§U²z«¬
³nÅé
¨t²Î
¤§
‹
»s»PÀ³
¥Î
3/3(Á`-pµe)
-p¹º
½s¸¹
: NSC 88-2213-E-002-004
°õ
¦æ
¤é
´Á
: 87/8/1 - 88/7/30
¥D«ù
¤H : ¶µ
¼ä
°ê
¥ß
¥xÆW¤j¾Ç
¸ê
°T¤u¾Ç
¨t±Â
hsiang@csie.ntu.edu.tw
¤@¡B ¤¤¤åºK-n ¡]ÃöÁäµü¡G§U²z«¬³nÅé¡B ¤H¤u´¼¼z¡B-Ó¤H¤Æ¡B¸ê°TÀ˯Á¡^ ¥»-p¹º¤§¥Ø¼Ð¬°³]-p¤@§U²z«¬³nÅ骺 µo®iÀô¹Ò¡A¥]§t«Ø¥ß¤@§U²z«¬³nÅéªÀ°Ï ( agent community ) ¤§µ²ºc ( architecture )¡B »y¨¥ ( agent task¡Aspecification language »P ontology) ¡BÅÞ¿è¡B±À²z¨t²Î (planning ¡A reasoning ¡Alearning ¡Aand personalization systems)¡A¤Î¹q¤l¹Ï®ÑÀ] (digital library)¡C¥» -p¹º¨Ã±N¦b¦¹Àô¹Ò¤U«Ø¥ß¼Æ-Ө㰪«×À³¥Î ©Êªº§U²z«¬³nÅé¡C ¥»-p¹º¥]¬A¤@-ÓÁ`Åé-p¹º¤Î¤--Ó¤l-p ¹º¡G (1) ¹ù¯Â¤¤¡NªL¤@ÄP §U²z«¬³nÅ餧ÅÞ¿è¨t²Î Logical systems for multi-agents (2) ³¯¤å¶i¤ä´©§U²z«¬³nÅ餧¹q¤l¹Ï®ÑÀ] Digital library for information agents (3) ³\¥Ã¯u
§U²z«¬³nÅ餧»s³y¤Î-Ó¤H¤Æ§Þ³N Implementation and personalization of software agents
(4) ¶µ¼ä
§U²z«¬³nÅéÀô¹Ò¤¤¬[ºc¤Î»y¨¥¤§¬ã»s Agent community architecture and agent task specification languages
(5) -b¦u·O
§U²z«¬³nÅ餤°ÊºA¥Ø¼Ð¾É¦V´¼¼z«¬³W¹º ¤§¬ãµo
Goal-oriented planning in dynamic agent environments
Agent community ¤Îagent ¥»¨-ªº architecture ±N¦b¤l-pµe¡]¥|¡^¤¤³]-p¡C Agent »Pagent ¶¡·¾³q©Ò¾D¹Jªºontology ¤è
-±ªº°ÝÃD«h¦b¤l-pµe¡]¤-¡^¤¤±´°Q¡CAgent ¤¤©Ò»Ý-nªºreasoning ¤Îplanning ³¡¥÷¥ç¦b ¤l-pµe¡]¤-¡^¤¤³]-p¡A¦Ó-«-nªº-Ó¤H¤Æ (personlization) ªº³¡¥÷«h¦b¤l-pµe¡]¤T¡^¤¤ ¬ã¨s¡CAgent µ{¦¡¼¶¼g©Ò»Ý-nªºtask specification language ¦b-pµe¡]¥|¡^¤¤³] -p¡C¦]¬°agents ¦³learning ªº¯à¤O¡Aagent µ{¦¡¥i¯à·|¸òµÛ¨Ï¥ÎªÌªº¨Ï¥Î¦Ó§ïÅÜ¡A¦] ¦¹³o-Ó»y¨¥ªº»y·N©Ò²o¯AªºÅÞ¿è¤ñ¤@¯ëªº µ{¦¡-n½ÆÂøªº¦h¡C³o¨ÇÅÞ¿è°ÝÃD¦b¤l-pµe ¡]¤@¡^¤¤±´°Q¡C¤l-pµe¡]¤T¡^¡B¡]¥|¡^»P ¡]¤-¡^¤¤·|³]-p¤@¨Ç¨ã°ª«×À³¥Î©Êªº agent ¡A¨ä¤¤³\¦h agents ©Ò»Ýªºdigital libraries ¦b¤l-pµe¡]¤G¡^¤¤¹ê²{¡C
¤G¡B-^¤åºK-n (Keywords: software agents,
digital library, artificial intelligence, software personalizaiton, information retrieval. )
We propose to study the design and applications of software agents. an agent is a goal-oriented application software capable of communication, learning, personalization, autonomous action, and can adapt to its environment. Unlike traditional software which is passive, agents actively learn,communicate and collaborate to satisfy users' need. Through the study of software agents, we also intend to reexamine how agent theory can change the fundamental philosophy of softwre design. In addition to therretical studies, we also plan to design and build prectical agents and apply them to information retrieval, office automation, network security, and manufactory automation. We hope that our project will have some inpact on the software industry in Taiwan.
One of the most important applications of agents is information retrieval. This has
become a critical issue since the emergence of Internet. an intelligent information retrieval agent, which knows the preference of each indicidual users, can be anindispensable aid to help them find precisely the information they want through the vast sea of information available on the Internet.
In the sub-projects to be described separately, we plan to study the architecture, language, logic, reasoning, design, and implementation issues of software agents. We will also build a digital libreary, which is to be used as the backbone for information retrieval.
