Hyperprogrammed agents for virtual environments

Behavioral animation is a fascinating combination of artificial intelligence and computer graphics. These systems typically use intelligent agents as a means to develop creatures which inhabit a virtual environment. An agent's behavior is implement though the use of competing reactive rules that control how it responds to both perceived external stimuli and internal states. Creating these rules takes time and effort to not only develop, but to understand how these rules interact. This dissertation describes a system for creating these rules using a programming-by-example approach.; This work extends the programming-by-example paradigm to work with 3D interactive graphics. These techniques allow a user to create an initial condition and then to demonstrate the action to be taken when this situation is encountered. Once a rule has been defined, it is available as an operator so that it can be combined to form higher level goal directed behavior.; The system consists of several parts: (1) a intelligent agent system, based on InterRap, which has been altered to support inclusion into a virtual environment; (2) a scripting language which is used a control language and means for world persistence; (3) a programming by demonstration system which allows users to develop reactive rules and higher level groups of rules.; Situations this system can design are: resolution of sensor input: (i.e. agent is blocked how does it respond), results of internal variable state changes, and changes in appearance. Complex rules can be designed by grouping previously defined rules to provide goal directed behavior. This system allows for complete environments to be constructed and populated with agents developed by the user. All rules created can be stored and reused for future use.; Additionally there is a VR-based execution environment, which allows a user to interact with worlds designed. It also servers as a debugger for it allows for the internal states of an agent to be adjusted and for new behaviors in the form of rules to be added.