Data Simulator for TLA (DATASIM)

An open source application that provides distributed learning scientists and stakeholders with the ability to produce large simulated datasets to benchmark and stress test the Total Learning Architecture (TLA).

The Challenge

Modernization efforts across the DoD require large datasets to properly research, test, evaluate, and integrate new IT components—especially in the development of the TLA. However, existing datasets do not represent the breadth of data used across the DoD, which makes it difficult to test and evaluate new tools and technologies against technical standards at scale.

The Solution

DATASIM is designed to create simulated performance data at scale for the purpose of improving distributed learning processes, systems, instructional design best practices, experimentation, and acquisition. DATASIM data-generated tools can be used to create large simulated datasets conformant with current TLA specifications and standards.

About the Project

DATASIM provides the foundation for data interoperability and conformance testing by generating tailorable streams of data based on candidate TLA specifications or user-submitted xAPI Profiles. DATASIM leverages xAPI Profiles to generate realistic xAPI Statement data for a population of simulated actors. These sets of statements may be used to benchmark and stress test components of the TLA and other distributed learning projects that use xAPI and xAPI Profiles as a solution. This project is intended to assist stakeholders in determining the effectiveness of their xAPI data design and implementations.

This project focuses on the creation of a software application to simulate learning activity at any scale, and to generate the activity’s event data as valid and meaningful xAPI Statements. DATASIM simulations are comprised of an environment made up of one or more xAPI Profiles. Within this environment reside artificial agents that simulate learners whose characteristics are set by the user, and are described by their behavior and activity. DATASIM leverages simulation parameters (e.g., xAPI Profile, simulation length, number of learners, learner attributes) which are set by the user to produce a median baseline dataset. That dataset provides a model against which user-defined simulations can be compared and validated.

DATASIM helps facilitate the adoption of xAPI and other TLA technologies by reducing the cost and risk of research and pilot projects, while also mitigating risk for full-scale technology modernization efforts throughout the DoD. DATASIM also provides a means of generating simulated data of a learner population without the cost of expensive live exercises or interruptive training events. The reduced need for human subjects allows for the repeated testing of tools, technologies, processes, and technical standards at scale.

R&D efforts in 2019 produced a limited-scale prototype of the DATASIM software that can generate a large-scale domain-based xAPI dataset. This prototype can be used to benchmark and stress-test applications with realistic, context-sensitive xAPI data. The DATASIM project is currently creating authoring tools to configure data streams specific to different specifications and standards (e.g., xAPI activities, paradata, metadata, mastery estimates). The software will align testable strategic learning and instructional approaches to new xAPI Profiles designed to support the automation of simulated training activity.

As DATASIM is further developed, users will have greater control over the simulation parameters and learner attributes, and will be able to reproduce simulations to further develop and fine-tune. The next period of work involves creation of an instructionally sound profile of performance data for Tactical Combat Casualty Care (TC3) courses which teach evidence-based, life-saving techniques and strategies for providing the best trauma care on the battlefield. These efforts will include modeling of the varying degrees of student behavior within the TC3 environment.