Unpredictable movements of autonomous robots increase human discomfort

＜Abstract＞

A research team from the Visual Perception and Cognition Laboratory and the Cognitive Neurotechnology Unit at Toyohashi University of Technology investigated how the movements of autonomous mobile robots influence human emotional responses during passing encounters in virtual reality (VR) environments. The study found that when the robot moved monotonously in a straight path, participants initially exhibited increased arousal and skin conductance responses, which gradually diminished over repeated trials—indicating habituation. In contrast, when the robot’s motion included unpredictable pauses and restarts, both arousal and skin conductance responses remained elevated throughout the experiment, suggesting a lack of habituation and a greater tendency for participants to experience discomfort. These findings highlight the importance of predictability in robot behavior as a critical factor affecting human comfort in shared spaces. The results of this study were published online in the International Journal of Social Robotics on January 19, 2026. https://doi.org/10.1007/s12369-025-01341-3

＜Main＞

In recent years, autonomous mobile robots designed for tasks such as food delivery, cleaning, and security have begun to enter everyday human environments. For these robots to be more widely accepted in human society, it is not sufficient to merely avoid physical collisions with people; their movements must also be designed to prevent feelings of discomfort or fear in human observers. The research team believes that incorporating human cognitive and emotional characteristics into robotic motion design can facilitate more comfortable and socially acceptable human–robot coexistence.

While previous studies have examined locomotion trajectories and avoidance patterns during human–robot passing scenarios, the emotional experiences of humans during such encounters have remained largely unexplored. To address this gap, the present study employed a virtual reality (VR) environment in which participants were asked to walk toward a goal while avoiding a robot approaching from the opposite direction. During the task, both subjective emotional evaluations (valence and arousal levels) and physiological indices of arousal—specifically, palm skin conductance responses (a measure known to correlate with sympathetic nervous system activity)—were recorded.

In the first experiment, participants encountered a robot that approached from one of three directions: straight ahead, from the left, or from the right. The results showed no significant differences in emotional or physiological responses based on the direction of approach. However, across repeated trials, participants’ arousal levels decreased—a habituation effect—suggesting that they became less tense over time. This is likely because the robot’s behavior was simple and predictable, consisting only of straight-line motion.

In the second experiment, behavioral uncertainty was introduced to manipulate the predictability of the robot’s movement. Specifically, the robot exhibited one of three behaviors: (1) a sudden stop followed by a restart, (2) no movement at all, or (3) continuous straight-line movement. The results revealed that when the robot performed the irregular behavior involving a stop-and-go motion, participants reported significantly higher levels of discomfort and showed increased arousal and skin conductance responses compared to the other two conditions.

Crucially, in contrast to the first experiment, no habituation was observed over repeated trials in the stop-and-go condition. This suggests that the unpredictability of the robot’s motion disrupted not only the current trial but also participants’ expectations in subsequent trials. These findings align with established psychological evidence that uncertain stimuli can elicit anxiety or aversive reactions. They underscore the importance of predictability in robotic behavior as a key factor for fostering a sense of safety and comfort in human–robot interactions.

Yuta Matsubara, the first author of this study and a recent graduate of the Master’s Program in Computer Science and Engineering as well as the Master of Science Program in Imaging and Light in Extended Reality (IMLEX), explains: “As robots become increasingly integrated into our daily lives, understanding how humans emotionally respond to them is essential. While there are still many challenges to be addressed before full-scale implementation in the real world, I believe that the insights gained from this study can contribute to the design of human–robot interactions that are comfortable and free of stress for both parties.”

＜Future Prospects＞

In this study, we conducted experiments using a simplified VR environment and a single autonomous robot. As a next step, we plan to extend our investigations to more realistic scenarios involving environmental complexities such as curved pathways and variable corridor widths, as well as interactions involving multiple robots or pedestrians. Through these future studies, we aim to establish design guidelines for robot behaviors that take human emotions into account, ultimately contributing to the creation of environments where humans and robots can coexist comfortably and without stress.

Figure: Robot model (left)．Equipment used (center)．Unpredictable movements are unpleasant and high-arousal (Right)

＜Acknowledgements＞

This study was based on the results obtained from the JPNP20004 project subsidized by the New Energy and Industrial Technology Development Organization (NEDO). This work was supported by the Nitto Foundation.

Matsubara, Y. ^†, Tamura, H. ^†* , Minami, T., & Nakauchi, S. (2026). Subjective Emotions and Physiological Responses During Collision Avoidance with a Virtual Autonomous Mobile Robot. International Journal of Social Robotics, 18(1), 5. https://doi.org/10.1007/s12369-025-01341-3

(*: Corresponding author, †: These authors contributed equally)