Intelligent behavior control and world modeling for virtual humans in virtual environments

Autonomous virtual humans must be able to solve problems in their virtual environment. Starting from an animation system that was developed in two other diplomna theses, a system for autonomous virtual humans was developed. The base of the system is a framework for behavior control of the virtual human. It consists of three parts: A perception module that gives information about the virtual environment and user commands to the virtual human, a planning system for the creation of plans, and an action manager that is responsible for the execution of the plans. The planner is a simple rule-based (reactive) planning system that can be extended with specialized planning algorithms. Those planning algorithms solve complex problems that are too hard for the reactive planner. Two complex problems that were dealt with are path-finding and grasping of objects. For path-finding one needs a proper representation of the virtual environment which is then used for searching a path. The A*-algorithm was used for the actual search on that representation; it was extended to take the height, width, step height and step length of the virtual human into account. Additionally, a connection had to be made between the path-finding module, the animation module, and the arbitrary scene. This enables the virtual human to follow a found path. The architecture of the system integrates the perception system, the planning system with specialzed planning algorithms, and the action manager with the visualization of the scene