[Paper Review] View Planning and Navigation Algorithms for Autonomous Bridge Inspection with UAVs

Kevin Yu, Prajwal Shanthakumar+4 more|arXiv (Cornell University)|Oct 3, 2019

Robotics and Sensor-Based Localization12 references4 citations

TL;DR

This paper presents an autonomous UAV system for full bridge inspection in GPS-denied environments using 2D LiDARs, a supervisor for switching between local navigation routines, and a planner that generates optimal inspection paths. The approach ensures complete coverage of box girder bridges with minimal flight time, validated through simulation on a real bridge structure.

ABSTRACT

We study the problem of infrastructure inspection using an Unmanned Aerial Vehicle (UAV) in box girder bridge environments. We consider a scenario where the UAV needs to fully inspect box girder bridges and localize along the bridge surface when standard methods like GPS and optical flow are denied. Our method for overcoming the difficulties of box girder bridges consist of creating local navigation routines, a supervisor, and a planner. The local navigation routines use two 2D Lidars for girder and column flight. For switching between local navigation routines we implement a supervisor which dictates when the UAV is able to switch between local navigation routines. Lastly, we implement a planner to calculate the path along that box girder bridge that will minimize the flight time of the UAV. With local navigation routines, a supervisor, and a planner we construct a system that can fully and autonomously inspect box girder bridges when standard methods are unavailable.

Motivation & Objective

• To enable fully autonomous UAV inspection of box girder bridges without relying on GPS or compass data.
• To develop local navigation routines for flight along girders and columns using 2D LiDARs.
• To design a supervisor that intelligently switches between navigation routines based on structural context.
• To create a planner that computes the optimal sequence of routines to minimize flight time and ensure full coverage.
• To validate the system on a real bridge model with simulated UAV flight under GPS-denied conditions.

Proposed method

• Two 2D LiDARs are used to detect bridge girders and columns, enabling real-time relative navigation.
• Local navigation routines (GL, GR, CU, CD) are defined for specific bridge surfaces, each with dedicated control laws.
• A supervisor module determines when to switch between routines based on LiDAR-based structural recognition and state estimation.
• A Global TSP (GTSP) solver is used to compute the optimal sequence of coverage for polygonal planner surfaces derived from the bridge geometry.
• The planner inputs are generated from a discretized bridge model, with clusters representing surface segments and nodes indicating coverage directions.
• The system integrates planner output with real-time UAV control via MAVROS, enabling execution of the full inspection path in simulation.

Experimental results

Research questions

• RQ1 How can UAVs autonomously inspect box girder bridges when GPS and compass are unavailable?
• RQ2 What local navigation strategies enable stable flight along bridge girders and columns using only 2D LiDARs?
• RQ3 How can a supervisor dynamically and safely switch between navigation routines during inspection?
• RQ4 What path planning approach guarantees full coverage with minimal flight time in the absence of global GPS knowledge?
• RQ5 Can a hybrid system of local routines, a supervisor, and a global planner achieve robust and optimal inspection in real-world bridge environments?

Key findings

• The system successfully planned and simulated a complete inspection path on a real box girder bridge model with 11 polygonal surfaces.
• The GTSP-based planner computed an optimal sequence in 1.92 seconds, ensuring full coverage with minimal flight time.
• The UAV executed the planned path in simulation, with mode transitions accurately triggered at green nodes, confirming supervisor functionality.
• Flight data showed stable navigation along girders and columns, with LiDAR-based detection enabling precise relative positioning.
• The approach guarantees full coverage of the bridge surface without requiring prior high-accuracy global maps or GPS.
• The system is extendable to other bridge types, with four tested routines (GL, GR, CU, CD) forming a foundation for eight total routines.