Autonomous Inspection of Power Line Insulators with UAV on an Unmapped Transmission Tower

Václav Riss; Vít Krátký; Robert Pěnička; Martin Saska

cs.ROFeb 27, 2026

Autonomous Inspection of Power Line Insulators with UAV on an Unmapped Transmission Tower

Authors: Václav Riss, Vít Krátký, Robert Pěnička, Martin Saska

Abstract

This paper introduces an online inspection algorithm that enables an autonomous UAV to fly around a transmission tower and obtain detailed inspection images without a prior map of the tower. Our algorithm relies on camera-LiDAR sensor fusion for online detection and localization of insulators. In particular, the algorithm is based on insulator detection using a convolutional neural network, projection of LiDAR points onto the image, and filtering them using the bounding boxes. The detection pipeline is coupled with several proposed insulator localization methods based on DBSCAN, RANSAC, and PCA algorithms. The performance of the proposed online inspection algorithm and camera-LiDAR sensor fusion pipeline is demonstrated through simulation and real-world flights. In simulation, we showed that our single-flight inspection strategy can save up to 24 % of total inspection time, compared to the two-flight strategy of scanning the tower and afterwards visiting the inspection waypoints in the optimal way. In a real-world experiment, the best performing proposed method achieves a mean horizontal and vertical localization error for the insulator of 0.16 +- 0.08 m and 0.16 +- 0.11 m, respectively. Compared to the most relevant approach, the proposed method achieves more than an order of magnitude lower variance in horizontal insulator localization error.

Figures & tables

Fig. 1: Deployment of the proposed pipeline for autonomous inspection of insulators onboard a UAV in a real-world scenario.

Panel of Figure 2 — Fig. 2: Illustration of the insulator detection and localization algorithmic pipeline. Points inside the bounding box marked in non-red color are clustered and utilized for insulator position and orientation estimation.

Panel of Figure 3 — Fig. 3: Simulated towers in Flight Forge simulator [ 22 ] , (a) shows tower A with twelve horizontal insulators, and (b) shows tower B with four vertical insulators.

Panel of Figure 4 — Fig. 4: A scheme representing the LiDAR frame $L$ , the body frame $B$ , and the camera frame $C$ . Fixed transformations between the frames allow projection of points ${}^{L}{\mathbf{p}}_{in}$ and ${}^{L}{\mathbf{p}}_{out}$ the image plane frame defined by I. Projected points ${\mathbf{s}}_{in}$ and ${\mathbf{s}}_{out}$ are then filtered out if they do not lie inside a green bounding box corresponding to detected insulator.

Fig. 5: Illustrative scheme of the proposed state machine for inspection of insulators. Whenever an insulator is detected, inspection waypoints are saved into the buffer

B

. State machine empties

B

by visiting the stored inspection waypoints.

Method	$w=8$ [m]	$w=9$ [m]	$w=10$ [m]
RANSAC	0.40 $\pm$ 0.20	0.40 $\pm$ 0.24	0.43 $\pm$ 0.30
DBSCAN + RANSAC	0.35 $\pm$ 0.17	0.37 $\pm$ 0.23	0.38 $\pm$ 0.22
DBSCAN	0.74 $\pm$ 0.57	0.66 $\pm$ 0.47	0.50 $\pm$ 0.33
DBSCAN + PCA	0.79 $\pm$ 0.65	0.60 $\pm$ 0.57	0.40 $\pm$ 0.35

TABLE I: Insulator localization error for a changing distance from insulator

w

on tower A. The statistics are computed based on 150 measurements.

Method	$w=8$ [m]	$w=9$ [m]
RANSAC	0.39 $\pm$ 0.24	0.43 $\pm$ 0.21
DBSCAN + RANSAC	0.34 $\pm$ 0.20	0.36 $\pm$ 0.23
DBSCAN	0.32 $\pm$ 0.21	0.41 $\pm$ 0.25
DBSCAN + PCA	0.33 $\pm$ 0.21	0.41 $\pm$ 0.26

TABLE II: Insulator localization error for a changing distance from insulator

w

on tower B. The statistics are computed based on 150 measurements.

Fig. 6: Trajectory of the UAV guided by our online inspection algorithm in Flight Forge simulator [ 22 ] computed by PMM planner [ 26 ] . Green points are the inspection waypoints suitable for obtaining detailed images of insulators. The starting position is depicted as a blue point.

Panel of Figure 9 — Fig. 7: Comparison of the duration $T_{fusion}$ achieved by our approach with the duration $T_{tsp}$ as optimal solution of TSP. The combined $T_{tsp}$ plus $T_{scan}$ of scanning the tower is also shown and included in the comparison. (a) depicts durations for inspection of tower A, (b) depicts durations for inspection of tower B.

