Abstract

A Knowledge-Based Diagnosis Approach for Real-Time Training Simulators

Madni, A.M., Meshkinpour, F., Giboney, V., Madni, C., McMeel, K., and Piper, A.

Abstract

It has been amply demonstrated that training simulators can be a cost-effective means of supporting otherwise costly training programs. The recent application of Artificial Intelligence techniques to the training problem has produced an Intelligent Computer-Assisted Instruction (ICAI) technology that supports individualized one-on-one instruction. The critical aspect of an ICAI system is the diagnosis of student performance errors and the identification of the causes of these errors. In this work, we present our overall approach for diagnosing individual deficiencies using knowledge-based ICAI techniques and an application of the approach within a real-time rifle marksmanship training simulator.

In our approach the trainee's performance is monitored via sensors, and the data is analyzed and compared to the simulated performance of an executable expert marksman model. The trainee interacts with the instrumented weapon and simulated target range via a real-time data acquisition system. The expert marksman's performance profile is generated by the executable expert model in response to the simulation scenario. The expert model is based on a composite knowledge representation formalism, and is fully inspectable and modifiable via an instructor/experimenter interface. In our approach, diagnosis is viewed as a process of evaluating competing hypotheses in the student model based upon a comparison of the student's performance with the expert model-generated performance profile. With minimal modifications, this approach can be customized to other diagnostic problem-solving domains.

From:

Madni, A.M., Meshkinpour, F., Giboney, V., Madni, C., McMeel, K., and Piper, A., Proceedings of the 1988 Summer Computer Simulation Conference, Seattle, Washington, July 25-28, 1988.