Usability Evaluation of Human–Robot Interaction in Manipulator Teleoperation via Virtual Reality in Inspection Tasks
Kotaro Kanazawa, Noritaka Sato, Yoshifumi Morita · 2024
Currently, numerous studies have compared conventional interfaces in terms of operability, demonstrating that VR interfaces outperform traditional methods in several aspects. However, research on best practices in VR remains limited. In VR interfaces, some robots move synchronously in real-time with the input, while others respond only after the operator sets the target posture. Achieving perfect synchronization between the operator's movements and the robot's actions is challenging due to limitations in robot speed and communication delays. The timing of the robot's movements can significantly impact operability in VR interfaces. This study investigates the effect of the timing of robot movement initiation on usability. Validation was conducted using a VR teleoperation interface developed for rescue robots. This interface offers two modes of operation: Move-During-and-After and Move-After. The former synchronizes the robot movement with the operator's actions; in the latter, the robot remains stationary until the operator specifies the target posture, after which it moves to reach the specified posture. We focused on the manipulation of the position and posture of the manipulator. Moreover, we compared the usability differences of inspection tasks using these two methods via experiments involving 13 participants. Furthermore, we evaluated the usefulness of the VR interface by comparing it with gamepad- and display-based interface. The mean task completion time and standard deviation for the Move-During-and-After (MDA) operation were the smallest, followed by Move-After (MA), and then the gamepad (MDA:M == 228.0s, SD == 62.2, MA:M == 294.2s, SD == 143.6, Gamepad:$M$== 426.8s, SD == 227.9). The results of the one-way ANOVA and Bonferroni test confirmed that the time required for the MDA was significantly shorter than that for the gamepad ($p$== 0.013). The findings of this study will facilitate the design of interactions with improved operability.