Horizontal constraints		Vertical constraints		Inspection parameters
$\mathbf{v}_{max}$ [ $\mathrm{ms^{-1}}$ ]	1	$\mathbf{v}_{max}$ [ $\mathrm{ms^{-1}}$ ]	1	$h_{max}$ [ $\mathrm{m}$ ]	25
$\mathbf{a}_{max}$ [ $\mathrm{ms^{-2}}$ ]	3	$\mathbf{a}_{max}$ [ $\mathrm{ms^{-2}}$ ]	3	$w_{max}$ [ $\mathrm{m}$ ]	10

TABLE III: Dynamic constraints and inspection parameters

Panel of Figure 11 — Fig. 8: The UAV guided by the online inspection algorithm with onboard sensor fusion in a real-world experiment. The blue FOV frustums illustrate the detection of each insulator in the picture (a). Each detection is also shown in the right column of images (b) - (d) together with the bounding box and the visualized filtered point cloud. The green markers in (a) represent the inspection waypoints used for image capturing and captured inspection images are shown in (e) - (j).

Panel of Figure 12 — Fig. 9: Real-world insulator localization errors for $xy$ plane in (a) and altitude $z$ in (b). Different colors correspond to different instances of an insulator.

Robotic Inspection

Robotic inspection uses autonomous or remotely operated robots to examine infrastructure, equipment, and environments, detect defects, and document condition.

See all 50 papers

A novel YOLO26-MoE optimized by an LLM agent for insulator fault detection considering UAV images
May 19, 2026João Pedro Matos-Carvalho, Laio Oriel Seman, Stefano Frizzo Stefenon +2Mixture of Experts Object Detection
LASIGE, Faculdade de Ciˆencias, Universidade de Lisboa, 1749-016 Lisboa, Portugal · Department of Automation and Systems Engineering, Federal University of Santa Catarina, Florian´opolis, Brazil · Instituto Superior de Engenharia de Lisboa, Instituto Polit´ecnico de Lisboa, Rua Conselheiro Em´ıdio Navarro 1, 1959-007 Lisboa, Portugal +1
Attention-Guided Autoencoder Fusion for Insulator Defect Detection Using UAV Transmission-Line Imaging
Jun 3, 2026Malak Allam, Khaled Shaban, Ali HamdiIndustrial Inspection Object Detection
Faculty of Computer Science MSA University Giza, Egypt · Department of Computer Science and Engineering Qatar University Doha, Qatar
Large-Scale Tunnel Air-Ground Collaboration With FLISP: Fast LiDAR-IMU Synchronized Path Planner
Jun 24, 2026Fenghe Guo, Runjie Shen, Chenyang Sun +4Multi-Robot Motion Planning Mobile Robotics
Shanghai Research Institute for Intelligent Autonomous Systems, Tongji University, Shanghai 201210, China · State Key Laboratory of Autonomous Intelligent Unmanned Systems, Tongji University, Shanghai 201210, China · College of Electronic and Information Engineering, Tongji University, Shanghai 201804, China

May 19, 2026J/K move · Enter open · S save

A novel YOLO26-MoE optimized by an LLM agent for insulator fault detection considering UAV images

João Pedro Matos-Carvalho, Laio Oriel Seman, Stefano Frizzo Stefenon, Mohammad Khalaf Mohammad Khreasat+1

LASIGE, Faculdade de Ciˆencias, Universidade de Lisboa, 1749-016 Lisboa, Portugal · Department of Automation and Systems Engineering, Federal University of Santa Catarina, Florian´opolis, Brazil · Instituto Superior de Engenharia de Lisboa, Instituto Polit´ecnico de Lisboa, Rua Conselheiro Em´ıdio Navarro 1, 1959-007 Lisboa, Portugal+1

Mixture of Experts Object Detection Aerial Object Detection Anomaly Detection

Open paper AskPDF arXiv

The inspection of electrical power line insulators is essential for ensuring grid reliability and preventing failures caused by damaged or degraded insulation components. In recent years, Unmanned Aerial Vehicles (UAVs) combined with deep learning-based vision systems have emerged as an effective solution for automating this process. However, insulator fault detection remains challenging due to small defect regions, heterogeneous fault patterns, complex backgrounds, and varying imaging conditions. To address these challenges, this paper proposes an optimized YOLO26-MoE, a novel object detection architecture that integrates a sparse Mixture-of-Experts (MoE) module into the high-resolution branch of the YOLO26 detector. The proposed modification enables adaptive feature refinement for subtle and diverse fault patterns while preserving the efficiency of a one-stage detection framework. Hyperparameter optimization, final training, and evaluation were coordinated through a tool-augmented Large Language Model (LLM) agent. The proposed model achieved 0.9900 [email protected] and 0.9515 [email protected]:0.95, outperforming the latest YOLO versions. These results demonstrate that the proposed model provides an effective and reliable solution for UAV-based insulator fault detection.

Abstract

Robotic Inspection

Explore similar work

A novel YOLO26-MoE optimized by an LLM agent for insulator fault detection considering UAV images

Attention-Guided Autoencoder Fusion for Insulator Defect Detection Using UAV Transmission-Line Imaging

Large-Scale Tunnel Air-Ground Collaboration With FLISP: Fast LiDAR-IMU Synchronized Path Planner