A rule-based approach to animating multi-agent environments

This dissertation describes ESCAPE (Expert Systems in Computer Animation Production Environments), a multi-agent animation system for building domain-oriented, rulebased visual programming environments. Much recent work in computer graphics has been concerned with producing behavioural animations of artificial life-forms mainly based on algorithmic approaches. This research indicates how, by adding an inference engine and rules that describe such behaviour, traditional computer animation environments can be enhanced. The comparison between using algorithmic approaches and using a rule-based approach for representing multi-agent worlds is not based upon their respective claims to completeness, but rather on the ease with which end users may express their knowledge and control their animations with a minimum of technical knowledge. An environment for the design of computer animations incorporating an expert system approach is described. In addition to direct manipulation of objects on the screen, the environment allows users to describe behavioural rules based upon both the physical and non-physical attributes of objects. These rules can be interpreted to suggest the transition from stage to stage or to automatically produce a longer animation. The output from the system can be integrated into a commercially available 3D modelling and rendering package. Experience indicates that a hybrid environment, mixing algorithmic and rulebased approaches, would be very promising and offer benefits in application areas such as creating realistic background scenes and modelling human beings or animals either singly or in groups. A prototype evaluation system and three different domains are described and illustrated with preliminary animated images.