Eye on the Ball: The efect of visual cue on virtual throwing Goksu Yamac Trinity College Dublin Ireland yamacg@tcd.ie Carol O’Sullivan Trinity College Dublin Ireland Carol.OSullivan@tcd.ie Figure 1: VR experiment: a participant grabbing (l), throwing (m) and receiving visual feedback (r) ACM Reference Format: Goksu Yamac and Carol O’Sullivan. 2022. Eye on the Ball: The efect of visual cue on virtual throwing. In SIGGRAPH Asia 2022 Posters (SA ’22 Posters), December 06-09, 2022. ACM, New York, NY, USA, 2 pages. https: //doi.org/10.1145/3550082.3564181 1 INTRODUCTION Despite rapid developments in AR devices, their feld of view (FOV) is still much lower than for VR headsets (e.g., the diagonal FOV of 52 ◦ for the Microsoft Hololens™ vs. 113 ◦ for the HTC Vive™ Pro 2). This reduction in visual feedback can be problematic for certain tasks, such as ball throwing. We present an experiment in VR, where participants threw a virtual ball at virtual targets, with diferent levels of visual feedback. The objective of the experiment was to investigate how visual cues afect the way that participants perform virtual throws, which may be useful for the design of AR/VR systems. Eighteen participants used their own body motion to throw a virtual ball at virtual targets and we simultaneously captured their full body motion (mocap) using an optical motion capture system for ofine analysis. Previously, Zindulka et al. [2020] found that people are less accurate when throwing in VR than for real throwing, while Butkus and Ceponis [2019] found that throwing accuracy in VR increased with distance and that throwing velocity was higher in VR than in reality. These studies used a device to control the ball, whereas in our study, we use VR gloves to emulate more closely the experience of a real throw. Nusseck et al. [Nusseck et al. 2007] demonstrated the difculty with predicting the properties of a bouncing ball during manipulations of the trajectory’s visibility. We also vary the visibility of a thrown ball’s trajectory in VR. Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s). SA ’22 Posters, December 06-09, 2022, Daegu, Republic of Korea © 2022 Copyright held by the owner/author(s). ACM ISBN 978-1-4503-9462-8/22/12. https://doi.org/10.1145/3550082.3564181 2 VR EXPERIMENT We developed the experiment in Unity™ and used the HTC Vive™ headset for display. The physics of the VE were governed by Unity’s built-in physics system, NVIDIA PhysX™. Our interaction and capture methods are similar to those of Yamac and O’Sullivan [2021]. Interactions were visualized using the virtual hand rendered from ManusVR™ glove data, which is generated from 10 sensors placed on the glove. A Vive controller attached to the participant’s arm was used for real-time arm tracking (to avoid interference from Vive trackers with the mocap system), which left the hands free to perform the throwing motions. A grey block, a 5cm diameter ball and 50cm diameter targets are displayed (see Figure 1). Participants stood next to the block (based on handedness), facing the target spawn area. For grabbing, the glove provides fnger fexion values for when the hand closes around the ball. This was set for each participant as they held a real tennis tennis ball. After a grab, the ball’s position was interpolated towards the predefned center of the hand and followed the hand thereafter. The velocity assigned to the ball at the point of release was estimated over a window of the previous nine frames. The release mechanism was implemented from the rate of change in fnger phalanges, which proved to be a good indicator of an opening hand. Once the rate of change passed a predefned threshold of 3 ◦ /sec, the ball was released from the hand and the estimated velocity was assigned to the ball. When a target was hit or missed, the target turned respectively green or red. Two levels of visual feedback Mode were presented: (i) in Full mode, the ball was visible throughout the full trajectory of the throw; and (ii) in Minimal mode, the ball was only visible until it was grabbed, after which it would gradually fade in the participant’s hand and remained invisible for the duration of the throw. Therefore, they only knew whether they hit the target or not when it changed color. To vary the Distance of the targets, we divided the foor surface into three regions with a width of 2m, referred to as Near (1.25-2m), Mid (3.5-4.25m) and Far (5-5.25m). The depth of the Far region was reduced in size during testing, as it proved to be very difcult to hit any targets beyond that distance. Targets were spawned randomly within these regions at run-time.