Tactical Explanations

Problem Context

In planning military operations, commanders devise courses of action (COAs). Artificial Intelligence can support by generating these COAs. In case AI is used, commanders also need to understand the (dis)advantages of these COAs to react to the predicted potential adversary actions. Therefore, COAs must be explained in the language of the commander.

Solution

The project focuses on explaining these COAs, using tactical terrain features and intuitive concepts of complex military considerations. The results are shown in a use case for the Robotic Autonomous Systems unit, where AI COAs for unmanned ground vehicles are generated and explained in military terms.

Results

2023 - In collaboration with the 13th Brigade Robotic Autonomous Systems (RAS), we settled on a “Seize” scenario to show case the developed functionality. Frequent interactions between developers and the (RAS) end users were held to ensure the suitability of results. Initial steps were made to create an ontology to be used for the explanation and a user interaction design. A developed prototype performs a statistical analysis in a model free approach for connecting data from the learning process to named geo-spatial features. This first step should be validated with the RAS unit in the future.