Understanding situational awareness in multi-unit supervisory control through data-mining and modeling with real-time strategy games

As robots become increasingly capable and autonomous, the role of a human operator may be to supervise multiple robots and intervene to handle problems and provide strategic guidance. In such cases, the extent to which HRI tools support the human supervisor's situational awareness (SA) and ability to intervene in an appropriate and timely fashion will constrain the scale of operations (e.g., the number of robots; the complexity of tasks) that can reasonably be supervised by a single person. One approach to understanding how humans might acquire, maintain, and use situational awareness in multi-robot supervision tasks is to look at video games that require similar activities. We describe our initial efforts at analyzing and modeling data from Real-Time Strategy (RTS) games with the goal of answering basic questions about the nature of situational awareness and supervisory control of multiple semi-autonomous agents.