¤T¡B-p¹º½t¥Ñ»P¥Øªº
¥»-p¹º¤§¥D¦®¬°µo®i-~²§©ó¶Ç²Î³nÅé Æ[©Àªº§U²z«¬³nÅé (software agent) §Þ³N¡A ¥H¦¹§Þ³N¶}µo·s¤@¥N³nÅé¨t²Î¡A¨Ã±N¹ê»Ú »s§@¼Æ-Ó agents¡A¬°¨ä°Ó·~¤Æ¥´¤U°ò¦¡C
ÄY®æ¨Ó»¡¡A software agent ¬O¤@-Ó¡u¨Ï ¥Î¡vµ{¦¡¤èªkªº·s·§©À¡C¦ý¬O¦]¬°¨Ï¥Î¤è ªkªº³Ð·s¡A¨Ï±oµ{¦¡¥»¨-»s§@¤W¤]»Ý-n·s ªº¤èªk¤Î·§©À¡C Software agent ¦³¥H¤U¤§ ¯S©Ê:
(1) ¥Ø¼Ð¾É¦V (mission-oriented)
¹ï¨Ï¥ÎªÌ¨Ó»¡¡Aagent ¥i¥H¹F¦¨¯S©w¥Ø ¼Ð (goal)¡A¨Ï¥ÎªÌ¥u-n§i¶D agent ¥Ø¼Ð¬°¦ó §Y¥i§¹¦¨¤u§@¡A¦Ó¤£»Ý»¡©ú¹F¨ì¥Ø¼Ðªº²Ó ¸`»P¨BÆJ¡C¦¬¨ì¨Ï¥ÎªÌ«ü©wªº¥Ø¼Ð«á¡A agent ¸g¥Ñ¥»¨-ªº±À²z¯à¤O¡A¥[¤W»P¨ä¥Lªº agent ¥æ´«°T®§¡A«K¯à¦Û°Êº¡¨¬¨Ï¥ÎªÌªº»Ý ¨D¡C
(2) ¨ã¾Ç²ß¯à¤O (adaptive & evolutionary)
Agent ¨ã¦³¤H¤u´¼¼z±À²z¤Î²³æªºÂk ¯Ç¯à¤O¡AÅý agent ¦b°õ¦æ¹Lµ{¤¤¡A¥i¥H
1. ¾Ç²ß¨Ï¥ÎªÌªº¾Þ§@²ßºD¡C 2. ¨ã³Æ¸û°ªªº®e§Ô«×¡A¯à°÷³B²z¤£§¹¾ã ©Î¦³Âø°Tªº¸ê®Æ¡C 3. ¦Û°Ê¨ú±o»P¥Ø¼Ð¬ÛÃöªº¸ê°T¡C (3) -Ó¤H¤Æ (personalized) ¥Ø«e¤j³¡¥÷ªº³nÅéµ{¦¡¡A¹ï¤£¦Pªº¨Ï ¥ÎªÌ¨Ó»¡¨Ã¨S¦³¥\¯à¤W©Î¾Þ§@¤Wªº¤£¦P¡A ¦ý agent «h¤£µM¡C¦b agent ªºÀô¹Ò¤U¡A¤@-Ó
¨Ï¥ÎªÌ¥i¥H¦Ûı¡]³z¹L option ªº³]©w¡^©Î ¤£¦Ûı¦a¡]Åý agent ¥»¨-¦Û§Ú¾Ç²ß¡^¨Ï agent Åܦ¨¬Û·í-Ó¤H¤Æªº¤u¨ã¡C»P²{¹ê¥Í¬¡ ¬Û¸û¡Aagent ´N¦p¦P¨p¤H¯µ®Ñ¤@¯ë¡A®É¶¡¤[ ¤F¦ÛµM´N¯à¼W¥[¹ï¤W¥qªº¤F¸Ñ¡C¹ï³nÅé¤u ·~¨Ó»¡¡A³o¼Ë¨ã¦³¾Ç²ß©M¸gÅç²Ö¿n¯à¤Oªº agent¡A±N§ó¯à°t¦X¨Ï¥ÎªÌªº°¾¦n»P²ßºD¡A ´£¨Ñ¤Íµ½¦Ó¦³®Ä²vªº¾Þ§@¬É-±¡C (4) ¦Ûµo©Ê (autonomous) ³o¬O agent §Þ³N»P¶Ç²Î³nÅéÆ[©À¬Û·í ¤£¦Pªº¤@ÂI¡C¶Ç²Î³nÅé¤j³¡¥÷³B©ó±µ¨ü«ü ¥Oªº³Q°Êª¬ºA¡A¨Ï¥ÎªÌ©¹©¹¥²¶·¦b³\¦h¥\ ¯à¤¤ºN¯Á¡B¹Á¸Õ¡A¤~¯à¹F¦¨¹w©wªº¤u§@¡C ¦Ó agent ¥i¦b¾Ç²ß¨Ï¥ÎªÌ²ßºD¤§«á¡A¥D°Ê ¹ï¨Ï¥ÎªÌ´£¥X«Øij©Î´£¨Ñ¬ÛÃö¸ê°T¡C¨ã¦³ ¦Ûµo©Êªº agent ±NÅý¤@¯ë¤H§ó®e©ö¨Ï¥Î¹q ¸£¡A¦]¬° agent ¥i¥HÀH®É´£¨Ñ¦UºØ¾Þ§@¸ê °T¥H¨Ñ¨Ï¥ÎªÌ°Ñ¦Ò¡C
(5) ¥i²¾°Ê©Ê (mobile)
¹ïºô¸ô¤è-±ªºÀ³¥Îµ{¦¡¨Ó»¡¡Aagent ±N ¨ã³Æ¥i²¾°Ê©Ê¡A¯à°÷¸g¥Ñºô¸ô¥D°Ê¨ì§O³B ´M§ä¸ê®Æ¡A¤]´N¬O»¡¡Aagent ¥i¥H³z¹Lºô¸ô ¨ì¹F»·ºÝ°õ¦æµ{¦¡¡C¦¹¶µ¯S©Ê°£¤F´£¨Ñ±j ¤jªº¥i²¾´Ó©Ê¥~¡A¤]Åý agent ªº°õ¦æ§ó¦³ ¼u©Ê¡C¦b¦¹¥i¯à¯A¤Îºô¸ô¦w¥þªº°ÝÃD¡A¦] ¦¹¡A¹ê»Ú¤W¥²¶·¦Ò¼{°õ¦æÅv--»P¦³®Ä½d³ò ªº³W½d¡C
