[Paper Review] Multi-UAV trajectory planning for 3D visual inspection of complex structures

Stefan Ivić, Bojan Crnković+2 more|arXiv (Cornell University)|Apr 21, 2022

Robotics and Sensor-Based Localization4 citations

TL;DR

This paper proposes a multi-UAV trajectory planning method based on the Heat Equation Driven Area Coverage (HEDAC) algorithm for 3D visual inspection of complex infrastructure. It uses potential fields and distance fields to generate collision-free, camera-optimized trajectories that ensure uniform surface coverage, outperforming state-of-the-art methods in inspection completeness and robustness across wind turbines, bridges, and synthetic structures.

ABSTRACT

The application of autonomous UAVs to infrastructure inspection tasks provides benefits in terms of operation time reduction, safety, and cost-effectiveness. This paper presents trajectory planning for three-dimensional autonomous multi-UAV volume coverage and visual inspection of infrastructure based on the Heat Equation Driven Area Coverage (HEDAC) algorithm. The method generates trajectories using a potential field and implements distance fields to prevent collisions and to determine UAVs' camera orientation. It successfully achieves coverage during the visual inspection of complex structures such as a wind turbine and a bridge, outperforming a state-of-the-art method by allowing more surface area to be inspected under the same conditions. The presented trajectory planning method offers flexibility in various setup parameters and is applicable to real-world inspection tasks. Conclusively, the proposed methodology could potentially be applied to different autonomous UAV tasks, or even utilized as a UAV motion control method if its computational efficiency is improved.

Motivation & Objective

• Address the lack of research on multi-UAV cooperative trajectory planning for 3D infrastructure inspection.
• Enable autonomous, safe, and comprehensive visual inspection of complex 3D structures such as wind turbines and bridges.
• Ensure all surfaces are captured in the camera’s field of view through dynamic camera orientation control.
• Improve coverage quality and reduce inspection time compared to existing path planning methods.
• Develop a flexible, computationally robust framework applicable to real-world UAV inspection tasks.

Proposed method

• Adapt the HEDAC algorithm to generate potential fields that guide UAVs toward under-covered regions in 3D space.
• Use distance fields to compute UAV-to-structure and UAV-to-UAV separation, enabling real-time collision avoidance.
• Dynamically orient UAV cameras toward the nearest point on the inspected structure using the gradient of the distance field.
• Integrate trajectory generation and path planning into a single process by solving for time-parametrized trajectories directly in the potential field.
• Model the inspection domain as a 3D target density field offset from the structure’s surface to guide coverage.
• Apply a modified HEDAC formulation that minimizes the difference between desired and achieved coverage over time.

Experimental results

Research questions

• RQ1 Can a HEDAC-based multi-UAV trajectory planner achieve uniform 3D surface coverage of complex infrastructure while avoiding collisions?
• RQ2 How does the proposed method compare to receding horizon planning in terms of inspection completeness and coverage quality?
• RQ3 To what extent does dynamic camera orientation toward the nearest surface point improve visual inspection effectiveness?
• RQ4 What is the impact of UAV count and structure complexity on coverage convergence and inspection efficiency?
• RQ5 Can the method be applied to real-world inspection tasks such as wind turbines and bridges with high fidelity?

Key findings

• The HEDAC-based method achieved superior surface coverage compared to receding horizon planning, with more complete inspection of all structural components.
• In the wind turbine inspection scenario, both UAVs successfully captured all components, demonstrating feasibility for real-world slender structures.
• The bridge inspection with five UAVs achieved near-uniform coverage of all surfaces within a limited time window.
• Coverage convergence was achieved across all test cases, with convergence speed and completeness depending on structure complexity and UAV count.
• The method produced stable, collision-free trajectories even in dense UAV configurations, such as 100 UAVs in a unit cube.
• Despite strong performance, the algorithm’s computational cost prevents real-time onboard motion control without optimization.