Abstract

Telepresence robots offer a promising solution to mitigate the challenges of physical absence in healthcare settings, particularly by enabling remote interaction between patients and caregivers or relatives. This thesis contributes to the ReduSys research project and focuses on enhancing Robody, a social telepresence robot, by integrating personalized, photorealistic 3D avatars to increase the sense of presence and reduce feelings of eeriness that previously arose from 2D avatars during interactions. Traditionally, avatar creation requires skilled designers and manual labor, making the process time consuming and impractical for clinical applications. To address this, the thesis explores the automatic generation of photorealistic 3D avatars from monocular smartphone recordings, making the process accessible to non-expert users. A mobile application was developed to guide users through the recording process, presenting established guidelines to ensure high-quality input videos for avatar generation. Three state-of-the-art avatar generation pipelines, GaussianAvatars, INSTA, and IMavatar, were evaluated in terms of quality, robustness, generation time and long-term viability. Dockerized environments were created to ensure reproducible results and fair comparison on the same hardware. Based on technical evaluation and empirical observations, GaussianAvatars emerged as the most suitable approach for the Robody platform, producing the qualitatively best avatars. A clinical study involving nursing staff, patients, and their relatives was conducted at Schön Klinik to assess the impact of Robody with personalized 3D avatars in the intended environment. Results suggested that while the system was easy to use and enjoyable, perceived usefulness and willingness to adopt the system in a hospital indicate areas for improvement. Nonetheless, the inclusion of 3D avatars significantly enhanced emotional engagement, demonstrating strong potential for telepresence robots in healthcare environments.

Results/Implementation/Project Description

Conclusion

[ PDF (optional) ] 

[ Slides Kickoff/Final (optional)]