(6) ¤¬¬Û¾Ç²ß©Ê (interactive)
Agent »P agent ¤§¶¡°£¤F¥i¥H·¾³qÁp ô¡A¥H¦@¦P¹F¦¨¯S©w¥Ø¼Ð¥~¡AÁÙ¥i¥H§Î¦¨ ¤¬¬Û¾Ç²ß¡B¥æ´«¨Ï¥Î¸ê°Tªº®ÄªG¡C¡]¥i¯à ¬O¥æ´«¦P¤@ºØ agent ¹ï¤£¦P¨Ï¥ÎªÌªº¸ê ®Æ¡A©Î¬O¥æ´«¤£¦P agent ¦b¤£¦Pªººô¸ô¤W ªº¸ê®Æ¡^¡A³oºØ¤¬¬Û¾Ç²ßªº¯S©Ê¡A±N´£¨Ñ ¤U¦CÀuÂI¡G
1. Agent ¥i¥H§ó¦³®Ä²v¦a¨ú±o©Ò-nªº¸ê °T¡C
2. Agent ¥i¥H²Ö¿n§O¤Hªº¸gÅç¡A¼W¶i¾Ç²ß ªº®Ä²v¤Î¯à¤O¡C
¤W-z¦U¶µ¯S©Ê¡A¦bÆ[©À¤W¹ï¶Ç²Î³nÅé ±N³y¦¨¬Û·í¤jªº½ÄÀ»¡CAgent -²©R©Êªº·§ ©À¡A°£¶}³Ð·s¿oªº¬ã¨s»â°ì¤§¥~¡A¤]´£¥X
³\¦h«E»Ý¸Ñ¨Mªº®Ú¥»°ÝÃD¡C¹ï©ó³o¨Ç°Ý ÃD¡A§Ú-̤u§@¤º®e¦³¡G
1. Agent µ{¦¡»y¨¥¡G
³o-Ó³¡¥÷¥D-nªº½ÒÃD¦b©ó¡GÅý¨Ï¥ÎªÌ »P agent¡Bagent »P agent ¤§¶¡¡A³£¯à¥H¦@³q ªº»y¨¥¶i¦æ¦³®Äªº·¾³q¡A¦P®É¡A»y¨¥¥» ¨-©Ò´£¨ÑªºÂX¥R©Ê¡A¥iÅý¥¼¨Ó·sªº agent ¯à °÷¥H Plug and Play ªº°µªk¡A»´©ö¦a¥[¤J¾ã -Ó²{¦³ªº agent ¨t²Î¤§¤¤¡C¹ï¦¹½ÒÃD¡A§Ú -̱N¤u§@¶µ¥Ø²Ó¤À¬°¤T:
(1) agent composition ¤§»y¨¥
(2)©w¸q agent specification ¤§¶¡·¾³qªº »y¨¥ (3)³]-p agent »P¨Ï¥ÎªÌ¤§¶¡ªº¤¶-± (interface) 2. Reasoning Engine : ¦³¤F²Ä¤@¶µ©Ò-zªº¨t²ÎÀô¹Ò»P·¾³q»y ¨¥¤§«á¡Aagent ³Ì-«-nªº³¡¥÷«h¥Ñ Reasoning Engine ¹F¦¨¡A¹ï agent ´£¨Ñ¾Ç ²ß¡B±À²z¡B¹w´úµ¥¦UºØ¤H¤u´¼¼z¤è-±ªº¥\ ¯à¡C¦b³o¤è-±»ÝµÛ¤â¸Ñ¨Mªº°ÝÃD¦³¡G
(1) relevance ªºÆ[©À (2) planning ªº¤èªk
(3) deductive ¤Î inductive reasoning ªº¯à ¤O¡C
3. Architecture :
-Ó§Oªº agent ¦¨§Î¤§«á¡A¦p¦ó¯Ç¤J²{ ¦³¨t²Î¡A¥²¶·±q¨t²Î¬[ºc¡]Architecture¡^ ¤è-±µÛ¤â³W¹º¡C¾ã-Ó agent ¬[ºc°£¤F´£¨Ñ agent ¹B§@ªº°ò¥»Àô¹Ò»P¨t²Î¸ê·½¤§¥~¡A³Ì -«-nªº¬Oºû«ù¨t²ÎªºÂX¥R©Ê»P©µ®i©Ê¡AÅý local ªº agent ¯à°÷«Ü®e©ö¦a¸ò·s¥[¤Jªº agent ¦@¦P¦X§@¡Aªº agent ÀH®É¥i³Q·sªº agent ´À´«¨ú¥N¡A¦Ó¹ïºô¸ô¤W remote ºÝªº agent¡A¤]¯à°÷¶¶§Q³q¹Lºô¸ô¥æ´«¬ÛÃö¸ê
°T¡A¹F¨ì agent »P agent ¤§¶¡¾Ç²ß¥æ ¬yªº¥Øªº¡C
4. Digital libraries agent ¤§ information agents
¬°¤F¨Ï agents ¯à¦³®Äªº¨Ï¥Î¡A¥¦-Ì»Ý -n¯à°÷Ą̊ú©Ò»Ý¤§¸ê°T¡C³Ì¦³®Äªº¤èªk¬°
