A Framework for Model-Based Adaptive Training

Scheduling System - Conclusion

The existing scheduling knowledge base has been created by experienced knowledge engineers. The knowledge base system, as it is implemented for the purpose of order scheduling, is described in [van de Brug & Brown 1995]. An extension to impose an explicit control structure on the knowledge base has been implemented by the author.

A re-engineered knowledge level design using problem spaces, tasks and problem solving methods has been linked to new symbol level code using the RIME methodology [van de Brug, Bachant & McDermott 1986]. It should be noted that within the MOBIT project, the term “expert” is domain knowledge implemented for the purpose of performing a demonstration to a trainee by making use of a simulation model [MOBIT 1994].

In the MOBAT framework the term “expert” is extended to also perform general reasoning at the domain level in the form of an expert system. This expert system can act as an intelligent assistant explaining its reasoning in the domain as well as performing a demonstration using a simulation model.

The MOBAT framework embodies principles of learning within the training system itself. Principles for transfer of learning based on the SOAR categorisation of within-trial, across-trial and across-task are intended to play a major role. The Scheduling-MOBAT analysis has identified these transfer of learning methods as attributes for measuring training unit learning.

This type of research is ongoing for model-based training techniques. Early results indicate that these transfer of learning methods are strengthening and reinforcing the classification of distinct expertise models. Within-trial learning can be recognised with small chunks of knowledge (principles) as they are re-used to accomplish a task.

Across-task learning can be recognised with chunks of knowledge for a particular situation (associations) where the same action is appropriate from one task to another. Across-trial learning is generally recognised when a particular task is repeated with a sequence of operations (procedures).

In the industrial world, a tool set approach for system development is a practical way to represent the various options in knowledge representation. This is supported within the training system architecture and methodology. A tool set approach can bring advantages in terms of reduced development time and reduced cost. For example, it may be possible to develop different expert models in parallel (see concurrent engineering in Section 3.5 Life Cycle Considerations).

A tool set approach supports the creation of domain experts independent of the instructional components, in a loosely coupled way. The important factor here is the coherent use of the different knowledge representation levels within the training framework. The mapping of expected trainee goal expertise and preferred trainee learning mode determines the type of knowledge representation within the MOBAT training framework.

Different expert models and instructional methods can be combined for a given training objective, enabling the system to be more effective and more general for a range of industrial training applications.

Flower Show
© | | Sitemap