Dynamic emergency exit displays are moving beyond static signage as a new AI model, Safe Step, offers real-time guidance during building fires. Developed by researchers at the National Institute of Standards and Technology (NIST) and detailed in the Journal of Building Engineering, Safe Step doesn’t simply react to current conditions; it anticipates fire spread to proactively adjust evacuation routes. “Fires can grow and spread,” said Hongqiang “Rory” Fang, a research associate at NIST and first author of the paper. “Our model forecasts how the fire is evolving and can help update emergency exit displays to direct people toward the safest exit.” Unlike traditional algorithms that prioritize the shortest path, Safe Step uses reinforcement learning to minimize exposure to hazards, calculating a fire safety metric called the fractional effective dose of toxic gases to guide occupants along safer routes.
Safe Step: AI for Predicting Fire Spread & Evacuation
Unlike previous systems reliant on identifying the shortest path to an exit, Safe Step utilizes reinforcement learning to assess cumulative hazards, factoring in the changing toxicity of gases over time. Researchers employed a metric called the fractional effective dose (FED) to quantify hazard exposure, with the model prioritizing routes exhibiting the lowest FED values. Testing against traditional algorithms in both simple and complex single-level building structures demonstrated Safe Step’s consistent ability to identify safer evacuation paths; for example, the AI can direct occupants away from a seemingly close exit if it anticipates that route will become compromised as the fire expands. Currently validated for single-story layouts, the team is now focused on extending Safe Step’s capabilities to multi-level buildings and incorporating a model of the behavior of numerous evacuees simultaneously. This advancement could address congestion at exits and even coordinate access for firefighters, facilitating rescue operations and fire suppression.
Reinforcement Learning Guides Safe Evacuation Route Selection
Building safety systems are becoming increasingly sophisticated, extending beyond static signage with new approaches leveraging artificial intelligence to guide occupants during emergencies. Current systems often rely on directing individuals to the nearest exit, but these methods fail to account for the dynamic nature of fire spread and resulting hazards along potential evacuation routes. Published in the Journal of Building Engineering, the model utilizes reinforcement learning, a technique where the AI learns through trial and error, to predict fire evolution and optimize evacuation strategies. In testing, Safe Step demonstrated an ability to identify safer alternatives even when the closest exit became compromised, for example, directing occupants away from a hallway filling with smoke toward a more distant but secure route. Future iterations envision an AI system where each occupant is represented as an individual, to better model congestion and coordinate access for firefighters. Fang stated, “This research is still in the early stages of research and development, but it represents an important step toward intelligent firefighting where effective use of advanced technologies can protect property and save lives.”
Fractional Effective Dose (FED) Metric Minimizes Hazard Exposure
This nuanced approach allows the AI to account for changing conditions, anticipating how toxic gas concentrations will evolve as an individual traverses a building during a fire. The implementation of FED allows Safe Step to make more informed decisions than simply reacting to current conditions. For instance, the model can assess a scenario where an initially clear exit becomes hazardous as a fire grows, directing occupants toward a more distant, but ultimately safer, alternative. NIST mechanical engineer Wai Cheong Tam explained, “We asked ourselves, ‘Can we build a better algorithm that predicts how the fire evolves, and in a way that helps save more lives?’” This predictive capability is crucial, as the model doesn’t require real-time fire simulations; instead, it relies on live sensor data to continuously refine its recommendations. During testing with complex single-level building structures, Safe Step consistently identified safe evacuation routes, outperforming traditional algorithms in scenarios where conditions rapidly deteriorated.
Dynamic Exit Displays & Multi-Agent System Future Development
These displays aren’t static signposts; they can now indicate whether an exit remains safe or redirect occupants toward alternative pathways, a capability tested in several “smart” buildings currently equipped with the technology. A significant future development involves transitioning from a single-agent model, one that calculates the optimal route for a single evacuee, to a multi-agent system. This approach envisions each building occupant represented as an individual “agent” within the AI, allowing for a more nuanced understanding of crowd dynamics and potential bottlenecks. Wai Cheong Tam notes the potential for coordinated responses, stating the team is exploring how the model could direct evacuees to different exits while simultaneously coordinating access points for firefighters. Such a system could alleviate congestion at exits, a common hazard during emergencies, and facilitate more effective fire suppression and rescue operations, particularly for vulnerable populations. Researchers anticipate that technologies like Safe Step could begin appearing in buildings within five to ten years, contingent upon regulatory approval and rigorous reliability testing, ultimately representing a crucial step toward intelligent firefighting and enhanced life safety.
“Fires can grow and spread,” said Hongqiang “Rory” Fang, a research associate at NIST and first author of the journal paper. “Our model forecasts how the fire is evolving and can help update emergency exit displays to direct people toward the safest exit.”
