...Weld
We thank Denise Draper, Steve Hanks, Terrance Goan, Nick Kushmerick, Neal Lesh, Rich Segal, and Mike Williamson for helpful discussions. This research was funded in part by Office of Naval Research Grants 90-J-1904 and 92-J-1946, and by National Science Foundation Grants IRI-8957302, IRI-9211045, and IRI-9357772. Golden is supported in part by a UniForum Research Award.

...planner
XII stands for ``eXecution and Incomplete Information.''

...,
Our thanks to Anthony Barrett and Daniel Weld at the University of Washington for providing us with the code for the UCPOP planner.

...planners.
In fact, our agent relaxes this assumption by associating expiration times with beliefs in and and by recovering from errors that result from incorrect information. However, a discussion of this mechanism is beyond the scope of this paper.

...LCW.
XII operator schemata explicitly distinguish between causal effects (that change the state of the external world) and observational effects (that only change the state of XII 's model) as explained in [4].

...details.
Note also that the classical universal base mechanism requires that a type's universe be static and finite. XII correctly handles dynamic universes. Furthermore, XII's policy of linking to effects handles infinite universes, but this is not of practical import.

...confrontation.
The first two techniques order the threatening action before the link's producer or after its consumer. Confrontation works when the threatening effect is conditional; the link is protected by subgoaling on the negation of the threat's antecedent [14].

...Q(foo)R(foo)
Note the difference between shrinking and protecting a link (section 3.1). Unlike the link case, shrinking does not have a disjunct corresponding to S(foo).

...redundant.
We cannot simply associate exactly one sensory action with each goal, a priori, because the agent may fail to satisfy that action's preconditions - in which case trying a different sensory action is warranted.

...set.
Since XII can subsequently lose LCW due to information loss or domain growth (Section 3.2), it has to record this pruning decision and recompute the options for once LCW(G) is lost. Doing this in an efficient but sound manner is actually quite complex - see [8] for the details.

bburnard@isx.com
Wed Feb 16 09:48:57 PST 1994