«Ø¥ß¤@-Ó¹q¤l¹Ï®ÑÀ]¡A¨ä¤¤°£¤F¹q¤l¹Ï®Ñ ¥~¡A³Ì¥D-nªº³¡¤À¬°¾ã²z¸g¹L Internet ©Ò ¨ú±oªº¸ê®Æ¡C¦b¥»-p¹º¤¤¡A§Ú-̱N¹ï³o¨Ç °ÝÃD°µ¹ý©³¦Ó²`¤Jªº¬ã¨s¡C ,¥|¡B¬ã¨s¤èªk¤Î¦¨ªG ¥»-p¹º¤¤¡A±N¹ï³o¨Ç°ÝÃD§@¹ý©³¦Ó²` ¤Jªº¬ã¨s¡C (¤@) §U²z«¬³nÅéÅÞ¿è¨t²Î¡þ¹ù¯Â¤¤¡NªL¤@ ÄP
Logical systems for multi-agents
§Ú-̱´°Q agent ¬ÛÃöªºÅÞ¿è¨t²Î¡A±N ¤§°Ï¤À¬°Âk¯ÇÅÞ¿è¤ÎºtöÅÞ¿è¨â¤jÃþ¡A¨Ã ±N¤§¾ã¦X¦b¤@-ӲΤ@ªº¬[ºc¤U¡A³o-Ó¬[ºc °ò¥»¤W¥Ñ agent ªº«H©À¡N°¾¦n¤Î¦æ°Ê¤T¤j ³¡¤À²Õ¦¨¡A¦A¥[¤W®É¶¡¤Î agent ªº¯à¤Oµ¥ ¦]¯À¡A«ØºÙ¬°¢Ð¢Þ¢Ï¬[ºc (Belief,
Preference, Action)¡C¦b¦¹¤@¬[ºc¤U¡A§Ú-Ì ±Nµo®i agent ±À²z¡N¨Mµ¦¤Î¾Ç²ß¼Ò¦¡¡A¨Ã §Æ±æ¥Ñ¦¹´£¨Ñ agent »y¨¥¤§»y·N°ò¦¡C
(¤G) ¤ä´©§U²z«¬³nÅ餧¹q¤l¹Ï®ÑÀ]¡þ³¯¤å ¶i
Digital library for information agents
Internet ´£¨Ñ¤F¤j¶q««¤â¥i±oªº¸ê·½¡A ¦ý¤]¥O¨Ï¥ÎªÌÃø¥H§ä´M©Ò»Ý¸ê®Æ¡C¥Ø«e¦b Internet ¤WÁöµM¦³¦U¦¡¦U¼Ëªº¸ê·½·j´M¤u ¨ã¡A¦ý¥¦-̳£µLªk´£¨Ñ¨Ï¥ÎªÌ¤@-Pªº¤ÀÃþ ªÅ¶¡¡A¨Ã·|³y¦¨ºô¸ôªº¨H-«-t¾á¡C¼Æ¦ì¹Ï ®ÑÀ]¬O±N¼Æ¦ì¸ê®Æ¤ÀÃþ³B²z¡A¨Ã´£¨Ñ¦UºØ ¤u¨ã¡A¤w¥ß¸ê°T¥æ¬y»P·j´MªºÀô¹Ò»P¬[ ºc¡C¦b³o-Ó¬[ºc¤U¡A¸ê°T¥i¥H§Ö³t¦Ó¦³®Ä ªº¬y³q¡A¶i¦Ó¥[³t§Þ³Nªºµo®i»P²£¹ïªº¤É ¯Å¡A¥i»¡¬O°ê®a¸ê°T°ò¦«Ø³]ªºÃöÁä§Þ ³N¡C
¥»-p¹º±N³]-p¨Ã¹ê§@¤@-ӼƦì¹Ï®ÑÀ] ªº¬[ºc¡A¥H¤ä´©§U²z«¬³nÅé¡A¦¨¬°¸ê°TÂ^ ¨ú§U²zªº°©·F¡A¨Ï¨ä¯à°÷¹L¤@-Pªº¬É-±¡A Â^¨ú¦UÃþªººô¸ô¸ê·½¡C³o-Ó¬[ºc¬O¥H¦h-Ó ¤À´²¦¡ªº¥N²zªÌ¬°¥D¡A¾ã¦X´¼¼z«¬§U²z¤Î ¥~¬Éªº¸ê®Æ®w¡Aºc¦¨¤@§¹¾ãªºÀô¹Ò¡C¥N²z ªÌ¥Ñ¬d¸ßºÞ²z-û¡N®Ñ¥Ø¸ê®Æ¦øªA¾¹¡N¤º®e µù¥U³B¤Î¸ô¥Ñ¾¹µ¥¥|¤j¼Ò²Õ²Õ¦¨¡A¬O§Ú-Ì ¬ã¨sªº-«ÂI¡C¨Ï¥ÎªÌ¥i³z¹L´¼¼z«¬§U²z¡A
¥H¹Ï§Î¨Ï¥ÎªÌ¤¶-±»P¤¬°Ê¼Ò¦¡©M¥N²zªÌ·¾ ³q¡A¥HÀò±o©Ò»Ý¤§¸ê°T¡C§@ªÌ©Î¹Ï®ÑÀ]-û «h³z¹L®Ñ°Ý¸ê®ÆªA¾¹¿é¤H°ª«~½èªº®Ñ¥Ø¸ê ®Æ¡A©Î±N¸ê®Æª½±µ¿é¤J¨ä±MÄݪº¸ê®Æ®w¨t ²Î¤¤¡CµL½×¹ï¨Ï¥ÎªÌ©Î§@ªÌ¦Ó¨¥¡A³£¥u¨£ ¨ì³æ¤@ªº¬É-±»P³æ¤@ªº¸ê°T¤ÀÃþ¡AµL¶·¦Ò ¶q¸ê®Æ©Ò¦bªº¦ì¸m¡A¤Î¤£¦Pªº¬d¸ß¤è¦¡»P ³q°T¨ó©w¡C§Ú-Ì»{¬°¦¹¬[ºc±N¨ã¦³©ö©ó¨Ï ¥Î¡N¥iÂX¥R¡N°ª®Ä¯àµ¥©Ê½è¡C¦³Ãö¸ê°TÂ^ ¨ú§U²z³¡¤À½Ð¨£¾ã¦X«¬-p¹º¡u§U²z«¬³nÅé ¨t²Î¤§¬ã¨î»PÀ³¥Î¡v¡C (¤T) §U²z«¬³nÅ餧»s³y¤Î-Ó¤H¤Æ§Þ³N¡þ³\ ¥Ã¯u
