Being effective when suspending goals

作者: Garrett A. Pelton , Jill Fain Lehman

DOI:

关键词:

摘要: In their everyday lives people flexibly handle manytasks. They work on one at a time, make quick de-cisions about which to on, and put asidetasks whenever attending them is not required toachieve the task’s goal. This last capability crit-ical because rather than fixate blocked task,a person can some other task. For exam-ple, making bean soup wouldn’t watch thebeans as they soak overnight. Instead she wouldtake inability affect progress of makingbean an opportunity other,possibly less important task where possi-ble. When taking advantage these opportunities,people don’t completely forget what were do-ing. Instead, put-aside tasks guide selectionof new tasks. After beans have soaked andthe simmering, might go out getother chores done, but probably would for-get eat out. By putting aside blockedtask remembering it appropriate aperson complete many still ac-complish in timely manner.This paper describes how artificial agent, calledLaureli, maintain goal-directed behavior whilesuspending her similarly that de-scribed above. Laureli serves exemplar agent,grounding detailed understandable exam-ples problems occur when suspendsher for later reactivation. creating Lau-tell, this extends [Hammond, 1989;Pryor, 1994] done recognize blockedso be until possi-ble again. Laureli’s suspension-reactivation mecha-nisms provide interleaving more available tasks"This research was supported part by grant fromMartin Marietta Wright Laboratory,Aeronautical Systems Center, Air Force Materiel Com-mand, USAF, ARPA under F33615-93-1-1330.The views conclusions contained documentare those authors should interpretedas necessarily representing official policies or endorse-ments, either express implied, Martin orthe United States Government.during slack time (while blocked).A task’s, availability defined metric howlikely agent expects working will makeprogress toward A availabilitychanges over depends both agent’sactions input from environment. Laureli’ssuspension-reactivation mechanisms are methodfor large changes availabilityover time.Representing isimportant, then better sched-ule its executes them. If avail-ability known advance, agentcan use knowledge generate schedule thatcan simply followed execution time. However,in cases doesn’t know itis difficult willchange The second step beansoup, "boiling hour", such anexample. Perhaps could measure water,look up specific heat usingthe effective transfer stove, calculatehow long boiling. However, also justput pan water stay inthe area, occasionally check see if waterwas advocates ap-proach, so monitor executionrather attempt toschedule task, theagent all knowledge.Laureli’s suspending reactivating tasksis similar deliberately excluding suspendedtasks decisions action take.However, based smaller num-ber functionally different simi-lar consider Since thesedecisions agent’s external behavior, sus-pending apparent ratio-nality achieving goals. Birnbaum [Birnbaum,1986] issues maintainingan rationality Us-ing assumptions Birnbaum, Lau-reli often act with ac-cess tasks, access the117

参考文章(12)
Louise Margaret Pryor, Opportunities and planning in an unpredictable world Northwestern University. ,(1994)
P. S. Rosenbloom, J. E. Laird, Integrating execution, planning, and learning in Soar for external environments The Soar papers (vol. II). pp. 1036- 1043 ,(1993)
Robert B. Doorenbos, Combining left and right unlinking for matching a large number of learned rules national conference on artificial intelligence. pp. 451- 458 ,(1994) , 10.21236/ADA278979
Daniel R. Kuokka, The deliberative integration of planning, execution, and learning Carnegie Mellon University. ,(1990)
Lawrence Albert Birnbaum, Integrated processing in planning and understanding Unknown Publisher. ,(1986)
Stuart Russell, Rationality and intelligence international joint conference on artificial intelligence. pp. 950- 957 ,(1995)
Michael E. Bratman, David J. Israel, Martha E. Pollack, Plans and resource-bounded practical reasoning computational intelligence. ,vol. 4, pp. 349- 355 ,(1988) , 10.1111/J.1467-8640.1988.TB00284.X
John Robert Anderson, Rules of the Mind ,(1993)