Multi-UAV Flood Monitoring via CVT with Gaussian Mixture of Density Functions for Coverage Control

TL;DR

The paper addresses real-time monitoring of unknown flood regions using a multi-UAV swarm. It develops a density-driven coverage framework that combines Centroidal Voronoi Tessellation with a Gaussian Mixture of Density Functions to represent evolving flood extents, updating as detections arrive and image moments determine μ_i and Σ_i. A centralized controller moves UAVs toward density-weighted centroids, and simulations in ROS/Gazebo with 16/20/24 UAVs show that GMDF improves coverage compared with axis-aligned Gaussian density. The work demonstrates adaptive, more accurate flood monitoring, while noting the need for decentralization to enhance scalability and resilience in real-world deployments.

Abstract

This study presents a control strategy for coordinating multiple unmanned aerial vehicles (UAVs) to monitor unknown flood regions and estimate the extent of inundation. The proposed method adopts a density-driven coverage framework based on Centroidal Voronoi Tessellation (CVT), in which the density function is modeled using a Gaussian Mixture of Density Functions (GMDF). This formulation provides a more accurate characterization of inundated areas compared to conventional axis-aligned Gaussian models. The performance of the two density modeling approaches is systematically evaluated under different UAV fleet sizes (16, 20, and 24), with multiple simulation trials conducted in the ROS/Gazebo environment. The results show that the GMDF-based formulation consistently achieves higher coverage rates, demonstrating its effectiveness in enhancing flood monitoring and improving UAV spatial distribution.

Figures (6)

• Figure 1: Illustration of a UAV swarm monitoring a flood-affected region. Multiple UAVs collaboratively cover the inundated area to overcome the limitations of single-UAV monitoring and satellite-based remote sensing.
• Figure 2: Illustration of the control strategy: (a) Step 1: UAVs are initially distributed within the workspace; (b) Step 2: a multimodal density function is generated, with each peak corresponding to a UAV that detects floodwater; the density represents the estimated distribution of inundated regions; (c) Step 3: a CVT is constructed based on the density, and the generated centroids serve as the target positions for UAVs, guiding those that have not yet detected flood areas toward the inundated regions; (d) Step 4: as UAVs move toward the flood regions, additional peaks are incorporated into the density whenever new flood detections occur, thereby continuously updating the estimation.
• Figure 3: Graphical explanation of how $\mu$ and $\Sigma$ are derived from UAV images: (a) $\mu$ is computed as the centroid of the segmented flood region, and $\Sigma$ from the image moments; (b) UAVs move toward flood areas based on the updated density function.
• Figure 4: Structure of the ROS/Gazebo-based simulator used for flood monitoring experiments.
• Figure 5: Snapshots of flood coverage (left), the corresponding CVT (middle), and the density function $\phi$ (estimation of the inundation area) at (a) $t=0$ s, (b) $t=25$ s, (c) $t=50$ s; (d) UAV paths; (e) Coverage comparison of the axis-aligned Gaussian density function and the Gaussian mixture of density functions under different UAV numbers.

Theorems & Definitions (2)

• Remark 1
• Remark 2