Implementation and personalization of software agents ÀHµÛ³nµwÅ骺¶i¨B¡A-Ó¤H¹q¸£»P¸ê°T ºô¸ô¥H¦¨¬°¤H-̤é±`¥Í¬¡¤¤-«-nªº¤@Àô¡C µM¦Ó·í¤j³¡¤À¹q¸£¨Ï¥ÎªÌ±ý¾Þ§@¬Æ©Î¸Ë³] ¥ô¦ó¤@¶µ·sªº³nµwÅé¨t²Î®É¡A¤´»Ý-n¸g¹L ¬Û·íªº¾Ç²ß¶¥¬q«á¤~¯à¶¶§Qªº¼ô±x¨ä¬ÛÃö «ü¥O¤w§¹¦¨¨ä¤u§@¡C¹ï©ó¤@-Ó«D¹q¸£±M·~ ¤§¨Ï¥ÎªÌ¦Ó¨¥¡A¦¹¶¥¬q©¹©¹¹L©óÁ}Ãø¦Ó¨Ï ¤§±æ¦Ó«o¨B¡F¹ï©ó¤@±M·~¨Ï¥ÎªÌ¦Ó¨¥¡A¦¹ ¶¥¬q¥ç©¹©¹¯Ó®É¶O¨Æ¦Ó¨Ï¤§-°§C¤u§@®Ä ²v¡C§U²z«¬³nÅé±N´£¨Ñ¹q¸£µ{¦¡³]-p»P¨Ï ¥Î¤@-ÓºIµM¤£¦Pªº¼Ò¦¡¡C¥»¤l-p¹º¦®¦b¬ã ¨s¶}µo§U²z«¬³nÅ餧»s§@»P-Ó¤H¹ºªº§Þ ³N¡A¥H²¤Æ¨Ï¥ÎªÌ¤§¤u§@-t¾á¡A¶i¦Ó¼W¶i ¨ä¤u§@®Ä²v¡C
¥»-p¹º±Ä¥Ñ¤U¦Ó¤W (bottom-up) ¤§µo®i ¼Ò¦¡¡A°t¦X¨ä¥L¤l-p¹º¤§¦¨ªG¡A³]-p¼Æ-Ó ¨ó§U¨t²Î¨Ï¥Î»PºÞ²z¤§§U²z«¬³nÅé¡C¤×¨ä ±N¯S§O¬ã¨s¦p¦ó¨Ï³nÅéÀHµÛ¦U¨Ï¥ÎªÌ¤§»Ý ¨D»P³ß¦n¦Ó§ïµ½¨ä¥\¯à¡A¶}µo-Ó¤H¤Æ§U²z «¬³nÅ餧°ò¥»§Þ³N¡C¥»-p¹ºÀÀ°õ¦æ¤§¤u§@ ¥i¤À¬°°ò¥»§Þ³N»P¹ê§@¨t²Î¨â¤è-±¡G 1. ¬ã¨sµo®i§U²z«¬³nÅ餧°ò¥»§Þ³N¡A¬ã¨s -«ÂI¥]§t¤U¦C¶µ¥Ø¡G
2. µo®i¾A·í¤§ª¾ÃѪí¹F¤è¦¡¡A¥H¾ã²zÀx¦s ¥ô°È¬ÛÃö¤§¸ê°T¡C 3. §Q¥Î¤l-p¹º¤@¡N¤-©Òµo®i¥X¤§±À²z¤Î³W ¹º¤èªk¡A¥H¶¶§Q¹F¦¨¥ô°È¡C 4. ³]-pÆ[¹î»P·j¶°¨Ï¥ÎªÌ¤u§@²ßºD©ÎÀu¥ý ¶¶§Ç¤§¤u¨ã¡C 5. ¬ãµo¾÷¾¹¾Ç²ß¤èªk¡A¤ÀªR©Ò·j¶°¤§¸ê®Æ ¨Ï¤§¦¨¬°¦³¥Îªºª¾ÃÑ¡A¥H§@¬°§U²z«¬³n Åé-Ó¤H¸Ü»P§ï¶i¨ä¤u§@¦¨®Ä¤§¨Ì¾Ú¡C ¹ê»Ú»s§@¼Æ-Ó§U²z«¬³nÅéµ{¦¡¡AÂÇ ¥H´ú¸ÕÅçÃÒ©Ò´£¥X§Þ³N¤§¥i¦æ©Ê¡A¨Ã¥B¶i ¤@¨Bµo±¸¹ê²{§U²z«¬³nÅé§Þ³N¤W¤§¬D¾Ô¡C ¥»-p¹ºÀÀ¬ã»s¤U¦C¼Æ-Ó agents¡G
1. ¸ê·½¤À°t printer agent ---¦Û°Ê¶i¦æÀÉ®× ªºÂà´«¡B¶Ç¿é¡A¨Ã¨Ì¨Ï¥ÎªÌ¤§°¾¦n¤ÎÀÉ ®×¯S©Ê¿ï¾Üºô¸ô¤W³Ì¦X¾A¤§¦Lªí¾÷¨Ó¦C ¦L¨Ï¥ÎªÌ¤§ÀɮסC
2. ¨t²ÎºÞ²z system information agent ---¨ó§U ¸Ñ¨M¨t²ÎºÞ²z»P³nµwÅé¦w¸Ë¤§ÃøÃD¡A´£ ¨Ñ¬ÛÃö¤§FAQ ÀÉ®×, on-line manuals,¤Î°Q ½×¸s¡C
¤@¥¹§¹¦¨³o¨Ç agents «á¡A§Ú-̱N§â»s §@ agent ªº§Þ³N¤@¯ë¤Æ¡A¥H©w¸q³]-p§U²z «¬³nÅé¨t²Î¤§°ò¥»¤èªk¡C©ó²Ä¤T¦~¤¤¡A¥» -p¹º±N§Q¥Î¤l-p¹º¤Gµo®i¤§¹q¤l¹Ï®ÑÀ]¨Ó Àx¦s»P¨t²ÎºÞ²z¬ÛÃö¤§¤åÄm»P¸ê®Æ¡A¨Ã±N ©Ò¦³ agents ¾ã¦X©ó¤l-p¹º¥|©Ò©w¤§ agent community ¬[ºc¤¤¡AÅý¥¦-̤¬¬Û·¾³q¥H¦@ ¦P´£¨Ñ¨Ï¥ÎªÌ¤@-Ó²©ö¥B¥\¯à±j¤j¤§¨t²Î Àô¹Ò¡C¥t¥~¡A§Ú-̱N¥H¤l-pµe¤@ªº²z½×¼Ò ¦¡¨ÓÅçÃÒ¦U agent ªº¥¿½T©Ê¡C (¥|)§U²z«¬³nÅéÀô¹Ò¤¤¬[ºc¤Î»y¨¥¤§¬ã»s ¡þ¶µ¼ä
