Traces: Embodied Immersive Interaction with Semi Autonomous Avatars

Abstract

Traces is an artwork for the CAVE that uses a novel machine vision system to enable unencumbered full body interaction with a range of semi-autonomous agents without the imposition of any sort of textual, iconic or encoded-gestural interfaces and without physically restrictive wiring, pointing devices, or headgear. Furthermore, Traces does not consist of a ``world'' which is ``navigated''; instead, the movement of the user through the space leaves volumetric and spatial-acoustic residues of user movement which slowly decay. This project was motivated by a desire to explore and critique four central issues in contemporary HCI: (a) embodied interaction with computational systems; (b) rapid and transparent learning of interfaces by untrained users (the autopedagogic interface); (c) immersive bodily interaction with software agents, (d) extension and elaboration of the general conception of ``interactivity'' itself. To explore these issues, we built an infra-red multi-camera machine vision system which constructs a volumetric model of the whole of the users' body in real time. We have also developed custom 3D vision tools, graphical techniques and a range of techniques for generating and managing semi-autonomous agents in immersive environments.