Description

This thesis targets collaborative decision-making for building renovation (or similar technical reviews) using a multi-user AR app on Snap Spectacles. Participants co-view a simple shared 3D model, toggle elements, and query a lightweight voice agent; explicit turn-taking policies (e.g., hand-raise) coordinate who may act or speak, reducing ambiguity and improving clarity.

Implement a Snap Spectacles app that loads a 3D model with per-element metadata and visibility toggles, with real-time, consistent state synchronization across at least three users. Add a voice agent that answers model-scoped questions; organize the networking (latency, conflicts) and run small multi-user sessions to compare turn-taking strategies (free-for-all, token/push-to-talk, hand-raise/queue, moderator).

Tasks

  1. Conduct a comprehensive review of the state of the art in agent-interaction turn-taking policies, with a particular focus on AR, and networked AR experiences.
  2. Networking & Sync: Implement real-time multi-user state synchronization (at least 3 users) for simple on/off element toggles; handle conflicts and maintain consistency across devices.
  3. Voice Agent: Add a basic QA agent that answers questions about the model.
  4. Roles & Floor Control: Design and integrate turn-taking policies (e.g., free-for-all, token/push-to-talk, hand-raise/queue) that govern who can issue changes and when.

This topic’s tasks can be used for a MA.

Requirements

  • English (source code, comments, and final report should be in English)
  • Lens Studio familiarity; networking basics required. Speech/agent basics helpful.

Contact

For more details, please contact Diana Marin – diana.marin@tuwien.ac.at or Peter Kán – peter.kan@tuwien.ac.at.