Agent community architecture and agent task specification languages
¥»¤l-p¹º¤§¥D-n¥Øªº¦³¤T¡G
l «Ø¥ß¤@§U²z«¬³nÅé¸sÅ骺¨t²Î¬[ºc (agent
community architecture)¡C
l ³]-p¤@¯à°÷°õ¦æ¡A¨Ã¥i¦Û°Ê evolve ªº
agent task specification »y¨¥¡C
l ³]-p¤@¸ê°T·j´M³nÅé (information retrieval
agent)¡C
²{¹ï¤W¤T³¡¥÷¤À-z¤§¡G
(1)¦b¤@ agent ¸sÅ餤¡A¨C-Ó agent °£¤F¥i ¥H¾Ç²ß¨Ï¥ÎªÌªº²ß©Ê¤§¥~¡A¨Ã-n¾Ç²ß¤@ ¸s¨Ï¥ÎªÌªº¦@¦P²ß©Ê¡CªÌ¨â¼hªº¾Ç²ß¦³ ¥¦-̪º¤£¦P³B¡C¦]¦¹¡A§Ú-̱N¤@-Ó agent ¤À¦¨¨â-Ó³¡¤À¡Gtask subagent »P
personal subagent ¡CPersonal subagent -t ³d-Ó§O¨Ï¥ÎªÌªº»Ý¨D¡CTask subagent «h -t³d¤@¸s¨Ï¥ÎªÌªº¦@¦P»Ý¨D¤Î¹ï¥~ªº¥æ ³q¡CTask subagent ¤§¶¡¦³¤@ managent -t ³d¥¦-Ì©¼¦¹¶¡ªºÁpµ¸¡C Personal subagent ¤§¶¡¥ç¦³¤@ organizer -t³d¥¦-Ì ¶¡ªºÁpµ¸¡C Organizer ¥»¨-¥ç¬°¨Ï¥ÎªÌ ªº interface agent ¡A-t³d«D±MÄÝ¥ô¦ó agent ¤§¨ä¥Lªº user ¦æ¬°¡Cagent (task subagent & personal subagent) ¥»¨-ªº¬[ ºc¦b-p¹º¤º®e¤¤¦³¸Ô²Ó»¡©ú¡C
(2) ¤@-Ó agent À³§@ªº¬O¥i¥Î¤@ task specification language ªí¹F¥X¨Ó¡C§Ú-Ì»Ý -nªº task specification language ¤£¦ý-n¬O executable ¡A¦Ó¥B-n¥R¤À high level ©M declarative ¡C³o¼Ëªºµ{¦¡¥i¦b¸g¹L reasoning¡]¥Î§Ú-̳]-pªº program evolve) «á²£¥Í autonomous actions ¡A¨Ã¼W¥[ agent ¹ï¨Ï¥ÎªÌªº¤F¸Ñ¡C
(3)¸ê°TĄ̊ú¤Î¤ÀªR¬O³Ì-«-nªº agent ¥\¯à ¤§¤@¡A§Ú-̳]-p¤@¯à search , navigate, personalize, auto-notify ªº¸ê°TĄ̊ú agent¡C¨Ã§Q¥Î¥¦¹ïuser group ¯S©w¤§¿³ ½ì«Ø¥ß¤pªÀ°Ï¹q¤l¹Ï®ÑÀ]¡C
(¤-)§U²z«¬³nÅ餤°ÊºA¥Ø¼Ð¾É¦V´¼¼z«¬³W ¹º¤§¬ãµo¡þ-b¦u·O
Goal oriented planning in dynamic agent environments
¤@¯ëagent-based system °ò¥»¤WÀ³¨ã¦³ ¥H¤U¯S©Ê¡G¦Ûµo©Ê (autonomy) ¡A¤ÏÀ³©Ê (reactivity)¡A·¾³q©Ê (social ability)¡A¤Î¥Ø¼Ð ¾É¦V (proactiveness)¡C´N¦p¦P¤@-Ó±q¨ÆªÑ²¼ ¤ÀªR¾Þ§@ªº¹Î¶¤¤¤¦U-Ó¶¤-û (human agent) ¶¡¦³µÛ¤£¦Pªº¤À¤u¡G¦³ªº¤H-t³dÂ^¨ú³Ì·s §Y®É¬ÛÃö¤§ªÑ²¼°T®§¡A¦³ªº¤H-t³d¤ÀªR°T ®§¡A«Ø¥ß-קï¤ÀªR¼Ò«¬¡C¦U-Ó¶¤-û§¡¯à¿W ¥ß¦a¦Ûµo©Ê¦a«ä¦Ò¤ÏÀ³¡A¦P®É¤]¤¬¬Û·¾³q ¨ó½Õ¦@¦P¦a°l¨D³Ì¨ÎªÑ²¼§ë¸ê¤§¨Mµ¦¡C¦Ó ¦p¦ó³]-pµo®i¤@-Ó³nÅé¡A¨Ï¨ä¯à°÷¦³®Ä¦a ¨Ó¹B§@¥B¦P®É¨ã³Æ¦³¤W-z agent ¤§¯S©Ê ©O¡H·N§Y¤°»ò¼Ëªº agent ¤º³¡µ²ºc (agent architecture) ¯à¦³®Ä¦a¨Ó¹F¨ì¤W-z agent ¤§ ¯S©Ê©O¡H
