Replan Understanding for Heterogenous Unmanned Vehicle Teams

As unmanned vehicles (UVs) become increasingly autonomous, the current multiple human to single vehicle ratio is likely to be reversed in that a single human operator may direct a team of UVs at the supervisory control level. A single human operator supervising UVs in the context of a time-critical mission will require dynamic adaptation of the mission plan to account for emergent and unexpected events. Such dynamic replanning can be aided by an automated planner (AP), capable of assisting the operator in generating mission schedules and adapting those schedules when the mission scenario changes. Such automated decision support should help an operator evaluate the AP‘s proposed changes to the mission schedule in light of emergent events, as well as understand the potential consequences and benefits for enacting the proposed changes. While an AP should produce recommendations based on exogenous mission changes, the AP should also accept input allowing for human-automation collaboration, since humans may need to input parameters for which the AP cannot account. To this end, this thesis describes an interface including a schedule management decision support tool (DST) designed to allow a single-operator to control multiple heterogeneous UVs performing a search and track mission. In particular, the DST allows for the operator to both compare schedules generated by the AP and collaborate with the AP such that human input can be provided. This Schedule Comparison Tool (SCT) contains configural displays which graphically show pertinent information common to all schedules in a manner that allows for efficient and effective decision making. Two evaluations were conducted in order to examine and critique the interface which included examining cognitive processes of expected users and a functional evaluation of the interface‘s efficiency. The interface was shown to support a concise and effective presentation of large amounts of critical information, which allows for effective management of the search and track mission. The interface also supports mission functionalities more efficiently than a previously designed engineering interface. While increasing the accessible functions for an operator, actual interactions with the interface were reduced, on average, approximately 50%. The work presented in this thesis also includes potential areas for future research. Thesis Supervisor: Mary Cummings Title: Associate Professor of Aeronautics and Astronautics