Artificial intelligence for tomorrow’s firefighters

Artificial Intelligence is more than SIRI on your iPhone and computers that can win at Jeopardy. A project where artificial intelligence assists firefighters during a call-out is running in the USA, and offers huge potential.

A firefighter enters a burning building. In addition to his fire-resistant clothing, boots, gloves and helmet, he also has a head-up-display (HUD), which presents key data in his field of vision. Moreover, his clothing is fitted with various sensors that feed the artificial intelligence that follows the firefighter with the firefighter’s position, temperature data, toxic gases and other hazard warnings. The artificial intelligence analyses the data, simultaneously collecting information from different sensors in the building and from databases with technical drawings of the structure. Based on all the data, the artificial intelligence sends instructions to the firefighter via his HUD, enabling him to navigate safely through the building.

Moreover, if there is a group of firefighters who, for example, need to fight the blaze or locate trapped occupants, the artificial intelligence can suggest ways in which the firefighters can work together to successfully perform their task.

AUDREY is here
The above example could actually become reality within a few years, as artificial intelligence is already under development. It – or should we say ‘she’ – is called AUDREY, which is an acronym for ‘Assistant for Understanding Data through Reasoning, Extraction, and sYnthesis’. And the above scenarios are just some examples of where AUDREY can be employed.

AUDREY is the result of a joint project between Jet Propulsion Lab (JPL), which is a part of NASA and administered by the California Institute of Technology (CALTECH), and the Department of Homeland Security in the USA. It is part of a larger project called Next Generation First Responders (NGFR), which will enhance safety for emergency services in the field through supporting and improving their focus and communication.

AUDREY is still at an early stage of development, and was tested last summer when it was fed data from various sensors and delivered recommendations to a mobile unit. And within the next year, AUDREY will be tested in real-life situations.

Firefighters’s guardian angel
AUDREY is based on a range of technological breakthroughs which will make it a possible assistant for tomorrow’s firefighters. For example, it is designed to integrate with the ‘internet of things’, where more and more everyday objects are connected to the internet, for example bathroom scales, underfloor heating systems, lighting, fridges and TVs. Even now, AUDREY is able to find the objects, gather data from their sensors and combine this data with data from the sensors which the firefighters carry as part of their equipment.

– When the firefighters are connected to all the sensors, AUDREY will in effect become their guardian angel. Thanks to the data which the sensors are registering, the firefighters will not, for example, run into a room where the floor is collapsing, says Edward Chow, manager at the JPL Civil Program Office and program manager for AUDREY.

Can observe and learn
Before data can be used, it must be filtered and processed.

– The prevalence of minisensors and ‘the internet of things’ can make a huge difference to first responder safety, how they are connected to one another and their understanding of the situation. However, the huge volumes of data are incomprehensible in their raw form and must be synthesised to usable, targeted information, says John Merrill, project manager for NGFR.

AUDREY can do this as well. It knows the different roles in connection with an emergency operation, and can thus provide relevant information to the right people without drowning all the firefighters connected to AUDREY in information. At the same time, AUDREY observes and learns during a call-out. And once it has acquired enough experience, it will predict which resources will be needed later in the call-out based on how previous incidents have progressed. Like all artificial intelligence, AUDREY is only as good as the data it receives. And the more data there is, the greater the likelihood of it being able to supply useful advice and instructions.

– Most artificial intelligence is rule-based: if x happens, then it does y. But what happens if it only receives some of the information? We use complex reasoning to simulate how people think. This makes it possible for us to provide more useful information to the firefighters than with conventional artificial intelligence, says Edward Chow.

06.09.17