2021 IEEE Virtual Reality and 3D User Interfaces (VR) Larger Step Faster Speed: Investigating Gesture-Amplitude-based Locomotion in Place with Different Virtual Walking Speed in Virtual Reality Pingchuan Ke* Kening Zhu† School of Creative Media School of Creative Media and City University of Hong Kong Shenzhen Research Institute City University of Hong Kong ABSTRACT walking travel in VR, how to control the virtual walking speed to In this paper, we present a series of user studies to investigate the reach the far-away target quickly and precisely remains to be solved. technique of gesture-amplitude-based walking-speed control for loco- For the limb-based gestures (i.e., arm swinging and foot stepping), motion in place (LIP) in virtual reality (VR). Our 1st study suggested existing works usually utilized the swinging/stepping frequency [54] that compared to tapping and goose-stepping, the gesture of marching or the amplitude of the foot/arm movement [6] to control the walking in place was significantly preferred by users across three different speed. Torso-leaning-based approach is another LIP option where the virtual walking speed (i.e., 1×, 3×, and 10×) while sitting and stand- angle of torso leaning could be adopted for speed control [22]. While ing, and it yielded larger motion difference across the three speed the leaning posture may provide a similar vestibular experience of levels. With the tracker data recorded in the 1st study, we trained a real-world walking [20], it is often used in specific scenarios with Support-Vector-Machine classification model for LIP speed control continuous movement, such as flying [20, 28, 60], dragon riding [42], based on users’ leg/foot gestures in marching in place. The overall and so on, rather than step-by-step walking. In addition, research accuracy for classifying three speed levels was above 90% for sitting showed that leaning-based locomotion may suffer from low accuracy and standing. With the classification model, we then compared the and speed [15]. Considering the step-by-step walking in place in VR, marching-in-place speed-control technique with the controller-based existing research shows that compared with the step-frequency-based teleportation approach on a target-reaching task where users were speed-control method, the step-amplitude-based approach could sitting and standing. We found no significant difference between the improve the effectiveness and the precision in the short-distance two conditions in terms of target-reaching accuracy. More impor- travel, and offer better speed control in the medium- and the long- tantly, the technique of marching in place yielded significantly higher range travel [6]. While both step-frequency- and -amplitude-based user ratings in terms of naturalness, realness, and engagement than speed-control approaches have been proposed for LIP in VR, the the controller-based teleportation did. maximum speed of these techniques was usually around 2 times of normal walking speed. Higher virtual walking speed could Index Terms: H.5 [INFORMATION INTERFACES AND PRE- potentially facilitate the large virtual world. At present, there is few SENTATION (e.g., HCI)]: Multimedia Information Systems— research on the method of switching across a wide range of the virtual Artificial, augmented, and virtual realities walking speed (e.g., from 1× to 10× or even higher) for LIP in VR. In this paper, we focus on the foot/leg-gesture-based LIP 1 INTRODUCTION techniques leveraging the footstep amplitude to control wide range Spatial locomotion/navigation is one of the essential tasks in immer- of virtual walking speed, from a normal walking speed to a much sive virtual reality (VR) [35]. With the explosion of immersive VR in higher speed level. In Study 1, we investigated how users may the consumer market, a number of hardware and software solutions perform their foot/leg-based gestures/movements to match the virtual have been developed to improve the efficiency and the experience of walking speed while sitting and standing. This was mainly to find spatial locomotion in VR. Locomotion control can rely on either using out the foot/leg gesture that is preferred by the users for matching physical peripheral devices (e.g. handheld controllers or joysticks) or with different walking speeds in VR (i.e., 1×, 3×, and 10× of the gesture-based interaction using the motion of body parts (e.g., physi- normal walking speed). The results showed that marching in place cal walking, and walking in place - WIP). WIP is a popular locomotion (MIP) received the highest ratings from the participants, and it technique that enables users to perform virtual locomotion by perform- yielded more distinguishable motion features for gesture-based gain ing a stepping-like gesture without physically moving around [34, 44, control. With the tracker data recorded in Study 1, we trained two 48]. Compared to using external handheld controllers/joysticks, WIP Support-Vector-Machine models for three-level speed classification allows for better spatial awareness [23] while approximating the body in standing