**Abstract**

Placing virtual objects accurately in Mixed Reality (MR) remains challenging, particularly when target locations are physically inaccessible or unsafe to reach. Conventional techniques, such as direct hand manipulation and raycasting, can limit placement precision, comfort, and control. Recent MR headsets provide reconstructed environmental geometry that can support context-aware object placement. However, these reconstructions often contain noise, missing regions, and geometric inaccuracies, potentially causing misplaced or incorrectly aligned virtual objects. Furthermore, placement techniques must accommodate objects positioned near physical surfaces without necessarily attaching them directly.

This internship aims to design, implement, and evaluate interaction techniques that leverage reconstructed environmental geometry to assist manual object placement in MR while preserving user control. The proposed techniques will support placement relative to physical surfaces and objects, including inaccessible locations. Particular attention will be given to robustness against reconstruction imperfections through approaches such as geometry estimation, noise reduction, adaptive placement constraints, and uncertainty feedback. The goal is to improve placement accuracy, usability, and robustness.

• Niveaux d’étude
• Stage Bac+5