Challenges and Research Directions from the Operational Use of a Machine Learning Damage Assessment System via Small Uncrewed Aerial Systems at Hurricanes Debby and Helene (2506.15890v1)

Abstract: This paper details four principal challenges encountered with ML damage assessment using small uncrewed aerial systems (sUAS) at Hurricanes Debby and Helene that prevented, degraded, or delayed the delivery of data products during operations and suggests three research directions for future real-world deployments. The presence of these challenges is not surprising given that a review of the literature considering both datasets and proposed ML models suggests this is the first sUAS-based ML system for disaster damage assessment actually deployed as a part of real-world operations. The sUAS-based ML system was applied by the State of Florida to Hurricanes Helene (2 orthomosaics, 3.0 gigapixels collected over 2 sorties by a Wintra WingtraOne sUAS) and Debby (1 orthomosaic, 0.59 gigapixels collected via 1 sortie by a Wintra WingtraOne sUAS) in Florida. The same model was applied to crewed aerial imagery of inland flood damage resulting from post-tropical remnants of Hurricane Debby in Pennsylvania (436 orthophotos, 136.5 gigapixels), providing further insights into the advantages and limitations of sUAS for disaster response. The four challenges (variationin spatial resolution of input imagery, spatial misalignment between imagery and geospatial data, wireless connectivity, and data product format) lead to three recommendations that specify research needed to improve ML model capabilities to accommodate the wide variation of potential spatial resolutions used in practice, handle spatial misalignment, and minimize the dependency on wireless connectivity. These recommendations are expected to improve the effective operational use of sUAS and sUAS-based ML damage assessment systems for disaster response.

Challenges and Research Directions from the Operational Use of a Machine Learning Damage Assessment System via Small Uncrewed Aerial Systems

The paper "Challenges and Research Directions from the Operational Use of a Machine Learning Damage Assessment System via Small Uncrewed Aerial Systems at Hurricanes Debby and Helene" investigates the deployment of a ML damage assessment system utilizing small uncrewed aerial systems (sUAS) during the operational response to Hurricanes Debby and Helene. This work represents a pioneering effort in orchestrating a real-world application of sUAS-based ML for disaster damage assessment.

Summary of Key Findings

This research identifies four predominant operational challenges encountered during these deployments:

1. Variation in Spatial Resolution: The paper identifies that a broad range of spatial resolutions (1.65 cm/px to 25.3 cm/px) were observed in the captured data. This variation was not solely attributable to differences in flight altitude or sensor specifications but was often a tactical response to limited wireless connectivity in the field. Such variability can significantly affect the consistency of ML model outputs.
2. Spatial Misalignment: Discrepancies between a priori building polygon data and captured orthomosaic imagery were noted to introduce misalignments that compromised the accuracy of automated damage assessments. The necessity for manual corrections increased data-to-decision time and suggested an area requiring further refinement in automated alignment techniques.
3. Limited Wireless Connectivity: The communication constraints hindered real-time data transfer, crucial for timely decision-making in disaster response scenarios. The research highlights a critical dependency on wirelessly transmitting large datasets, emphasizing the need for advanced on-site processing capabilities.
4. Insufficient Data Product Formats: The traditional GIS formats proved challenging for field use, particularly where bandwidth and specialized knowledge were lacking. The shift to more universally understandable data product formats (e.g., CSV, PDF) was proposed to facilitate better integration into existing decision-making workflows.

Research Directions and Implications

Given the identified challenges, the authors propose three primary research directions:

1. Accommodating Spatial Resolution Variability: The research emphasizes the necessity for ML models capable of accommodating diverse image resolutions. Such advancements would allow for more reliable performance regardless of data quality variations imposed by operational constraints.
2. Automating Spatial Alignment: Developing robust methodologies for automated correction of spatial misalignments could significantly improve the accuracy and reliability of inferred damage assessments, thus expediting the data processing pipeline.
3. Reducing Wireless Connectivity Dependency: The integration of local Edge computing could mitigate the inherent delays associated with data transmission, thereby improving operational responsiveness. This would enable real-time processing and immediate access to actionable insights directly within the disaster zone.

Overall, this paper underscores the complexity and potential of utilizing sUAS-based ML systems for real-time damage assessment in disaster response. Implementing the proposed research directions could significantly enhance the efficacy and reliability of such systems. This work offers a pragmatic perspective on the operationalization of ML in high-stakes environments and provides a foundation for further exploratory work in this evolving field. The findings and proposed directions are invaluable for the robotics and ML communities as they advance toward robust, scalable solutions for disaster management and response.