¥»¤l¬ã¨s-p¹º¹w¦ô¥H¤T¦~®É¶¡¨Ó³]-p µo®i¥X¤@-Óµ²¦X¥H¥Ø¼Ð¾É¦V´¼¼z«¬³W¹º (goal-oriented planning) ¬°°ò¦¤§ planning task agent »P«eºÝ¤§§Y®É¦³®Ä·j´MInternet ¸ê°T¤§ integrated information agent ¤§ agent architecture¡A¥]¬A¨ä±À²z¾÷¨î¤Î¨äµo®iÀô ¹Ò¡C¦¹ agent architecture °£¤F¨ã¦³¤W-z agent ¤§¤@¯ë¯S©Ê¥~¡F¥t¥~¨ä¥ç¹Á¸Õ¸Ñ¨M¨â -Ó©|«Ý¸Ñ¨M¤§°ÝÃD¡G(1) ·í agent ¥\¯à«D±` ±j®É¡]¥ç§Y¡A¥¦¯à¹F¦¨¥ô¦ó¨Ï¥ÎªÌ¤§¥ô¦ó ¥Ø¼Ð¡^«h¥²¶·ª`·N¨ì¨ä©Ò±À²z¥X¤§°õ¦æ¤u §@²Ó¸`»P¨BÆJ¥²¶·¬O¦X²zªº( acceptable solution) ¡]¦¹¤D«ü·í¹F¦¨¨Ï¥ÎªÌ¥Ø¼Ð¤§¤â ¬q¦³¦h¼Æ-Ó¡A¦ý¦³¨Ç¤â¬q¬O¤£¦X²z¤§±¡ªp ®É¡^¡C¥ç§Y agent ¥²»Ý¨ã¦³¦w¥þ©Ê¡F¦P®É ¨ä¤£·|¦]¬°ª`·N¨ì¦w¥þ©Ê¦Ó©¿²¤¤F±À²z¤§ ®Ä²v¡C(2) ¦b multi-agent ¤§¬[ºc¤¤¡A agent ¶¡¥²»Ý¬Û¤¬¦X§@¥H§¹¦¨¨Ï¥ÎªÌ¤§»Ý¨D¡C¦] ¦¹¥²¶·¯à°÷±N agent architecture ¤§¥\¯à©µ ¦ù¦Ü¤À´²©ÊÀô¹Ò¸Ì (distributed environment)¡C
¬°¨Ï¥»-pµe©Òµo®i§Þ³N¤£¸¨©ó¤£¹ê¥Î ¤§¦a¨B¡A¥»-pµe¥H¡u¤¤µØ§ë¸ê«H°U¤½¥q¡v ¨ä¬ã¨sµo®i³¡ªù¤ºªÑ²¼°òª÷¤ÀªR¾Þ§@¤§¹Î ¶¤¤u§@¬°®×¨Ò¡]¦]¬°ªÑ²¼°òª÷¤ÀªR¨Mµ¦¸g ±`²o¯A¨ì§Y®É°ÊºA¤§¸ê°T¡^¡C´Á±æ¥»-pµe ¤§ agent reasoning engine ¯à°÷²£¥Í¦U-Ó¤£ ¦P¤À¤uªº task agents ¡C¥¦-̤]¯à°÷¹³ human agents ¤@¼Ë¯à°÷¤¬¬Û·¾³q¡B¨ó½Õ¡A ¦@¦P¦a¨Ó¨ó§U°l¨D³Ì¨ÎªÑ²¼°òª÷§ë¸ê¤§¨M µ¦¡C ¤-¡B°Q½×»Pµ²½× ¥»-p¹º¬°°ê¤º²Ä¤@-Ó§U²z«¬³nÅé¬ã¨s ªº¤j«¬¸sÅé-p¹º¡C¤T¦~«e²Ö¿n¤F³\¦h¬ã¨s ¸gÅç»P¦¨ªG¡A¨Ã¦b°ê¤ºÁ|¿ìAgent Technology Workshop¡A¨C¦~§¡¦³¤W¦Ê¤H°Ñ ¥[¡A²{¤w¶i¤J²Ä¤T©¡¡A°£¤F¬ã¨s¥»¨-¥~¡A ¹ï°ê¤º§U²z«¬³nÅé¬ã¨sªº±À°Ê¤]¦³©Ò°^ Äm¡C §U²z«¬³nÅé²{¦b¤w¦¨¬°ºô»Úºô¸ô¤WÀ³ ¥Î³nÅé·§©Àªº¥D¬y¡A¦b©¹«áªºµo®i·|§ó¥[ ½´«k¡C¥»-p¹º¬°°ê¤ºµo®i¥´¤U¤@¨Çªì¨Bªº °ò¦¡A¨Ã´£¿ô°ê¤º¬ã¨sªÌ¹ï³o-Ó¥@ÁͶժº
ª`·N¡A§¹º¡ªº¹F¦¨¹w©w¥Ø¼Ð¡C
¤»
¡B°Ñ
¦Ò
¤å
Äm
[1] Y. Shoham, Agent-Oriented Programming, AI, 60(1):51-92, 1993.
