Year: 2025
Type: preprint
Abstract: Avatar-based mixed reality (MR) telepresence has shown significant promise in simulating in-person interactions from remote locations. However, translating the user's locomotion into the avatar's environment remains a challenge due to differences in spatial and component configurations between the two environments. A common situation arises when obstacles exist in the avatar's space but not in the user's. Since the user is unaware of these obstacles, directly translating their position to the avatar's environment can cause the avatar to unnaturally pass through them, disrupting the realism of the experience. Traditional robotic obstacle avoidance solutions are not applicable in telepresence scenarios, as the avatar does not move independently, but must continuously represent the user's real-time motion. To address this challenge, we build upon a real-time mesh deformation framework to dynamically map a user's path onto the avatar's space. We propose a novel two-part solution: (i) a method for generating smooth, natural-looking avoidance paths around any irregularly shaped obstacles using minimum-area ellipses (MAEs), and (ii) an Augmented Radial (A2R) projection technique that maps the user's path through the obstacle onto the avoidance path, effectively reducing discontinuities for common walking paths found in traditional radial projections. Through simulated and real human walking paths across diverse obstacle shapes, we demonstrate that the A2R method produces more continuous and efficient paths, reducing traversal time while preserving synchronization between the user and avatar. This work significantly enhances the spatial coherence and realism of avatar locomotion in MR telepresence, broadening its applicability to richer, more embodied remote interactions.
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Funder: National Science Foundation
OpenAlex ID: W4411396508