Abstract:

We consider coordinated aerial inspection of large-scale infrastructure using a team of explorer (LiDAR-equipped) and photographer (camera-equipped) drones. We propose a hierarchical framework that combines constrained trajectory generation with multi-agent task assignment to perform inspection efficiently and safely. A key feature of our approach is the use of sequential convex programming for photographer trajectory generation under dynamics, perception, and communication constraints. We also validate our approach in a ROS2/PX4 high-fidelity simulator that extends an existing ROS1-based CARIC benchmark (Cao et al., 2025). The proposed approach typically achieves shorter scoring duration than a baseline A*-based planner while maintaining comparable inspection quality and balanced drone utilization. The code for this work is publicly available at https://github.com/merlresearch/ros2_caric.