and sitting respectively, to support the automatic speed motion of real-world walking [40], thus offering the proprioceptive switching based on the footstep amplitude. At the current stage, we feedback to reduce the motion sickness [18]. Unlike unrestricted real- focused on the discrete speed classification/selection instead of the world physical walking, which may require a large physical space to continuous speed estimation, as previous research showed that the match the VR space or enable walking redirection [38], the WIP tech- discrete speed selection in VR locomotion could potentially reduce nique allows users to perform the VR locomotion in small physical ar- the motion sickness [10,51]. Lastly, we conducted the user study to eas. Researchers also proposed other locomotion-in-place techniques, compare the effectiveness of virtual target reaching using three tech- such as leaning, which require even less body movement [25, 28, 53]. niques: controller-based teleport (Tele), MIP with controller-based In addition, researchers proposed the VR locomotion technique of speed control (cMIP), MIP with gesture-based speed control (gMIP) arm swinging without the need of foot motion [27, 31, 37]. However, supported by the classification model. The results showed that gMIP the requirement of arm/hand movement for LIP in VR may affect the achieved a comparable target-reaching accuracy as Tele did, and was simultaneous hand-based tasks, such as object manipulation [13, 52]. rated to be significantly more natural, real, and engaging. While a user is performing the gestures/postures for locomotion The paper makes contributions in three folds: in place (LIP), one of the challenges could be the ability to naturally control the virtual walking speed. For instance, in a long-distance • We investigated the users’ behavior and preference towards *[email protected] gesture-amplitude-based speed control for LIP in VR. †Corresponding author: [email protected] • We experimented and developed the data-driven gesture-amplitude- based speed-control models for LIP in VR. • We evaluated the effectiveness of gesture-amplitude-based LIP speed control for a target-reaching task in VR. 2642-5254/21/$31.00 ©2021 IEEE 429 DOI 10.1109/VR50410.2021.00067 2 RELATED WORK Translational Gain [55]. Bolte et al. proposed the Jumper Metaphor The presented research is highly inspired by the existing works on which combines natural direct walking with teleport through locomotion in place (LIP) and walking-speed control in VR. large-scale virtual environments [3]. By predicting the user’s walking direction and target location, the Jumper Metaphor directly translate 2.1 Locomotion in Place in VR the user to the target in a high speed with a fast and blurred animation of movement. Wilson et al. found that in an interactive room-scale Researchers have proposed various types of LIP techniques, virtual environment, when using the large virtual walking speed, including button trigger in the physical controllers (e.g., teleport and positional accuracy diminishes at speed gains beyond 2× [56]. joystick), stepping in place (e.g., marching and tapping in place), arm Abtahi et al. investigated the speed-gain-visualization methods on swinging in place, body leaning, and so on. Buttussi and Chittaro the gain levels of 1×, 3×, and 10× for real walking in VR [1]. provide a thorough literature review on various types on locomotion Enlarging the translational speed may induce motion sickness techniques in VR [7]. The common movement trigger was pressing while using VR. To address this issue, Interrante et al. proposed Seven the direction buttons in the handheld controllers or joysticks [41]. League Boots [17], which amplifies movements along the users’ However, this may introduce serious motion sickness in VR due to walking direction, without modifying the sideways movements that the mismatching between the still body in real and the moving visual naturally occur when walking. Interrante et al. have also shown that content in VR. To address this problem, the locomotion technique users maintain better spatial awareness when using Seven-League of Point&Teleport was proposed to reduce the visual motion and Boots, compared to flying with a magic wand [16]. Moreover, they support point&click-based instant locomotion in VR [4]. To support have found that users overwhelmingly prefer Seven-League Boots hand-free locomotion in VR, foot-based teleportation was also to 2D Translational Gain [17]. Later, Bhandari et al. introduced proposed and studied [52]. However, existing research showed that Legomotion to allow users switch between real walking and walking the Point&Teleport technique still may not provide as much presence in place to enable navigation at scale in VR [2]. as body-motion-based locomotion techniques [8]. In certain contexts, While above-mentioned existing work focused on different the Point&Teleport technique may also negatively affect the time presentation/visualization techniques for different speed levels, they spent on target reaching, especially with obstacles on the way [5].