Ensuring Robust Agent Control in Uncertain Environments

Agent control involves reasoning about local problem solving ac tivities interacting with other agents planning for a course of action and contingencies in the event of failure of the action and nally carry ing out the actions with limited resources and uncertainty about agent outcomes and the actions of other agents The growing complexity and dynamics of agents and the environments in which they interact requires robust agent control where the chance of complete failure of an agent s plan to achieve the high level goal is minimized Design to Criteria is a soft real time agent control process where schedules are built to meet dynamic client goal criteria including real time dead lines using a task model that describes alternate ways to achieve tasks and subtasks In this paper we describe a post scheduling con tingency analysis process that can be employed in deadline critical situations where the added computational cost is worth the expense We describe the uncertainty representation and how it improves task models and the scheduling process and provide empirical examples to show how robustness is built into agent control E ort sponsored by the Defense Advanced Research Projects Agency DARPA and Air Force Research Laboratory Air Force Materiel Command USAF under agreement number F and by the National Science Foundation under Grant number IIS and number IRI The U S Government is authorized to reproduce and distribute reprints for Governmental purposes notwithstanding any copyright anno tation thereon The views and conclusions contained herein are those of the authors and should not be interpreted as necessarily representing the o cial policies or endorsements either expressed or implied of DARPA NSF USAF or the U S Government