Intelligent agents can use plans to reach their goals. However, sometimes plans fail during execution due to the agent's lack of knowledge or an unexpected outcome of an agent's action. Replanning is an expensive albeit necessary remedy for this problem.;In this dissertation, the author examines components of a plan creation and execution system built specifically for agents that exist in dynamic environments.;The system that is presented here has as its foundation decision networks which are grounded in probability and decision theory. The decision networks are used to infer preconditions and effects of actions, store plans and guide execution.*;*Originally published in DAI Vol. 61, No. 7. Reprinted here with corrected author name.
展开▼
