Plan Understanding for Mixed-initiative control of Autonomous systems (PUMA)

Future naval operations in the littoral environment will make use of unmanned vehicles to support a range of operations, including intelligence, surveillance, and reconnaissance (ISR) and mine warfare. Many challenges that arise in pursuing such capabilities are rooted in the need for effective communication between human operators and autonomous mission planning/execution systems, such that the human operator can clearly and rapidly understand mission plans generated by autonomous systems, and interact with those systems decisively. To address this need, we propose an iterative design, development, and evaluation process focused on Plan Understanding for Mixed-initiative control of Autonomous systems (PUMA). Our technical approach has three components: 1) a suite of advanced visualization tools supporting intuitive interactions between operator and autonomous systems; 2) an agent-based computational plan analysis module that provides cognitively congruent semantic overlays on top of system-generated plan descriptions to better enable rapid human understanding; and 3) enhanced mixed-initiative interaction mechanisms that support the operator's role as mission monitor in a manner that does not degrade situation awareness nor introduce cognitive overload. Finally, we propose to employ a structured technology evaluation process based on cognitive walkthroughs of developing PUMA features by subject matter experts.