Laboratory data collected by cameras, seismic and acoustic sensors, and radar. 

He will also conduct indoor and outdoor experiments at WPI using wheeled robots on the ground and miniature UAVs equipped with cameras.

The project “could help those who want to use drones to safely deliver packages in a neighborhood or those who want to plan for the optimal distribution of medical resources during a crisis,” he was cited by wpi.edu. “We can’t simply try to build bigger and more powerful computers. We need to find ways to deliberately collect the most relevant information in order to make faster, automated decisions.”