Developing virtual environments for industrial training
Introduction
Virtual reality-based training systems (VRTSs) are advanced computer-assisted training systems using Virtual Reality (VR) technology. Compared with traditional training approaches, these systems would allow trainees to properly operate new equipment before it is actually installed. The important perceptual cues and multi-modal feedback (e.g., visual, auditory, and haptic) provided to trainees enable VRTSs to more effectively transfer virtual training to real-world operation skills. More importantly, the systems can provide higher degree of freedom for operation and the results of improper operation can be simulated without incurring the associated costs in terms of human injury and equipment repair.
The objectives of this research are to develop an architecture of VRTSs and a knowledge modeling technique to overcome the difficulties arising from complexity of systems and knowledge variety, and to fill the gap between abstract task model and detailed implementation. A Petri net (PN) theory [1] is used as a tool for specifying virtual training task plans and training scenarios. An experimental VR-based Computer Numerical Control (CNC) milling operations training system (VR-CNC) [2] is developed to show the feasibility and effectiveness of the proposed architecture and approach.
Section snippets
Research background
Over the past few years, we have witnessed that increasing research progress on VRTS has been shown in a variety of applications [2]. The degree of interconnection among functions of systems would seem to motivate the development of a tightly integrated design system [3]. However, the concerns of system robustness, scalability, reusability, and maintainability lead to the adoption of a loosely coupled set of cooperating modules [4]. VRTSs are advanced computer-assisted training systems. By
A VRTS architecture
In this approach, a VRTS is composed of five interactive modules: a training task-planning module, a simulation module, a performance evaluation module, an instruction module, and an interface module. Fig. 1 illustrates the modules and the message flow among them. Briefly, the functionality of the modules is as follows.
The responsibility of the training task-planning module is to adaptively generate appropriate training task plans according to training goals input by a trainee and the profile
Virtual training task-planning knowledge
Training tasks are the basis of the training scenario modeling. Training tasks are constructed and organized toward some training goals. Sets of relevant training tasks are called training plans (TP). These training plans are then specified as specialized PNs, TP-nets, and stored in the knowledge base of the system. The following exploits the training goal decomposition and TP-nets construction.
Task-oriented training scenario models
To reduce the complexity of training scenario modeling and analysis, a training scenario is conducted for each task. From the TP-net and these task-oriented training scenario models, the whole training scenario model for a given training goal is obtained (see Fig. 5). A task-oriented training scenario model that is a PN consists of four sub-nets linked together: simulation sub-net, interface sub-net, instruction sub-net, and evaluation sub-net.
Background description
CNC machines are commonly used manufacturing devices. The training of CNC milling and drilling operations is one of the more important industrial training functions. Unfortunately, it is potentially hazardous and complex. Conventionally, if a casual user wants to learn how to operate a CNC milling machine, he/she must have some knowledge of hardware components and task procedures. The traditional approach would be to find an operating manual and follow the instructions to practice. It is not
Discussion
In comparison with other existing works, the distinctive contribution of this work is system and human-interface interaction modeling and formal specification. The main benefit of the architecture proposed is software reuse and usability. While the initial application considered is a single machining operation training, and more specifically CNC machine training, it appears that the architecture and mechanisms could be applied to other industrial training including manufacturing cell
Conclusion
We have described an architecture of VRTSs and a knowledge modeling approach to designing VRTSs. The PN theory has been used as a unique tool for representing training task plans and training scenarios. The application in the CNC operations training using VR shows the effectiveness of the approach. Future work includes the extension of the architecture and the approach to networked virtual environments and other types of industrial training tasks.
Acknowledgements
We would like to thank Prof. Mitchell Tseng and Dr. Benjamin Yen for their helpful comments during the planning and development of the software as well as the technicians Charles, Denil, Tin and Yung from the manufacturing labs of the Department of Industrial Engineering and Engineering Management at HKUST. As well, we would like to thank Asian Journal of Ergonomics for their permission to reprint Fig. 7, Fig. 8 in this paper.
References (18)
- et al.
Modeling production rules by means of predicate transition networks
Information Sciences
(1985) - et al.
Petri net application in flexible process planning
Computers in Industrial Engineering
(1994) - et al.
A new collision detection method for CSG-represented objects in Virtual Manufacturing
Computers in Industry
(1999) Petri Nets: An Introduction
(1985)- F. Lin, The development of Intelligent Virtual Reality-based Industrial Training Systems, unpublished Ph.D. Thesis, The...
- et al.
An intelligent training system for space shuttle flight controllers
Telematics and Informatics
(1988) - et al.
Petri net based modeling of the cooperative interaction in a multi-agent based intelligent learning environment
CESA'96 IMACS
(1996) The defining characteristics of intelligent tutoring systems research: ITSs care, precisely
International Journal of Artificial Intelligence in Education
(1999)Investigating the applicability of petri nets for rule-based system verification
IEEE Transactions on Knowledge and Data Engineering
(1993)
Cited by (107)
Fostering short-term human anticipatory behavior in human-robot collaboration
2022, International Journal of Industrial ErgonomicsCitation Excerpt :On the other hand, studies dealing with the human-centered perspective, i.e. human anticipation and predictability of robots' future moves are limited. Until recently, HRC researchers mainly focused their attention of fostering human's abilities as a collaborator, through training (Freedy et al., 2007; Lin et al., 2002; Matsas and Vosniakos, 2017; Nathanael et al., 2016b). However, an effective way to achieve successful HRC while reducing the aforementioned efficiency – safety tradeoff may involve interaction cues similar to the ones observed in Human – Human (H–H) collaboration.
Testing public interventions for flash flood evacuation through environmental and social cues: The merit of virtual reality experiments
2020, International Journal of Disaster Risk ReductionEXTENDED REALITY IN INDUSTRY: PAST, PRESENT AND FUTURE PERSPECTIVES
2023, Proceedings of the Design Society