In this case study, we focus on the autonomous mobile robot (AMR), which can navigate and perform tasks in active environments without direct human control. Safe navigation and obstacle avoidance are fundamentally important for AMRs in the workplace.
Case Study Evaluation of an Auditory Alert for an Autonomous Mobile Robot
Justin Haney, Marvin Cheng, and Emily J. Haas | Centers for Disease Control and Prevention (CDC)
The expanded use of mobile robots in the manufacturing and warehousing industries is reshaping how certain tasks are performed in the workplace. A critical question that may benefit from detailed investigations is what are the potential hazards that might emerge when robots collaborate with human workers to complete tasks? In this case study, we focus on the autonomous mobile robot (AMR), which can navigate and perform tasks in active environments without direct human control. Safe navigation and obstacle avoidance are fundamentally important for AMRs in the workplace [1]. Consequently, predicting an AMR’s movement is critical for preventing collisions.
Research to date has shown that an AMR’s movement is more legible when the robot projects an arrow on the floor [2,3]. While some studies have shown that turn signals are capable of conveying movement intention and improving perceived comfort [4,5], other studies have shown such methods might not always be effective [6] and require the visual attention of nearby workers. Based on the inconsistent findings around the predictability of AMR movement, more research would be beneficial to support developing an effective hearing, or auditory, alert for conveying an AMR’s movement intention and the impact of that alert on human workers’ perceptions and behaviors. To this end, the objective of this case study was to investigate the use of an auditory alert for an AMR on perceived safety, mental workload, workers’ trust of the robot, and human behaviors. This study was reviewed and approved by the Centers for Disease Control and Prevention (CDC) National Institute for Occupational Safety and Health (NIOSH) Institutional Review Board (§ See 45 C.F.R. part 46; 21 C.F.R. part 56).
CASE STUDY DESIGN
We recruited nine adults (four males and five females; mean age: 22.0 ± 2.1 years) with experience working in manufacturing, warehousing, or in a stockroom (mean: 2.0 ± 0.7 years) to participate in the three-hour study. The study occurred at the NIOSH Division of Safety Research robotics laboratory located in Morgantown, WV. The laboratory is equipped with a Vicon motion capture system, which measured participant movement at a rate of 120 Hz, or hertz, which is movements per second.
To simulate human-robot interaction in a typical manufacturing or warehouse work environment, we used the AMR Freight100 (Zebra Technologies) [7] to interact with participants during requested tasks. We attached a Bluetooth speaker to produce an auditory alert signaling the AMR’s movement intention. The alert consisted of three pulses at a frequency of 1000 Hz (i.e., three sharp, high-pitched tones played one after another). Participants completed a set of tasks that involved transferring two boxes, one at a time, from a pick-up cart to a receiving cart, located 5.5 m apart.
The AMR started at a waypoint next to the receiving cart. First, we conducted a baseline trial where the AMR remained stationary. For the remaining eight trials, the AMR moved toward a second waypoint once the participant was approximately 2.7 m away from the receiving cart while they were delivering the second box. The trial was completed once the second box was placed on the receiving cart. The independent variables used in the study design were:
- AMR Height:
- Short: 0.36 m (AMR base with no cart attachment)
- Tall: 1.32 m (AMR base with cart attachment)
- Movement Path:
- Parallel
- Perpendicular
- Movement Communication:
- Alert on
- Alert off
The AMR moved either parallel or perpendicular to the participant’s movement between two waypoints. When the auditory alert was on, it was initiated once the participant picked up the second box from the pickup cart. Participants completed one baseline trial and eight experiment trials (2 heights x 2 movement directions x 2 movement communications). The order of the conditions was randomized.
The dependent variables included self-reported measures, task time, and behavioral measures. Specifically:
- Perceived Safety [8]: Evaluated by ratings from 1 (Never) to 6 (Very Much) for surprise and fear.
- Mental Workload [9]: Measured using the NASA Task Load Index (NASA-TLX), which consists of 6 subscales each rated on a 1 to 20 scale with 1 being “low” and 20 being “high.” The six subscales are combined using a weighting system to achieve an overall TLX score ranging from 0 to 100. A higher score indicates a higher perceived workload [9]:
- Mental demand
- Physical demand
- Temporal demand (i.e., time pressure)
- Performance
- Effort
- Frustration
- Robot Trust [10]: Corresponded to the average ratings (0-100%) of 12 items.
