Enhancing Harbor Workers' Safety through Remote and Immersive Inspection with a Quadruped Robot

Enhancing Harbor Workers’ Safety through Remote and Immersive Inspection with a Quadruped Robot Ali Yousefi DIBRIS, RICE Lab University of Genova Genova, Italy ali.yous[email protected] Zoe Betta DIBRIS, RICE Lab University of Genova Genova, Italy zoe.bet[email protected]ige.it Davide Corongiu Autorità di Sistema Portuale del Mar Ligure Occidentale Genova, Italy [email protected] Carmine Tommaso Recchiuto DIBRIS, RICE Lab University of Genova Genova, Italy [email protected] Antonio Sgorbissa DIBRIS, RICE Lab University of Genova Genova, Italy [email protected] Abstract— The study tested an immersive teleoperation system for quadruped robots in harbor container parks, comparing a Head-Mounted Display (HMD) with 5G control to a standard gamepad and 2D display. Nine harbor workers performed hazard detection tasks while researchers measured success rate, time, workload, usability, and simulator sickness. Both setups showed low workload and good usability, with the immersive system offering slight improvements in performance. However, simulator sickness with the HMD was a drawback. Workers found the system relevant for real inspections. Overall, immersive teleoperation may improve safety by limiting human exposure to hazards, but larger studies and solutions for simulator sickness are needed before real-world deployment. Keywords— Teleoperation, Immersive Interfaces, Robotics in Hazardous Fields, Quadruped Robots, Human Robot Interaction. I. INTRODUCTION This study addresses workplace safety risks in harbors by examining the use of quadruped robots for remote inspections. While drones and wheeled robots offer certain advantages, quadrupeds are especially well-suited to hazardous container parks due to their stability, obstacle-navigation capabilities, and safer operation in close proximity to workers. Although standard remote inspection systems using gamepads and 2D displays already provide significant safety benefits by removing workers from hazardous areas, this study explores whether immersive teleoperation can offer additional advantages. To this end, the Spot robot was deployed in a harbor environment using an immersive teleoperation framework that combines an HMD with 5G connectivity. Operators control the robot’s movements via head tracking and handheld controllers, enabling remote inspection of simulated hazards such as unstable stacks, leaks, or blocked paths. Building on prior interviews with port operators [1], the research focuses on whether immersive teleoperation reduces workload, improves usability, avoids simulator sickness, and enhances hazard detection compared to a standard 2D display and gamepad interface. The contributions of this work are twofold: (1) the deployment of Spot for safe remote harbor inspections, and (2) a user study with harbor workers evaluating usability, workload, and hazard detection effectiveness. In terms of related work, robotic systems have been explored for infrastructure inspections in civil environments [2], but harbor operations remain underrepresented. Existing studies primarily address underwater inspections or aerial surveys, which are effective for large areas but limited for close-range, detailed inspections near workers. Quadruped robots have been tested in offshore platforms and construction monitoring, showing strong navigation capabilities and potential for immersive inspection [3]. However, most prior works lacked direct involvement of end-users, limiting their practical relevance. To date, only one study has surveyed port operators’ actual needs regarding quadruped robots [1]. The novelty of the present work lies in deploying a quadruped robot for immersive teleoperation specifically in harbor environments, streaming stereo video to an HMD, and conducting the first user study with professional port terminal operators to evaluate usability, workload, and hazard detection while ensuring system acceptance and real-world applicability. II. METHODOLOGY A. Immersive Teleoperation System The immersive teleoperation system (see Figure 1) is structured around dual-component architecture, comprising the Robot Side and the Operator Side, interconnected through a VPN tunnel over the internet. The robot is equipped with a ZED 2 stereo camera and an NVIDIA Jetson Nano, enabling high-resolution image acquisition, compression, and transmission over 5G network. Two parallel threads manage streaming and control operations: the Streaming Thread captures and transmits stereo imagery to the operator, while the Control Thread receives control inputs for both locomotion and head orientation tracking. The operator side features a PC connected to a Meta Quest 2 headset, with dedicated threads for retrieving video frames and sending control inputs, including IMU-based head orientation and thumbstick commands, to the robot. Real-time communication is maintained via a GStreamer pipeline for streaming and MQTT topics for control messages. Fig. 1. System architecture of the immersive teleoperation system. This work was carried out within the framework of the project “RAISE - Robotics and AI for Socio-economic Empowerment” and has been supported by European Union – NextGenerationEU. 