Auditory Localization and Assessment of Consequential Robot Sounds: A Multi-Method Study in Virtual Reality

TL;DR

This study investigates how consequential robot sounds influence human auditory localization and subjective evaluations in a VR setting, comparing three locomotion types (quadruped Go1, wheeled Turtlebot 2i, wheeled HSR) with and without AVAS across two speeds and two trajectories. Using a within-subjects design, participants localize moving robot sounds over 240 trials and then rate annoyance, valence, and trust for each sound, enabling analysis of both objective localization performance and subjective perception. Results show that robot type markedly affects localization accuracy and precision, with the HSR generally hardest to localize, while AVAS effects are inconsistent; intriguingly, participants rate the HSR sound as more positive and trustworthy despite poorer localization. The findings highlight a tension between subjective sound evaluations and objective perceptual performance, suggesting that hubris about perceived sound pleasantness may not translate to safer or more intuitive spatial awareness in HRI. These insights inform robot sound design and social navigation strategies, emphasizing the need to consider both perceptual effectiveness and user experience when integrating robots into shared spaces.

Abstract

Mobile robots increasingly operate alongside humans but are often out of sight, so that humans need to rely on the sounds of the robots to recognize their presence. For successful human-robot interaction (HRI), it is therefore crucial to understand how humans perceive robots by their consequential sounds, i.e., operating noise. Prior research suggests that the sound of a quadruped Go1 is more detectable than that of a wheeled Turtlebot. This study builds on this and examines the human ability to localize consequential sounds of three robots (quadruped Go1, wheeled Turtlebot 2i, wheeled HSR) in Virtual Reality. In a within-subjects design, we assessed participants' localization performance for the robots with and without an acoustic vehicle alerting system (AVAS) for two velocities (0.3, 0.8 m/s) and two trajectories (head-on, radial). In each trial, participants were presented with the sound of a moving robot for 3~s and were tasked to point at its final position (localization task). Localization errors were measured as the absolute angular difference between the participants' estimated and the actual robot position. Results showed that the robot type significantly influenced the localization accuracy and precision, with the sound of the wheeled HSR (especially without AVAS) performing worst under all experimental conditions. Surprisingly, participants rated the HSR sound as more positive, less annoying, and more trustworthy than the Turtlebot and Go1 sound. This reveals a tension between subjective evaluation and objective auditory localization performance. Our findings highlight consequential robot sounds as a critical factor for designing intuitive and effective HRI, with implications for human-centered robot design and social navigation.

Figures (6)

• Figure 1: This study examines the impact of consequential sound of three mobile robots on human localization and subjective evaluation. It indicates a tension between objective and subjective measures but consistently highlights consequential robot sounds as key factor for designing a harmonious and effective HRI.
• Figure 2: Schematic illustration of experimental study. a) Participant localizes an invisible robot by its sound in VR and subsequently evaluates the different robot sounds. b) Top-down view on audiovisual VR scene. The invisible robot moves towards (head-on trajectory) or around the participant (radial trajectory). For both trajectories, the robot stops at a constant distance of 7 m from the participant and the participant estimates the final pose of the robot. The angular position of the robot was randomized within a span of -90$^\circ$ and +90$^\circ$ (participants’ default orientation was 0$^\circ$). The absolute angular deviation of the estimated from the actual robot pose represents the localization error ($\delta$) in each trial. c) Participants start each trial by pulling the trigger on the controller. The sound of the moving robot is presented for 3.0 s. After the sound has stopped, participants indicate where they last localized the robot and proceed with the next trial. d) Fully-crossed combinations of 4 independent variables (IVs): trajectory, velocity, AVAS and robot (a total of 24 experimental conditions for the localization task, each was repeated 10 times per participant).
• Figure 3: Normalized sound pressure spectrograms of the wheeled HSR (top left), wheeled Turtlebot 2i (top right), quadruped Go1 (bottom left), and isolated AVAS signal (bottom right). In the study, the AVAS signal was combined with the three robot sounds.
• Figure 4: Localization performance for head-on trajectories as a function of velocity, AVAS, and robot. a) Localization accuracy as measured by the mean absolute localization error (blue). b) Localization precision as measured by the intra-individual standard deviation (SD) of the absolute localization error (green). Error bars indicate ±1 $SE$ of the 24 individual values. The descriptive pattern consistently indicates a poorer performance for the HSR (particularly without AVAS) compared to the Turtlebot and Go1.
• Figure 5: Localization performance for radial trajectories as a function of velocity, AVAS, and robot. a) Localization accuracy as measured by the mean absolute localization error (blue). b) Localization precision as measured by the intra-individual standard deviation of the absolute localization error (green). Error bars indicate ±1 $SE$ of the 24 individual values. The descriptive pattern consistently indicates a poorer performance for the HSR than for the Turtlebot and Go1, as well as for the higher than for the lower velocity.