- Task Time: Refers to the duration of the second box transfer phase.
- Behavioral Measures:
- Average walking speed
- Maximum walking speed
- Maximum deviation from the straight-line path during the second transfer phase.
Outcome
Data analysis involved mixed effects models in IBM’s Statistical Package for the Social Sciences (SPSS) software application. Covariates included age, sex, trial number, and years of experience. Given the small sample size, we only report results for the aggregate sample. The supplementary Table 1 shows a summary of the p-value results for the mixed effects models for the self-reported measures. There was no significant difference in ratings for surprise across the independent variables. Briefly, key results included:
- Perceived Fear Significantly Increased in Certain Scenarios
- Fear was significantly higher when:
- A cart was present (1.71) vs. baseline (1.07; p = 0.010)
- No cart was present (1.38) vs. baseline (1.07; p = 0.014)
- AMR movement was parallel to the participant (1.69) vs. baseline (1.07; p = 0.014) and vs. perpendicular (1.41; p = 0.047)
- Additionally, alerts tended to slightly increase fear compared to baseline and no alert conditions.
- Fear was significantly higher when:
- Mental Workload Significantly Increased (NASA-TLX)
- Overall perceived workload (measured via the 100-point NASA-TLX) was significantly higher with alerts (9.99) than baseline (5.72; p = 0.008), but there was no difference between the alert and no alert conditions.
- Using the 20-point temporal demand subscale, perceived time pressures to complete tasks significantly rose with alerts (13.45) vs. baseline (7.40; p = 0.025) and vs. no alert (9.71; p = 0.007)
- Frustration levels were significantly higher with alerts (9.06) when compared to baseline (1.42; p = 0.004)
- Trust in Robot Significantly Increased
- Total trust rose from 75.31 (no alert) to 84.10 (alert; p < 0.001)
- Alert use improved trust ratings for (see supplementary Table 2 for more detailed results):
- Predictability (p = 0.005)
- Dependability (p = 0.008)
- Communicating intentions (p < 0.001)
- Providing useful feedback (p = 0.004)
- Giving appropriate information (p < 0.001)
- Task Time and Movement Behavior
- Task completion time: No significant change across conditions
- Average walking speed: Decreased with alert (873 mm/s) vs. no alert (905 mm/s; p = 0.011)
- Max speed: Dropped with alert (1351 mm/s) vs. baseline (1422 mm/s; p = 0.029)
- Path deviation: Tended to increase with alert (265 mm) vs. no alert (225 mm; p = 0.095)
Summary and future work
This case study sought to extend previous research about safety and trust during human-robot interaction with AMRs and inform future research [11]. We evaluated the ability and impact of an AMR auditory alert to warn nearby workers of the robot’s movement intentions. First, AMR alerts may help workers trust robots more by making their movements more predictable. However, this finding should be balanced with other results showing that alerts can also increase overall mental workload, temporal demand, and worker frustration. These findings agree with prior research which shows that warnings can increase workload and temporal demand while driving [12]. To this end, sound design matters when identifying auditory alerts for AMRs. It is possible that some tones may increase stress or workload more than others, identifying the importance of further exploring in future studies.
Generally, participants reported the AMR’s movement to be more predictable and dependable when an auditory alert was active. Specifically, our findings revealed that workers may adjust their movements and behaviors upon hearing an alert. This action may improve safety behavior but also impact movement speed (i.e., workflow), as shown in our results. These findings agree with past research investigating the use of turn signals or light projectors on the legibility of an AMR’s movement intention and the comfort of nearby workers [2-5].
These findings provide preliminary evidence that an auditory alert can improve the overall safety of an AMR in the workplace while being aware of some limitations. Although this case study was limited by a small sample size, results offer a snapshot into how basic auditory alerts may influence worker behaviors, perceptions, and experiences with robots in the workplace. Future research may seek to identify if trends become significant with a larger sample size. Additionally, future research may also consider other background workplace noise when auditory alerts may be used. To this end, future research in this area may try to refine the types of auditory alerts delivered such as varying tones or voice-based cues and their impact on worker trajectories and speed profiles. In conclusion, this case study illustrates both the technical and human challenges to consider when integrating robots into the workplace. Our findings highlight the importance of intentional design choices that are grounded in empirical, behavioral evidence to support safe human worker-robot collaboration.