2025 I-RIM Conference October 17-19, Rome, Italy ISBN: 9788894580570 10.5281/zenodo.17629664 75 Fig. 2. Snapshots from the experiments at the Port of Genoa. Left: Terminal worker remotely controlling the robot from an office, Middle and Right: Spot robot operating between containers. B. System Performance Network performance was evaluated using custom Python scripts, measuring metrics including glass-to-glass latency (~500 ms), round-trip time (~378 ms), throughput (~3.3 Mbps), jitter (~93 ms), frame rate (~15 Hz), and packet loss (~1.3%). It is expected that the performance metrics will not significantly affect the operator’s performance, in line with previous studies in the literature [4]. C. Experimental Procedure A preliminary user study was conducted with 𝑁 = 9 port operators to evaluate the effectiveness of an immersive teleoperation system compared to a standard 2D/gamepad interface for remote harbor inspections. The experiments took place in two designated inspection areas approximately 190 meters from the operator site. Eight printed shapes (triangles, squares, circles, and stars) were used as representative cues for simulated hazards, arranged in four configurations over two days to reflect the dynamic nature of the port environment. Following a training session, each participant performed an inspection task involving the detection of simulated hazards (shapes) under the following Conditions:  A: using the immersive system with an HMD and head-tracked control;  B: using a standard gamepad with a 2D display. The order of Conditions A and B and shape configurations (1-4) was randomized to minimize bias. Objective measures included the number of shapes identified and task completion time, combined into a “Success Rate” metric defined as shapes found per minute. Subjective measures involved the NASA Task Load Index (NASA-TLX) to assess workload across six dimensions (mental, physical, temporal demand, performance, effort, and frustration), the System Usability Scale (SUS) to evaluate overall usability, and the Simulator Sickness Questionnaire (SSQ) to measure nausea, oculomotor disturbance, and disorientation before and after HMD use. III. RESULTS AND CONCLUSIONS Figure 3 presents the measured results, showing bar plots that compare the average task success rate, completion time, and system usability, and a radar plot illustrating the average workload scores in both conditions. A. Workload A one-sample Wilcoxon signed-rank test and a onesample t-test indicated that the average workload score was “low” in Condition A (M = 32.96, SD = 23.09; Z = 2.31, p = .021) and also in Condition B (M = 35.37, SD = 24.08; t(8) = 3.44, p = .009). Condition A’s slightly lower score suggested a reduced workload. However, a paired samples t-test revealed that the difference between the two conditions was not statistically significant (t(8) = 0.43, p = .680). B. System Usability One-sample t-test results indicated that the average system usability score was “good” in Condition A (M = 78.89, SD = 16.96; t(8) = 5.11, p < .001) and in Condition B (M = 80.03, SD = 18.29; t(8) = 4.92, p < .001). However, a paired-samples t-test showed that the slightly lower score in Condition A was not statistically significant and therefore did not support higher usability for the immersive system (t(8) = 0.32, p = .760). C. Distress Symptoms One-sample t-test results indicated that the average distress symptoms score was “severe” after using the immersive system (M = 31.17, SD = 26.11; t(8) = 2.43, p = .041), underscoring a significant limitation of the HMD. D. Task Performance The results showed a higher mean task success rate in Condition A (M = 0.64, SD = 0.33) compared to Condition B (M = 0.61, SD = 0.25), and a lower mean task completion time in Condition A (M = 361.89 s, SD = 91.33) than in Condition B (M = 369.89 s, SD = 89.62). However, paired-samples ttests indicated that neither the difference in task success rate, t(8) = 0.19, p = .852, nor the difference in task completion time, t(8) = 0.16, p = .877, was statistically significant. E. Conclusions In conclusion, the immersive system showed a slight trend toward lower workload and improved task performance compared to the gamepad interface, though differences were not statistically significant. Both systems exhibited good usability, while severe simulator sickness in the HMD condition remains a major barrier. Future work will explore mitigation strategies for HMD-related discomfort, such as optimized control mappings, reduced high-velocity inputs, and potential shared-control modes, alongside larger participant studies to validate whether immersive teleoperation can meaningfully enhance hazard detection and safety in harbor operations. REFERENCES [1] Z. Betta, D. Corongiu, C. T. Recchiuto, and A. Sgorbissa, “Enhancing worker safety in harbors using quadruped robots,” in Proc. 22nd Int. Conf. Ubiquitous Robots (UR 2025), College Station, TX, USA, Jul. 2025, pp. 1–10. [2] D. Lattanzi and G. Miller, “Review of robotic infrastructure inspection systems,” in J. Infrastructure Syst., vol. 23, no. 3, p. 04017004, 2017. [3] M. Ramezani, M. Brandao, B. Casseau, I. Havoutis, and M. Fallon, “Legged robots for autonomous inspection and monitoring of offshore assets,” in Offshore Technology Conf., 2020, pp. D011S006R005. [4] M. Moniruzzaman, A. Rassau, D. Chai, and S. M. Islam, “Teleoperation methods and enhancement techniques for mobile robots: A comprehensive survey,” in Robot. Auton. Syst., vol. 150, p. 103973, Apr 2022. Fig. 3. Results of pairwise comparisons between the conditions. 76