AI-Enabled Drone Swarms for Fire Detection, Mapping, and Modeling
This project will develop an integrated AI-based formation and onboard computing method for a fleet of heterogeneous drones to enhance fire detection and mapping while ensuring efficient data transmission. Currently, Unmanned Aerial Systems (UAS) face limitations in wildfire monitoring due to bandwidth constraints and the necessity for human operators. This initiative aims to create a hierarchical platform of multiple UAVs for long-term fire coverage, develop low-computation real-time collaborative learning methods for onboard fire detection and mapping, and transmit final fire maps to relevant parties. This project will also develop a digital twin environment for interactive fire management and fire behavior modeling.
This AI-based UAS framework, developed from a scientific perspective, aims to provide high-resolution, frequent coverage in high-risk fire zones. It integrates deep learning algorithms for precise fire detection and behavior prediction and employs data fusion techniques to assimilate information from various sensors. This framework is designed to autonomously navigate challenging terrains and atmospheric conditions, enhancing data accuracy and timeliness. It also incorporates advanced communication protocols to efficiently relay critical data, reducing reliance on human operation and traditional communication infrastructure.
This project harnesses advanced UAS technology to revolutionize wildfire management. High-altitude platforms (HAP) will function as command centers, directing lower-altitude drones for targeted operations. The project emphasizes AI integration for on-the-fly fire analysis, leveraging advanced algorithms for real-time detection, mapping, and predictive modeling of fire spread. This technological approach aims to optimize drone efficiency and streamline data processes. The synergy of AI-processed data with physics-based simulations will culminate in a digital twin interface, providing an immersive, data-rich operational tool.
- An edge distributed AI model will segment ignited areas and share fleet to fleet and fleet to HAP to reduce communication burden and compute the fire growth indicator (FGI).
- A rapid short term AI-based physics-aware fire spread prediction model will be developed from UAS-based fire mapping data that considers highly dynamic environmental changes.
- A wildland fire digital twin environment will create an immersive environment that jointly takes into account the demonstration of dynamic and exogenous data and provides a simulation platform for UAS fleet control and fire-distinguishing maneuvers.
Fatemeh Afghah is an Associate Professor with the Electrical and Computer Engineering Department at Clemson University and the director of the Intelligent Systems and Wireless Networking (IS-WiN) Laboratory. Her research interests include wireless communication networks, decision-making in multi-agent systems, and UAV networks. Her recent project involves autonomous decision-making in uncertain environments, using autonomous vehicles for disaster management and wildfire management. She is the recipient of several awards including the Air Force Office of Scientific Research Young Investigator Award in 2019, NSF CAREER Award in 2020, NAU’s Most Promising New Scholar Award in 2020, NSF CISE Research Initiation Initiative (CRII) Award in 2017 and Best Paper Award at INFOCOM WiSRAN in 2022. She is the author/co-author of over 150 peer-reviewed publications, and 7 US patents.
P. Haeri, A. Razi, S. Khoshdel, F. Afghah, J. Coen, L. O’Neil, P. Fule, A. Watts, N. Kokolakis, K. Vamvoudakis, “A Comprehensive Survey of Research towards AI-enabled Unmanned Aerial Systems in Pre-, Active-, and Post-wildfire Management“, under review 2024.
Xiwen Chen, Bryce Hopkins, Hao Wang, Leo O’Neils, Fatemeh Afghah, Abolfazl Razi, Peter Fule, Janice Coen, Eric Rowell, Adam Watts, “Wildland Fire Detection and Monitoring using a Drone-collected RGB/IR Image Dataset“, IEEE ACCESS, 2022.
A. Shamsoshoara, F. Afghah, E. Blasch, J. Ashdown, M. Bennis, “UAV-assisted Communication in Remote Disaster Areas using Imitation Learning”, IEEE Open Journal of the Communication Society, Special Issue on Aerial Wireless Networks: Drones for Communications and Communications for Drones, 2021.
A. Shamsoshoara, F. Afghah, A. Razi, L. Zheng, P. Fule, E. Blasch, “Aerial Imagery Pile burn detection using Deep Learning: the FLAME dataset”, Computer Networks, 2021.
H. Rajoli, P. Afshin, F. Afghah, “Thermal Image Calibration and Correction using Unpaired Cycle-Consistent Adversarial Networks”, IEEE Asilomar Conference on Signals, Systems, and Computers, ASILOMAR 2023.