Abstract
Teleoperation is increasingly recognized as a viable solution for deploying robots in hazardous environments. However, controlling high-degree-of-freedom (DOF) robots can pose significant challenges for teleoperators. To design an assistive robot controller, it is essential to gain a comprehensive understanding of the intricate interplay between the robot's autonomous behavior and the operator's internal state. Previous research has investigated cognitive load as a metric for quantifying user demand, engagement, and perception during these interactions. Additionally, other studies have explored trust within shared-autonomy systems. While there is recognition of these factors' influence on human performance, their specific effects and the dynamic interrelationship between cognitive load, trust, and the autonomy level of the robot remain unclear. Our results show that autonomy level influences the teleoperator's perceived cognitive load and trust. However, we establish that there is no interaction between these factors. This suggests that the relationship between autonomy, cognitive load, and trust is not directly dependent on one affecting the other. Instead, these elements appear to operate independently, highlighting the need for a nuanced approach to designing assistive robotic systems that consider both cognitive load and trust as distinct but interrelated factors. This insight is crucial for the development of more effective and adaptable robot controllers in teleoperation scenarios.
Novint Falcon haptic controller (3-DOF)
Franka Emika Research 3 Robotic Arm (7-DOF)
Virtual Rendering + Point Cloud Visualization (Rviz)
Tobii Pro X2 Screen-Based Eye-Tracker
Logitech Keyboard
Cognitive Load
The NASA-TLX questionnaire has been widely used to capture cognitive load information, including in HRI and robot teleoperation settings
Pupil Dilation has been associated with more intense cognitive processing
Dual-Task Method: The secondary task consists of the "Rhythm Method" designed to assess and measure the participant's cognitive load. This method involves the participant tapping a pre-recorded rhythm from memory while concurrently performing a primary task.
The MDMT questionnaire is composed of 8 sub-scales in each of the Capacity Trust and Moral Trust categories, and has been widely employed in HRI studies to capture trust information.
Single-Scale Trust question: "How much do you trust the robot?" reported using a 10-point discrete scale.
The primary task required the participant to teleoperate the robot to track a trajectory of 10 seconds duration using the robot's end-effector. Participants were asked to "as accurately as possible" follow the 3-dimensional time-parameterized trajectory.
The pattern consisted of a short inter-tap interval of duration t, followed by a long inter-tap interval of duration 3t. Hence, the speed of the rhythms were inversely related to and directly determined by the duration of the short inter-tap interval, t. Participants followed the rhythm by tapping "space" on a keyboard.
Each session had two phases: Training and Main.
Within each phase, participants completed multiple different rounds - 3 rounds in the training phase and 2 rounds in the main phase - where each round contained 5 trials.
During individual trials, the task performance and pupil diameter data were continuously recorded, while participants' self-reported their cognitive load and trust after each round.
Results
We performed two-way mixed ANOVAs with autonomy as the within-subject factor and order as the between-subject factor for each measure. Results were considered significant at the threshold α < 0.05.
effect -> NASA-TLX
effect -> MDMT
no effect -> Pupil Diameter
no effect -> Tapping Error
We conducted a Pearson correlation analysis between MDMT score and data from the Single-Scale Trust question, in which the two measures were found to be strongly positively correlated (r(46) = 0.873, p < .001).
With high robot autonomy, there existed a trend where participants reported lower trust as they experienced higher cognitive load.
With low robot autonomy, trust remained approximately constant across increasing cognitive load levels.