Dynamic Risk-Aware Graph Cross-Attention for Resilient UAV–UGV Disaster Navigation and Relay Coordination
Contributors
Rajanikanth Javvadi
Shiva Shankar Reddy
Keywords
Proceeding
Track
General Track
License
Copyright (c) 2026 Sustainable Global Societies Initiative

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
Abstract
In post-disaster environments, real-time multi-robot navigation is difficult due to the dynamic nature of the terrain, the need to always maintain communication between the UAV and UGV through debris and the limited battery life of the UAV, which necessitates timely UAV handoffs. This paper introduces a Graph Neural Network (GNN) approach for multi-agent path planning (MAPPs) on a dynamically changing risk graph, continuously updated with UAV sensor data, while coordinating ground paths, aerial relay locations, and battery-aware task allocation. It includes five modules: Dynamic Risk Map Encoder (DRME), Multi-Agent Graph Cross-Attention (MGCA), Risk-Aware Local Action Block (RLAB), Communication-Relay Routing System (CRRS), and Battery-Aware Assignment Module (BAAM). DRM-PathNet was evaluated in Gazebo/ROS across five disaster scenarios with 1,200 simulation episodes and achieved a 94.8% path success rate, a mean survivor reach time of 7.8 min, 97.6% communication uptime, and 43.2 ha/UAV-hour area coverage. These improvements were both statistically significant ((p<0.001) and showed the effectiveness of the framework for coordinated and resilient multi-robot disaster response.