Gregg Collins

Learn More
A planner in the real world must be able to handle uncertainty. It must be able to reason about the eeect of uncertainty on its plans, select plans that avoid uncertain outcomes when possible, and make contingency plans against diierent possible outcomes when uncertainty cannot be avoided. We have constructed such a planner, Cassandra, which has these(More)
A fundamental assumption made by classical AI planners is that there is no uncertainty in the world: the planner has full knowledge of the conditions under which the plan will be executed and the outcome of every action is fully predictable. These planners cannot therefore construct contingency plans, i.e., plans in which di erent actions are performed in(More)
One of the most common modifications made to the standard STRIPS action representation is the inclusion of filter conditions. A key function of such filter conditions is to distinguish between operators that represent different context-dependent effects for the same action. We consider how filter conditions may be used to provide this functionality in a(More)
1 Chicken and water chestnuts A case-based planner devises new plans by retrieving and adapting old ones from memory (see, e. Ideally, the plan it retrieves from memory should be one that requires the least adaptation to t the current circumstances. Finding this plan in a cost-eeective way is thus a key issue in case-based reasoning (see, e. The approach(More)
Human agents typically evolve a set of standard routines for carrying out often-repeated tasks. These routines effectively compile knowledge about how to carry out sets of interacting tasks without causing harmful interference. By modifying its routines in response to observed failures, an agent can refine enlarge and refine its planning repertoire over(More)
We have developed a model-based approach to learning from plan failures in which an agent uses a model of itself to determine where in its planning or execution the cause of a failure lies. We believe that such an approach constitutes the most promising basis for developing learning models that are capable of decid ing for themselves what needs to be(More)