Comparative Assessment of Update Methods Using Static Domain Constraints

The paper analyzes and compares methods for ramiication, or for update, which are based on minimization of change while respecting static domain constraints. We analyze how the choice of minimization strategy aaects the set of preferred new states. Besides the classical case of straight minimization, we analyze alternative cases where state variables ((uents) are partitioned into two or three kinds, such as occluded/unoccluded, persistent/nonpersistent, frame/nonframe, or released/nonreleased, and where minimization treats changes in these categories diierently. This includes the methods previously proposed by del Val and Shoham, and by Kartha and Lifschitz. We also discuss and analyze an additional technique, called changeset-partitioning, where minimization is done separately for each outcome of a nondeter-ministic action. We relate it to the previous methods, and argue that this new method is generally to be preferred over them.