The development of a practical system for intelligent Virtual Reality-based industrial training (VR-ITS) is a difficult undertaking because of its inherent complexity both at the conceptual and implementation level. From this decade, although the hardware technology for the VR applications has been available, very few software tools have been developed for this new technology. From literature review, most of the researches on VR has concentrated on the presentation technology required to create immersive experiences, there is little attention has been paid on the high level modeling and dynamic realism at the viewpoint of system architecture of VR systems. As a result, there is a gap between high-level task model and lower-level implementation level. The main reason is that the developm...[ Read more ]

The development of a practical system for intelligent Virtual Reality-based industrial training (VR-ITS) is a difficult undertaking because of its inherent complexity both at the conceptual and implementation level. From this decade, although the hardware technology for the VR applications has been available, very few software tools have been developed for this new technology. From literature review, most of the researches on VR has concentrated on the presentation technology required to create immersive experiences, there is little attention has been paid on the high level modeling and dynamic realism at the viewpoint of system architecture of VR systems. As a result, there is a gap between high-level task model and lower-level implementation level. The main reason is that the development of these tools has been hindered by the lack of models and specification techniques as the basis for such kind software tool development.

First, in this research, a multi-agent system (MAS)-based framework for generic intelligent VR-ITS is proposed. By extending the ordinary Petri Nets model and defining a set of basic design modules, a Recursive and Extended Petri Net (REPN) model is presented. The REPN are used as a unified model of agents modeling to facilitate training task plan modeling, pedagogical knowledge acquisition and representation, and trainee behavior modeling. In addition, the combination of object-oriented technology (OOT) and REPN formalism could explicitly represent the static structure and dynamic behaviors of the virtual objects with precise semantics and at a manageable level. The coordination among agents and objects is formalized with extended "token"-passing mechanism and shared transitions.

Second, as real-time collision detection plays a critical important role for enabling and the interaction between the trainee and virtual training equipment, triggering the machine behavior simulation, an algorithm for detecting collision between two objects represented by Constructive Solid Geometry (CSG) is proposed. The method described here may be of interest to VR-ITS developers attempting to enhance the usability of VR-based training systems by reducing interaction time lag.

Third, a VR-based CNC milling operation training prototyping system developed at VR Applications Lab of the department of Industrial Engineering and Engineering Management at the Hong Kong University of Science and Technology using the proposed framework is expounded to demonstrate the effectiveness of methodologies proposed. This experience may provide a useful reference for researcher and developers attempting to design an intelligent and complex VR-ITS cost-effectively.