OSCAR is a computer-based architecture for an artificial rational agent, based upon a general-purpose automated reasoner. The objective of this project is to develop a decision-theoretic planning system within the OSCAR architecture. Existing automated planning systems tend to make unrealistic assumptions about the planning environment, the most serious being (1) that the planner has a fixed knowledge base that does not change over time, and (2) that it is known with certainty what the result will be of taking any relevant action in any describable situation. This project is aimed at relaxing these assumptions by making use of innovative features of the OSCAR architecture. The automated reasoner upon which OSCAR is based is capable of more sophisticated reasoning than has been available in the past to automated planners. This will culminate in an automated planning system capable of operating in richer environments than has been possible in the past, and will be useful for diverse tasks ranging from automating manufacturing to launching a space shuttle.
