Iteratively Adapting Avatars using Task-Integrated Optimisation

Abstract

Figure 1: (a) A person (b) embodies an avatar in VR that matches their morphology; however, many non-matching avatars could be better for the given task. (c) Our approach iteratively adapts the avatar to optimise the user's task performance. (d) Here, it produces an avatar with longer arms, helping the user reach a set of targets.

Beyond Being Real: A Sensorimotor Control Perspective on Interactions in Virtual Reality

“…For example, Ninja Hands maps the movement of a single hand to multiple hands to ease distant target selection [135]. Another paper iteratively adjusts the length, and therefore range of motion, of the avatar's forearms and fingers to achieve better performance on specific tasks [96]. Note that users may perceive space scaling transformations as body scaling and a form of alternate morphology.…”

Modeling the Noticeability of User-Avatar Movement Inconsistency for Sense of Body Ownership Intervention

“…One of the most widespread applications is input augmentation. Researchers leveraged the modified movement to change the movement speed or the arm length [8,24,29,37,59]. The Go-Go [37] technique extends the position of the hand in-depth non-linearly, enabling users to interact with objects beyond reach.…”

“…As previous studies explored the effect of scaling the lengths of the upper arm and the forearm on embodiment [29,59], we focus on how rotational changes at the shoulder joint and the elbow joint affect the sense of body ownership. We thus employed the spherical polar coordinate system, which defines the position of a point with its distance to the origin 𝑟 , its polar angle 𝜃 , and its azimuthal angle 𝜙, instead of the Cartesian coordinate system, which represents a point in space with a triplet (𝑥, 𝑦, 𝑧).…”

“…We acknowledge that there are offsets of other types that can also be applied to modify the movements. Existing research [29,37,59] has explored stretching the arm without a length restriction to enable reaching the space far away. We expect the research approach that we adopted in this study can also be applicable to the investigation of other offsets, including length and the combined usage of length and rotation offsets.…”

“…The motion consistency between the user and the avatar is essential to maintaining immersion in VR. However, breaking the consistency can enable new interactions with powerful functionalities, including input augmentation [29,31] and pseudo-haptic feedback such as weight simulation in VR [45] at the expense of body ownership. Yet, there lacks a thorough understanding of how this cost of body ownership changes as the movement inconsistencies become more significant, or more fundamentally, how the noticeability of the inconsistency reacts to the amplitude enlargement.…”