Responding to unanticipated goals when planning travel for autonomous underwater vehicles

Agents must be able to react to the unpredictable elements of their world. At the same time, they can take advantage of reliable predictions that can be made about the world. In this paper, we present a method for taking advantage of such predictions to respond to unanticipated goals that become active after the agent has begun performing its task. The application we discuss is sequencing locations that an autonomous underwater vehicle (AUV) must visit to achieve its goals. Our method is embodied in NBA-PLANNER.<
>