Researchers led by Ammaar Zaman have compared ways of controlling a robot through a virtual reality headset, according to a paper on arXiv. Seventeen participants guided a four-legged robot through a physical maze, found marked symbols on the walls and photographed them, using a joystick or an omnidirectional treadmill to move and a joystick or their head to aim the camera.
Aiming the camera with the head led to faster runs, less time touching the walls, a lower workload and better ratings for presence, usability and acceptance than aiming it with a joystick. Walking by joystick with head-guided viewing gave the best measured performance, while the treadmill with head-guided viewing gave the best overall experience.
With the head controlling the camera, people switched views less and relied more on the pan-tilt view. Nearly all targets were captured in every setup, and simulator sickness stayed low. The authors conclude that how people control the view matters more than how they move the robot for this kind of task.



