A Framework for Model-Based Adaptive Training

Introduction - The Training System Architecture

The overall structure is based on the ITSIE architecture [ITSIE 1992, D7], this augmented earlier intelligent tutoring system architectures to satisfy the requirements of industrial intelligent training systems.

The MOBAT framework requires extensions to the MOBIT training system architecture. The separation of MOBIT Expert and Trainer system modules has been taken a step further.

In the MOBAT framework there can be independently operating Expert and Trainer agents. Intelligent training agents need to operate with a strategy, they need to be based on models, they need to comply with an evolving rule set and they need to pursue training goals jointly with other agents. I.e., communicate and negotiate with one another on how the training is to be provided.

The specification framework takes into account the design of separate agents. The MOBIT training architecture has interfaces for the trainee and trainer.

In the MOBAT framework there is an additional interface for knowledge acquisition by the domain expert(s). To evolve a training knowledge base, the MOBAT framework recognises the need for an ongoing knowledge acquisition capability in the training system architecture.

The trainer and domain expert interfaces promote the use of distinct training agents.

The primary components in the training system architecture are:

  1. domain simulation – a representation of the training domain which supports the demonstration and validation of expertise by the trainee;
  2. domain expert agent(s) – explicit representations in an executable form of the expertise which is to be conveyed to the trainee; and
  3. trainer agent – the component in overall control of the training process.

The relationships between these components, together with the user-interface modules is shown in Figure 1-1. The domain components, the simulation module and the expert agent(s), provide the training system with executable models of the domain system and the human domain expert. The trainer agent provides a mechanism for controlling the interactions between the system and the trainee based upon measures of both a trainee's performance and hypotheses regarding a trainee's beliefs. Each of the main components is described further below.


Figure 1- 1 Training System Architecture (adapted from [Slater et al. 1994] )

Flower Show
© | | Sitemap