Integrated Virtual Humans
Building embodied virtual humans is a complex multi-disciplinary effort that requires many components to be connected together. Additionally there is a great need to support both the research and the transition of this technology. To this end a common general architecture was researched and developed for the overall virtual human research and development effort.
The architecture allows researches to further understand all of the intricate interconnections between the various components in the system, to enable researchers to test their components with the system as a whole and to facilitate interesting synergies between the various research groups.
To demonstrate the virtual human system and components, a set of negotiation scenarios have been constructed for a large interactive, immersive, virtual training environment prototype. In this environment, a trainee uses natural language to interact with a life-sized virtual human agent and perform a negotiation task.
The first scenario is the Stability and Support Operations--Simulation and Training (SASO-ST) where a user negotiates with a single agent to move a clinic out of harms way.
The second scenario is the Stability and Support Operations -- Extended Negotiations (SASO-EN) where a user negotiates with multiple agents to convince them both to move the clinic to another part of the town.
This project differs from others in the following ways:
- It is comprehensive in integrating all of the various components required to build Virtual Human agents and embedding them into virtual and real environments. (Few projects attempt to bring together so many pieces of technology.)
- It is highly interactive. Users interact with a life-sized virtual human naturally with multi-modal input in an immersive environment.
- It contains full-body avatars. The virtual humans consists of a complete virtual body, with facial expressions, non-verbal behavior, walking and motion, speech input and output, emotion modeling, decision making and goal-directed behavior.
- It allows multi-party interactions. The immersive environment is capable of supporting multiple virtual humans and multiple users in the same virtual world.
- It is a dynamic environment. The virtual world can be modified internally by the agents or externally by a user or trainer for more story-oriented scenarios.
Tags: architecture, human, immersive, prototype, scenario, virtual
Goals
- Develop a general virtual human architecture to be used by all projects that are developing virtual humans. The architecture should be plug and play, distributed and flexible. Additionally a common set of messages developed in conjunction with international research groups is used between the components. See Figure 3 for the architecture. These tasks will research new ideas, improve and enhance the architecture and to investigate designs or components that will allow the virtual humans to be utilized better for training purposes.
- To develop a suite of tools called the Virtual Human Toolkit, which also includes the architecture. This set of tools for both developers and end users facilitate building virtual humans for different domains and automate some of the hard tasks of creating the knowledge and tasks for the characters.
- To develop knowledge, tasks, goals, verbal ad non-verbal behavior for new and exciting domains for entertainment, teaching and military training.
- To support the integration and testing of the whole virtual human system and supporting research components.
- Support the construction of virtual environments with all the components required to support interaction with Virtual Human agents, including vision and speech of the trainee or user.