Framework for Authoring Repeatable Scenarios in Virtual Environments Populated with Stochastic Virtual Agents

Abstract

Virtual Environments (VEs) are used extensively for training systems because of their ability to provide interactive experiences. Effective use of such environments requires the ability to generate consistent scenarios, which involve one or more realistic yet repeatable interactions between the trainee and the virtual humans in the training environment. This paper presents a framework that facilitates the development of rich interactive scenarios between one or more virtual humans and a trainee. The framework utilizes a scenario authoring system with runtime elements that ensure the trainee encounters realistic and stochastic, yet repeatable, well-orchestrated and precisely choreographed situations to meet specific training goals. Core elements of the framework include the use of probabilistic state machines for controlling the behavior of virtual elements along with triggers which ensure pre-determined outcomes even when utilizing stochastic behavioral models for the virtual humans.