A Method for 3D-Scene Generation Using a Human-Like Language

We propose a mechanism by means of which a user with little expertise of programming can easily produce custom-made virtual scenes. A typical scene comprises virtual entities that take decisions in real time in response of perceived environmental changes. The actions performed lead to the fulfillment of one or more goals specified by the scenarist. We have developed an open agent-oriented system. It provides a number of tools useful to create virtual scenes, including a rendering and a scene descriptor. Besides, it allows the addition of new services, like behavior controllers, 3D-object descriptors and virtual worlds. The design of a world not only involves specifying the environment graphical features, but also the rules that control the interaction between entities. This paper deals with the system architecture and the language provided to describe both a scene and a virtual world