Extending HLSR to Support Scaling Up to Complex Models for Training, Simulation, and Robotics

A long-term, cost-effective approach to addressing increased mission complexity and cost is to increase automation in operations and training. Although a variety of automation exists, the next significant advance will be to automate decision-making processes that currently rely on human experts. Such experts take years to train and are only available for a limited time, whereas intelligent software systems can be copied without limit, and they do not age or retire. Although now technically feasible to build intelligent decision-making systems, it remains expensive and difficult to engineer them, as well as to carry out the necessary supporting research into psychological modeling. However, High-level languages for software engineering have proven extremely effective at reducing costs for the development of complex software. Soar Technology, Inc. proposes to bring similar effectiveness to decision-making system engineering, reducing the cost of development and maintenance by 4-5 times. We will accomplish this by extending an existing HLSR language and compiler with knowledge patterns for building increasingly complex and human-like models. While this effort will primarily payoff for engineered systems, we argue that it also will improve the cost effectiveness and scientific consistency of cognitive modeling research.