Integrating Virtual Humans into Complex Scientific and Technical Applications
Virtual Reality (VR) is on its way towards being a serious and powerful tool in scientific and industrial applications. As such, embedding Virtual Humans (VHs) becomes increasingly important: By means of VHs, advanced, emotional human interfaces can be designed and evaluated for increased usability and acceptance. In the medical field, a greater complexity of psychological and social research questions can be addressed by integrating VHs into behavioral studies. In the analysis of complex, heterogeneous data, VHs may advise and monitor users during the data exploration process. Additionally, VHs might serve as training partners and instructors, assisting users in executing complex, collaborative tasks or in improving certain personal skills. Furthermore, the development and evaluation of digitized production environments, i.e., cyber-physical systems, represents a highly relevant example: In production sites of the future, workers are expected to collaborate in teams, including humans as well as intelligent mobile robots. By means of VHs, such scenarios can be designed, evaluated and trained in a fully controllable virtual environment. Consequently, integrating VHs into scientific VR applications will open up new and innovative research opportunities.
In the scope of this project, we will lay the software foundation for integrating VH functionality into our existing VR framework. Additionally, we will focus on two research questions: (1) In order to be accepted by users, VHs need to show believable, human-like behavior. To this end, we will investigate how VHs have to react to an approaching user. Here, aspects like respecting the user’s personal space as well as automatically recognizing the reason for approaching, e.g., passing by or starting a conversation, and reacting appropriately are challenging. (2) Furthermore, predefining VHs’ behavior only by scripts is too inflexible for most use cases and using approaches purely based on artificial intelligence is not yet technically feasible. To this end, a hybrid approach is required: VHs execute predefined behaviors while being temporarily superimposed by external operators. Here, we are facing the challenges of how a single operator can efficiently and effectively control a set of VAs and which support an intelligent system needs to provide.