Integrating realistic human group behaviors into a networked 3D virtual environment

Abstract

Virtual humans operating inside large-scale virtual environments (VE) are typically controlled as single entities. Coordination of group activity and movement is usually the responsibility of their real world human controllers. Georeferencing coordinate systems, single-precision versus double-precision number representation and network delay requirements make group operations difficult. Mounting multiple humans inside shared or single vehicles, (i.e. air-assault operations, mechanized infantry operations, or small boat/riverine operations) with high fidelity is often impossible. The approach taken in this thesis is to reengineer the DIS-Java-VRML Capture the Flag game geolocated at Fort Irwin, California to allow the inclusion of human entities. Human operators are given the capability of aggregating or mounting nonhuman entities for coordinated actions. Additionally, rapid content creation of human entities is addressed through the development of a native tag set for the Humanoid Animation (H-Anim) 1.1 Specification in Extensible 3D (X3D). Conventions are demonstrated for integrating the DIS-Java-VRML and H-Anim draft standards using either VRML97 or X3D encodings. The result of this work is an interface to aggregate and control articulated humans using an existing model with a standardized motion library in a networked virtual environment. Virtual human avatars can be mounted and unmounted from aggregation entities. Simple demonstration examples show coordinated tactical maneuver among multiple humans with and without vehicles. Live 3D visualization of animated humanoids on realistic terrain is then portrayed inside freely available web browsers.