PerFECT: An Automated Framework for Training on the Fly

Abstract

Currently available cognitive training systems can highly benefit from more adaptable and encapsulated frameworks that include better performance assessment methods, robust feedback mechanisms and automated mechanisms that reduce the manual intervention and curriculum management required during training sessions. In short, there is an ardent need for an automated human in the loop training system that can effectively train cognitive skills required for military operations. An automated training system would be extremely beneficial if it can be easily coupled with a synthetic learning environment to function autonomously is an entirely data driven manner. Such a system would enable rapid deployment of key training scenarios, skills and tactics to war fighters and help them maintain a superior level of competence in the battlefield. An automated framework for training on the fly also known as performance feedback engine for conflict training (PerFECT) which includes key components for simulating training scenarios, measuring trainee’s performance, providing relevant feedback and dynamic curriculum management is discussed in this chapter. First, the training system comprises of custom plug-in interface that allows components of the training framework to readily interface with a simulated virtual learning environment. Second, it has a “Performance Evaluator” that enables automated, real-time and objective evaluation of a trainee’s performance grounded within an objective framework known as time window and enables run-time evaluation of performance skills based on a skills matrix. Third, PerFECT has a “Feedback System” that can provide contextual and immediate feedback to trainees based on process measures. Finally, PerFECT includes a “Curriculum Manager” that dynamically selects appropriate training scenario from a template library with varying levels of complexity. The selection algorithm for training scenario is based on the trainee’s historical performance scores and complexity of the earlier scenarios. We also present the initial findings from a pilot study which helps illustrate the capabilities of the framework and conclude with future directions in this area of research.