Online character motion adaptation to similar indoor environments

Abstract

With the advance of augmented reality (AR) technologies, the demand for people in different spaces to communicate through avatars has increased. When a remote user is augmented as an avatar in a local space, the avatar is expected to move around and interact with objects in the local space following the remote user’s movements. Due to the environmental differences between two spaces, the remote user’s motion should be adapted according to the configuration of the local space. To achieve this, we propose the environment mapping function that defines the correspondences between two spaces. Given the remote user’s placement, we determine the avatar’s global placement using the environment mapping function. Following the basis of the existing approaches, we estimate the adjustment level for each joint and determine the desired joint positions where the avatar can preserve the context of the original motion. Then with the proposed optimization-based motion adaptation scheme, we adjust the remote user’s pose so that the avatar can naturally interact in the local space. The results show that the proposed method places the avatar in spatially appropriate placements, and ensures natural interaction with local objects which have different configurations and shapes with remote objects. The proposed method is useful for the real-time telepresence applications.