FireAIDSS: Real-Time and In-Situ Reconstruction of Wildfire Thermofluid Dynamics Through Physics-Constrained Inverse PDE Modeling
ISEF · 2026 Physics and Astronomy
Overview
Wildfires pose significant threats to ecosystems and communities worldwide. While visible flames can be readily captured, the underlying thermofluidic processes governing atmospheric fire spread, today the leading cause of fire-related property destruction, remain uncharted. This project presents FireAIDSS, a novel framework for real-time and in-situ computation of thermofluidic wildfire dynamics using sparse data and conservation laws. FireAIDSS consists of a perception module and an action module. The perception module includes a Physics-Informed Attention-CNN model capable of reconstructing and predicting the transient evolution of temperature and velocity fields from sparse observations. The AI model is trained with CFD simulations and a multi-stage curriculum to enhance physical fidelity. The action module comprises an intelligent drone swarm built from low-cost platforms for multi-agent data collection. The two modules are linked by a feedback-based search strategy where predictions of the AI model optimize the next-step motion of the drone swarm, enabling efficient information gathering and rapid convergence to accurate reconstructions. In full-scale experimentation, FireAIDSS achieves an average temperature error of 2.97 K and wind velocity error of 0.08 m/s in a space of 2×2×1 m³ within 58s, meeting emergency response requirements in accuracy and efficiency. Across 200+ runs, it consistently maintains an update rate of 1.1 FPS and demonstrates collision-avoidance behaviors, highlighting its environmental robustness. As the first system making wildfire dynamics spatiotemporally resolved in real time, FireAIDSS democratizes wildfire intelligence and offers a scalable and cost-effective solution to building wildfire resilience across wildland and urban settings.
Awards (1)
- Third Award of $1,200 $1,200
Competition history
- ISEF 2026
Resources
Related projects
ISEF · 2022
It's Flaming Out: Using Artificial Intelligence To Emulate Critical Aspects of Wildfire Growth
ISEF · 2025
An AI-Driven Thermodynamics Based IOT Sensor Network for the Ultra-Early Detection of Wildfires Without Line of Sight View of Fire
ISEF · 2021
FyreWatch: Deep Learning for Accurate Wildfire Environmental Conditions Detection
CWSF · 2026
Phoenix: Wildfire Prevention using an Autonomous Bionic Morphing Ornithopter and AI Models
ISEF · 2024
Developing a Novel Wildfire Spread Prediction Model Using Spatio-Temporal Analysis and Time-series Forecasting
ISEF · 2023
Developing a Deep Learning Model to Approximate the Convolutional Geometry of Wildfires
JSHS · 2024
Predicting Next-Day Wildfire Spread with Environmental Data and Machine Learning
ISEF · 2018
Stop the Flame in Its Infancy! Multivariable Early-Warning System for Low-Cost Prevention of Wildfire Proliferation
Closest projects by meaning, across every fair and year in the corpus.
Browse more like this
Source: Regeneron International Science and Engineering Fair