東華大學圖書館 |

語系: 繁體中文

說明(常見問題)

回圖書館首頁

手機版館藏查詢

登入

回首頁

切換: 標籤 | MARC模式 | ISBD

Execution resource allocation for di...

Li, Haksun.

FindBook

Google Book

Amazon

博客來

Execution resource allocation for distributed real-time controllers.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Execution resource allocation for distributed real-time controllers./
作者:	Li, Haksun.
面頁冊數:	258 p.
附註:	Source: Dissertation Abstracts International, Volume: 65-06, Section: B, page: 3003.
Contained By:	Dissertation Abstracts International65-06B.
標題:	Computer Science. -
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3137874
ISBN:	049685044X

Execution resource allocation for distributed real-time controllers.
Li, Haksun.

Execution resource allocation for distributed real-time controllers. - 258 p.

Source: Dissertation Abstracts International, Volume: 65-06, Section: B, page: 3003.

Thesis (Ph.D.)--University of Michigan, 2004.

Automated agents operating in a real-time environment face the challenge of not having sufficient resources to execute their preferred control plans to monitor and react to all possible hazardous situations rapidly enough. In this dissertation, we develop computationally tractable algorithms that the agents can use to modify their preferred control plans to produce schedulable control plans which satisfy their local execution resource constraints, and still respond to the most probable hazards appropriately.

ISBN: 049685044XSubjects--Topical Terms:

626642
Computer Science.

Execution resource allocation for distributed real-time controllers.
LDR:03126nmm 2200313 4500 001 1846628
005 20051103093530.5
008 130614s2004 eng d
020 $a 049685044X
035 $a (UnM)AAI3137874
035 $a AAI3137874
040 $a UnM $c UnM
100 1 $a Li, Haksun. $3 1934733
245 1 0 $a Execution resource allocation for distributed real-time controllers.
300 $a 258 p.
500 $a Source: Dissertation Abstracts International, Volume: 65-06, Section: B, page: 3003.
500 $a Chairs: Edmund H. Durfee; Kang G. Shin.
502 $a Thesis (Ph.D.)--University of Michigan, 2004.
520 $a Automated agents operating in a real-time environment face the challenge of not having sufficient resources to execute their preferred control plans to monitor and react to all possible hazardous situations rapidly enough. In this dissertation, we develop computationally tractable algorithms that the agents can use to modify their preferred control plans to produce schedulable control plans which satisfy their local execution resource constraints, and still respond to the most probable hazards appropriately.
520 $a We prove that the agents can use certain information to efficiently transform their initially unschedulable plans into schedulable ones. Specifically, when a real-time agent in a multiagent environment is unschedulable, it will first analyze the cost-benefit tradeoffs of its resource usages to attempt to schedule a subset of the actions in its plan that still preempt all conceivable hazards. If this fails, the agent exchanges partial plans with other agents to discover and prune away the unnecessary actions that were planned due to ignorance about other agents' plans. As the agent now becomes more aware of the interactions among the agents, it may realize that some action decisions it has made are now suboptimal. The agent searches for local changes to its plan to correct these suboptimal decisions. If the agent still remains unschedulable, it can analyze its probabilistic temporal trajectory so it can drop the least likely needed actions repeatedly until its plan finally becomes schedulable.
520 $a We conclude that using our algorithms an agent, whether it is acting alone or in a multiagent environment, can construct a more accurate local model of the world to make resource allocation decisions and better assess their merits. Our experiments show that an agent can, on average, considerably improve its performance by using our techniques. The contributions in this thesis consist of a set of carefully characterized, analyzed, and evaluated data structures and algorithms for discovering, representing, and using new knowledge that a resource-limited real-time agent can efficiently apply to effectively improve resource allocation.
590 $a School code: 0127.
650 4 $a Computer Science. $3 626642
650 4 $a Artificial Intelligence. $3 769149
690 $a 0984
690 $a 0800
710 2 0 $a University of Michigan. $3 777416
773 0 $t Dissertation Abstracts International $g 65-06B.
790 1 0 $a Durfee, Edmund H., $e advisor
790 1 0 $a Shin, Kang G., $e advisor
790 $a 0127
791 $a Ph.D.
792 $a 2004
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3137874