From Laboratory to Field: Benchmarking Motion Capture Technologies for Offshore Wind Turbine Maintenance Ergonomics

Abstract

Offshore wind turbine maintenance poses significant ergonomic challenges due to physically demanding tasks performed in dynamic environments. This study benchmarks three motion capture technologies—marker-based, flexible sensor-based, and phone video-based systems—for evaluating biomechanical demands in simulated maintenance activities. Sagittal plane lumbar flexion angles derived from sensor- and video-based systems closely match those from the marker-based system in simpler tasks, with root mean square error (RMSE) values under 7° and high correlation coefficients. However, accuracy declined in complex or multi-plane motions, particularly for axial rotations. These findings support the use of sensor- and video-based systems as portable, costeffective alternatives to marker-based motion capture for field applications, while underscoring the need for cautious interpretation for dynamic or asymmetric movements.

Introduction

The growing demand for renewable energy has positioned offshore wind turbines as a critical component of the global energy transition. However, the maintenance of these structures presents significant challenges due to the harsh and dynamic environments in which they operate. Maintenance tasks often require workers to perform physically demanding activities, such as climbing, lifting, and operating heavy machinery, which can lead to musculoskeletal injuries if proper ergonomic practices are not followed. As indicated by a previous study (Oestergaard et al., 2022), lower back disorders should be a primary focus for monitoring, as they affect work at twice the rate of the second most impacted body region. Understanding and analyzing the biomechanics of these tasks is essential for preventing work-related injury and optimizing task performance. A critical first step is, however, to capture the motions involved in performing these tasks.

Traditional motion capture technologies, such as optical marker-based systems, are impractical for field use. Emerging alternatives, such as flexible sensor-based systems (Yin et al., 2023) and phone video-based methods (Uhlrich et al., 2023; Wang et al., 2025), offer promising solutions for real-world applications, but their feasibility for and performance in challenging offshore environments have not been evaluated. This study aims to bridge this gap by benchmarking the performance of three motion capture systems—marker-based, flexible sensor-based, and video-based—in acquiring kinematic data for offshore wind turbine maintenance tasks.

Background

Marker-based analysis continues to serve as the ‘gold standard’ for in-depth biomechanical analysis in laboratory settings, but its limitations in field applications have driven the development of alternative technologies (Chen et al., 2025). A Flexible sensor-based system, which integrates an accelerometer, a gyroscope, and a biopotential measurement unit, has shown potential for capturing movement in complex environments (Yin et al., 2023), while phone video-based methods leverage advances in computer vision to provide accessible and cost-effective motion analysis (Uhlrich et al., 2023). However, the accuracy, reliability, and practicality of these systems in the context of offshore maintenance tasks remain unknown. By simulating such tasks in the laboratory and by quantifying the offsets and correlations between kinematics measured by these motion capture technologies, this study also seeks to pave the way for laboratory-to-field translation in future applications.

Approach

Three healthy male participants (age: 30.7 ± 1.9 years, weight: 86.9 ± 10.9 kg, height: 1.76 ± 0.06 m) were recruited for this study. The experiment protocol was approved by the Institutional Review Board of Texas A&M University, and written informed consent was obtained from all participants prior to their involvement.

Participants were instructed to perform three representative occupational tasks (Milligan et al., 2024; Milligan et al., 2019). Task 1: Symmetric lifting; Task 2: Stepping onto and down a ladder; Task 3: Climbing up and down three rungs. Each task was repeated three times.

Thirteen BioStamp nPoint sensors were strategically placed on the subject to capture kinematic data from multiple body segments (Figure 1). Building upon the previous eleven-sensor model (Chen et al., 2024), two additional sensors were positioned on the foot to record ankle motion. Forty reflective markers (Figure 1) were attached to anatomical landmarks for motion tracking. Additionally, three iOS devices (two iPhones and one iPad) were positioned to ensure minimal obstruction of body segments by the ladder.

Figure 1.

Sensor and reflective marker placement. Pictures were created by adapting images from MC10 Inc., MA, USA.

To calibrate and synchronize the flexible sensor-based system, one static trial and one synchronization trial were conducted for each participant. Both marker-based and sensor-based motion data were recorded to ensure precise alignment and synchronization. To synchronize the video-based motion data with marker-based motion, another synchronization task (raising the right hand above the head) was adopted based on the OpenCap recommendation.

Outcome

The sensor-based and video-based methods demonstrated strong agreement with the marker-based system when measuring sagittal plane lumbar flexion angles. As shown in Figure 2 and Table 1, both systems closely mirrored the marker-based system, especially during simpler tasks where the amplitude of motion was more substantial. However, this agreement diminished as non-sagittal plane motion (e.g., the lumbar axial rotation) or asymmetry increased and when measuring the axial rotations themselves (Figure 3). Surprisingly, the video-based system performed better in measuring lumbar axial rotation during task (3) where the sensor-based system failed to capture its substantial variation (Figure 3). These results highlight the potential of sensor- and video-based systems for capturing sagittal plane kinematics yet advise caution when extending their use to more complex movements or multi-plane analyses.

Figure 2.

Lumbar flexion angle profiles of a representative participant (subject 02) during tasks (1), (2), and (3) derived by marker-based, sensor-based, and video-based methods. The shaded areas are ± one standard deviation.

Table 1.

The Grand Mean (Standard Deviation) of RMSE and Pearson Correlation Coefficient r Between Sensor-Based or Video-Based System and Marker-Based System for Lumbar Flexion Angle Across All Subjects.

Task	System,	RMSE (°)	Coefficient r
(1)	Sensor	6.17 (2.91)	0.976 (0.014)
	Video	5.92 (2.97)	0.990 (0.005)
(2)	Sensor	10.59 (4.3)	0.631 (0.247)
	Video	5.08 (2.05)	0.616 (0.157)
(3)	Sensor	11.71 (4.75)	0.443 (0.285)
	Video	6.25 (2.27)	0.564 (0.106)

Figure 3.

Lumbar axial rotation profiles of a representative participant (subject 02) during tasks (1), (2), and (3) derived by marker-based, sensor-based, and video-based methods. The shaded areas are ± one standard deviation.

Conclusion

This study benchmarks the performance of three motion capture technologies—marker-based, flexible sensor-based, and phone video-based systems—for acquiring kinematic data during simulated offshore wind turbine maintenance tasks. The findings demonstrate that both sensor-based and video-based systems exhibit strong agreement with the marker-based system in measuring sagittal plane lumbar flexion angles. However, the agreement decreases in more dynamic or multi-plane movements, emphasizing the need for careful consideration of task complexity and motion characteristics when selecting a motion capture system for biomechanical analysis. These results highlight the promise of sensor- and video-based systems as cost-effective, portable alternatives to traditional marker-based motion capture, though further technical innovation and validation are warranted to enhance their reliability in real-world applications.