Trailblazers: AI ‘guardian angel’ guides firefighters through burning buildings
AUDREY (the Assistant for Understanding Data through Reasoning, Extraction, and sYnthesis) is designed to collect data on temperatures, gases and other danger signals and transmit them to individual firefighters, guiding them safely through a blaze and recommending how they can best work with their team.
"As a firefighter moves through an environment, AUDREY could send alerts through a mobile device or head-mounted display," said Mark James of NASA's Jet Propulsion Laboratory, lead scientist for the AUDREY project.
The cloud-based technology could also work with wearable sensors placed on the firefighters’ clothes, where it could identify an individual’s GPS location, heat levels in other rooms, the presence of dangerous chemicals and gases, and find satellite imagery of a location.
"When first responders are connected to all these sensors, the AUDREY agent becomes their guardian angel," said Edward Chow, manager of JPL's Civil Program Office and program manager for AUDREY. “Because of all this data the sensor sees, firefighters won't run into the next room where the floor will collapse."
The technology also allows the system to filter and customize information to each individual user, sending only what is deemed relevant to their role.
Chow noted that the quality and quantity of data available is pivotal for the AI to be effective.
"Most AI projects are rule-based - if this, then that," he said. "But what if you're only getting part of the information? We use complex reasoning to simulate how humans think. That allows us to provide more useful information to firefighters than a traditional A.I. system."
AUDREY is part of the Next Generation First Responder (NGFR) program, launched by Homeland Security in 2015 for use by first responders, including police, in the field.
A virtual demonstration of the technology took place in June in San Diego, when the program was provided with data from a variety of sensors and asked to make safety recommendations which it sent to a mobile device. After nine months of development the DHS-funded project is in line for field tests next year.