Details published in the Journal of Building Engineering reveal a new approach to fire evacuation that goes beyond simply identifying the nearest exit. Researchers at the National Institute of Standards and Technology (NIST) have developed Safe Step, an AI model that predicts fire outbreaks and dynamically updates safe evacuation routes within buildings. Unlike previous algorithms that focused on the shortest path, Safe Step considers how hazards accumulate over time and moves occupants away from exits that may become unsafe due to fire spread. “Fires can grow and spread,” said Hongqiang “Rory” Fang, a researcher at NIST and lead author of the journal paper. “Our model can help predict how a fire will progress and update exit signs to direct people to the safest exit.” Some smart buildings are currently testing technology with dynamic exit displays that leverage real-time temperature and air quality data.
Safe Step: Fire spread prediction and evacuation using AI
The ability to accurately predict the spread of a building fire and direct occupants to safety represents a significant advance in fire protection technology. NIST researchers have developed an AI model to accomplish this. The model uses reinforcement learning, a type of artificial intelligence, to make decisions through trial and error and learns evacuation routes based on the building layout and data from NIST’s proprietary fire simulation tools. Importantly, Safe Step does not require real-time fire simulation. We rely on live sensor data that measures temperature and air quality to continually refine our recommendations. Researchers use fractional effective doses (FEDs) of toxic gases to quantify safety, with the goal of minimizing long-term crew exposure. In tests compared to traditional algorithms in a one-story building, Safe Step demonstrated a consistent ability to identify safer evacuation routes, even when initially nearby exits were compromised.
For example, the model can predict scenarios where as a fire intensifies, nearby exits become unsafe, directing occupants to alternative exits further away, but ultimately safer. NIST is currently working to expand the model’s capabilities to handle tall buildings and include multiple agents that simulate individual occupant behavior and potential congestion points, potentially adjusting firefighter access and evacuation routes.
Reinforcement learning guides safe evacuation route selection
New artificial intelligence models take into account the evolution of fire hazards and go beyond simply identifying the nearest evacuation site to dynamically guide building occupants to a safe location. This is in contrast to traditional algorithms that prioritize the shortest path based on initial conditions, which can lead people down dangerous paths when conditions change. “We asked ourselves, ‘Can we build better algorithms to predict how fires will progress and save more lives?'” said NIST mechanical engineer Wai Cheong Tam. Consider a scenario where a fire breaks out across the hallway. A traditional algorithm might tell you to cross immediately to the nearest exit. However, SafeStep can predict situations where that exit becomes untenable as the fire intensifies.
“Fires can grow and spread,” said Hongqiang “Rory” Fang, a researcher at NIST and lead author of the journal paper. “Our model predicts how a fire will progress and helps us update emergency exit signs to direct people to the safest exit.”
Minimize hazardous exposure with fractional effective dose (FED)
NIST researchers are refining fire evacuation strategies, focusing instead on minimizing cumulative hazardous exposure using a metric called fractional effective dose (FED). The approach, detailed in the Journal of Building Engineering, goes beyond traditional algorithms that prioritize shortest paths and takes into account the evolution of toxicity of fire byproducts generated during evacuation. The team’s AI model, called Safe Step, uses FED to quantify the severity of fire danger experienced by evacuees over time. A lower FED indicates decreased exposure and therefore increased safety. Safe Step does more than just calculate a safe route on the fly. Continuously assess risk based on real-time data from building sensors that monitor temperature and air quality. This dynamic assessment is critical because previously clear trails can quickly become dangerous as the fire spreads.
The model might instead recommend a longer route to the exit further down the hallway, recognizing that by the time the evacuee reaches the exit, the first, closer option may no longer be tenable. This predictive ability is achieved by assigning numerical values to the hazards, allowing the algorithm to choose the path with the lowest overall FED. Researchers are now expanding Safe Step’s capabilities to handle multi-story buildings and model the interaction of multiple evacuees, aiming for a system that can coordinate the safe egress of all occupants while facilitating access for firefighters.
“Although this work is still in the early stages of research and development, it represents an important step toward intelligent firefighting that can protect property and save lives through the effective use of advanced technology,” Huang said.
Dynamic exit indication and multi-agent system development
The integration of artificial intelligence into building safety systems is moving beyond theoretical models to practical testing. Dynamic exit displays are currently being evaluated in some smart buildings to provide real-time guidance to occupants during emergencies. More than just lighted signs, these displays are responsive interfaces that use data from sensors that monitor temperature and air quality to assess exit safety and dynamically guide people away from dangerous routes. This proactive approach represents a significant change from traditional evacuation strategies that prioritize the shortest route regardless of the progression of the hazard. This allows the model to account for crowding and adjust firefighter access, potentially directing evacuees to less crowded exits while allowing emergency responders to effectively enter the building.
