John F. Horty, Martha E. Pollack
Progress in cognitive robotics can come either from efforts to provide physical robots with more autonomy, or from efforts to make AI techniques more generally applicable to the needs of robotic agents. We are conducting research on dynamic plan management, which is an instance of the latter type of effort. In this paper, we provide an overview of our research project. Plan management is the process (or set of related processes) by which a goal-directed agent coordinates, updates, and monitors its plans in response to ongoing changes in its environment. We discuss the main challenges in automating plan management, and then illustrate our approach by focusing on one plan management task: evaluating options for action in the context of existing plans. We also briefly describe the Plan Management Agent (PMA), a system we are building as a testbed for our plan management theories. Although PMA is not a robotic system, we are interested in also applying dynamic plan management techniques to mobile robots, which is why we are participating in this symposium.