Fig 1.
(a) Autonomous mobile robot (AMR) in experiment. (b) Illustration of environment. Participant moves from starting point toward goal. Ratio of distances used in this panel differs from actual ratio used in study.
Fig 2.
Walking velocity and distance to AMR.
(a) Average velocity profile (thick) and that for each participant (thin). Horizontal axis represents normalized time between beginning and end of trials. Vertical axis represents velocity. (b) Average distance to AMR. Other formats are same as those for (a).
Fig 3.
Human collision-avoidance behavior.
(a) Walking trajectory. Horizontal and vertical axes represent mediolateral (ML) and anteroposterior (AP) displacements, respectively. Lines represent walking trajectories (red: rightward avoidance, blue: leftward avoidance) on average (thick) and for each participant (thin). Black cross represents location of theoretical point of collision (TPC). (b) Histogram of estimated starting angles at which participants showed no consistent side preference when avoiding the AMR. The horizontal axis represents the starting angle; negative values indicate rightward avoidance. The number of participants with rightward and leftward avoidance is six (red) and nine (blue), respectively. The red vertical line represents the average estimated starting angle (–3.47°). The image in the top-right corner illustrates the fitted individual psychometric curves used to estimate these angles. (c) Average probability of rightward avoidance (vertical axis) at initial AMR positions (horizontal axis). Error bars represent standard error of mean. (d) Probabilities of rightward avoidance for each participant (horizontal axis) in head-on direction. Vertical axis is same as in (c). Green horizontal line represents average (same as 0° in (c)).
Fig 4.
(a) Changes in rotation angle and ML displacement at waist. Horizontal axis indicates normalized time between beginning and end of trials. Vertical axes indicate waist rotation angle (top) and ML displacement (bottom). Format is same as in Fig 3a. (b) Comparison of onset times between rotation angle and ML displacement in leftward avoidance case. Error bars represent standard error of mean. (c) This panel is same as (b) but for different avoidance direction. Asterisks indicate statistically significant differences in t-test; *** p < 0.005.
Fig 5.
(a) Illustration of predicting avoidance direction. (b) Different timings to detect waist rotation angle. (c) Two questionnaires after each trial.
Fig 6.
(a) Average ratings of timing for avoidance among each condition. Horizontal axis indicates each condition. Left vertical axis indicates speed from very fast to very slow. Right vertical axis indicates probability of “no avoidance” responses. Error bars represent standard error of mean. (b) Average ratings of comfort for trial. Format was same as (a) except left vertical axis indicates comfort level from very uncomfortable to very comfortable.
Fig 7.
(a) Autonomous mobile robot (AMR) in Experiment 3. (b) Comparison of waist rotation angles obtained from tracker and camera. Horizontal axis indicates time (seconds), while left vertical axis indicates waist rotation angle (degrees). Orange line represents data from motion tracker, and blue line represents data from MediaPipe Pose.
Fig 8.
(a) Average ratings of timing for avoidance in each condition during tracker session. Horizontal axis indicates each condition. Left vertical axis indicates speed from very fast to very slow. Right vertical axis indicates probability of “no avoidance” responses. Error bars represent standard error of mean. (b) Average ratings of comfort for trial in tracker session. Horizontal axis indicates each condition. Left vertical axis indicates level of comfort from very uncomfortable to very comfortable. (c) Average ratings of timing for avoidance in each condition during camera session. (d) Average ratings of comfort for trial in camera session.