A data-driven approach for ranking entry and exit points in UAV-assisted firefighting missions

Wildfires are a growing threat as they can lead to significant casualties and result in damages to the economy and environment. Although risk mitigation and prior preparation are important, some wildfire causes make the disaster difficult to predict and therefore to prevent, hence the importance of improving disaster response capabilities. In this paper, we tackle the problem of determining the entry and exit points for firefighting Unmanned Aerial Vehicles (UAVs) when approaching and leaving the wildfire zone. The entry and exit point are scored based on the time the UAVs spend in the fire zone and the time to reach the fire zone. The problem is formulated as a regression model, which is tackled using machine learning algorithms, namely decision trees and random forest. The methods are simulated and evaluated on synthetic data, and the results show that the approach was able to provide accurate rankings of the entry and exit points.