Disclaimer: The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the National Institute for Occupational Safety and Health, Centers for Disease Control and Prevention. Mention of any company or product does not constitute endorsement by NIOSH, CDC.
Justin Haney, Biomedical Engineer, NIOSH Division of Safety Research
Marvin Cheng, Team Leader, Mechanical Engineer, NIOSH Division of Safety Research
Emily J. Haas, Associate Director for Science, NIOSH Division of Safety Research
References
- Pandey, A., Pandey, S., & Parhi, D. (2017). Mobile robot navigation and obstacle avoidance techniques: A review. International Robotics & Automation Journal, 2(3), 96–105. https://doi.org/10.15406/iratj.2017.02.00023.
- Shrestha, M. C., Onishi, T., Kobayashi, A., Kamezaki, M., & Sugano, S. (2018). Communicating directional intent in robot navigation using projection indicators. In 2018 27th IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN) (pp. 746–751). IEEE. https://doi.org/10.1109/ROMAN.2018.8525528.
- Matsumaru, T. (2006). Mobile robot with preliminary-announcement and display function of forthcoming motion using projection equipment. ROMAN 2006 - The 15th IEEE International Symposium on Robot and Human Interactive Communication (pp. 443–450). IEEE. https://doi.org/10.1109/ROMAN.2006.314368.
- Shrestha, M. C., Kobayashi, A., Onishi, T., Uno, E., Yanagawa, H., Yokoyama, Y., Kamezaki, M., Schmitz, A., & Sugano, S. (2016). Intent communication in navigation through the use of light and screen indicators. In 2016 11th ACM/IEEE International Conference on Human-Robot Interaction (HRI) (pp. 523–524). IEEE. https://doi.org/10.1109/HRI.2016.7451837
- May, A. D., Dondrup, C., & Hanheide, M. (2015). Show me your moves! Conveying navigation intention of a mobile robot to humans. In 2015 European Conference on Mobile Robots (ECMR) (pp. 1–6). IEEE. https://doi.org/10.1109/ECMR.2015.7324049.
- Fernandez, R., John, N., Kirmani, S., Hart, J., Sinapov, J., & Stone, P. (2018). Passive demonstrations of light-based robot signals for improved human interpretability. In 2018 27th IEEE International Symposium on Robot and Human Interactive Communication (RO-MAN) (pp. 234–239). IEEE. https://doi.org/10.1109/ROMAN.2018.8525728.
- Zebra Technologies (2023). Create the new world of robotics automation. https://www.zebra.com/content/dam/archive_zebra_dam/en_gb/brief/application/oem-brief-application-robotics-automation-en-gb.pdf.
- Nonaka, S., Inoue, K., Arai, T., & Mae, Y. (2004, April). Evaluation of human sense of security for coexisting robots using virtual reality. 1st report: evaluation of pick and place motion of humanoid robots. In IEEE International Conference on Robotics and Automation, 2004. Proceedings. ICRA'04. 2004 (Vol. 3, pp. 2770-2775). IEEE. https://doi.org/10.1109/ROBOT.2004.1307480.
- Hart, S. G., & Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research. Advances in Psychology, 52(C), 139–183. https://doi.org/10.1016/S0166-4115(08)62386-9.
- Schaefer, K. (2013). The perception and measurement of human-robot trust. Dissertation. Orlando, FL: University of Central Florida.
- Haney, J. M., & Liang, C. J. (2024). A Literature Review on Safety Perception and Trust during Human–Robot Interaction with Autonomous Mobile Robots That Apply to Industrial Environments. IISE Transactions on Occupational Ergonomics and Human Factors, 12(1–2), 6–27. https://doi.org/10.1080/24725838.2023.2283537.
- Koniakowsky, I. M., Hamm, S. J., Forster, Y., & Keinath, A. (2024, September). From Warnings to Workload: The Impact of Driver Monitoring System Warnings on Drivers' Self-Reported Workload. In Adjunct Proceedings of the 16th International Conference on Automotive User Interfaces and Interactive Vehicular Applications (pp. 68-71). https://doi.org/10.1145/3641308.3685025.
The content & opinions in this article are the author’s and do not necessarily represent the views of RoboticsTomorrow
Comments (0)
This post does not have any comments. Be the first to leave a comment below.
Featured Product