[2] C. J. Liau , On the possibility theory-based semantics for logics of preference, International Journal of Approximate Reasoning , 20(2), 173-190,1999.
[3] C. J. Liau, A logic for reasoning about action, preference, and commitment, ECAI'98, pp.552-556, 1998.
[4] C. J. Liau, A semantics for logics of preference based on possibility theory, Proc. of 7th International Fuzzy Systems Association World Congress, Vol I, pp.243-248,1997.
[5] C. J. Liau, A logic for reasoning about fuzzy truth values, IEEE-FUZZ'98, pp.903-908, 1998.
[6] C. J. Liau, Possibilistic residuated implication logics with applications, International Journal of Uncertainty, Fuzziness and Knowledge-based Systems, 6(4), 365-385,1998.
[7] C. J. Liau, On rough terminological logics, 4th International Workshop on Rough Sets, Fuzzy Sets, and Machine Discovery, pp. 47-54,1996
[8] C. J. Liau , On rough quantifiers, Cybenetics and Systems'98, pp.175-180,1998.
[9] C. J. Liau, Modal reasoning and rough set theory, Proc. of 7th International Conference on Artificial Intelligence: Methodology, Systems, and Applications, pp. 317-330, LNAI 1480, 1998.
[10] C. J. Liau and D.R. Liu, A logical approach to fuzzy data analysis, Proc. of 3rd European Conference on Principles and Practice of Knowledge Discovery in Databases, 1999. [11] §f«CÂE, ¾ðª¬°t¹ïªk©óHTML ¤å¥ó¸ê °TµÑ¨ú¤§¬ã»s»PÀ³¥Î. ºÓ¤h½×¤å, °ê¥ß ¥xÆW¤j¾Ç¸ê°T¤uµ{¾Ç¬ã¨s©Ò. ¤¤µØ¥Á°ê ¤K¤Q¤C¦~¤»¤ë [12] §õ¨B°·, °Ï°ìºô¸ô¤§¥N²z«¬µ{¦¡»PªA°È ¦@¨É¬[ºc. ºÓ¤h½×¤å, °ê¥ß¥xÆW¤j¾Ç¸ê °T¤uµ{¾Ç¬ã¨s©Ò. ¤¤µØ¥Á°ê¤K¤Q¤C¦~¤» ¤ë [13] ©ö¤åÃü, ¾ðª¬HTML ¤å¥ó¤§¸ê°TÂ^¨ú. ºÓ¤h½×¤å, °ê¥ß¥xÆW¤j¾Ç¸ê°T¤uµ{¾Ç¬ã ¨s©Ò. ¤¤µØ¥Á°ê¤K¤Q¤»¦~¤»¤ë [14] ½²©v¿«, ¾÷¾¹¤ÀÃþ¾Ç²ß§Þ³N©ó-Ó¤H¤Æ¹q ¤l¶l¥ó§U²z¤§¬ã¨s. ºÓ¤h½×¤å, °ê¥ß¥x ÆW¤j¾Ç¸ê°T¤uµ{¾Ç¬ã¨s©Ò. ¤¤µØ¥Á°ê¤K ¤Q¤K¦~¤»¤ë [15] ¾H¦t¶¯, -Ó¤H¤Æ¹q¤l¶l¥ó§U²z. ºÓ¤h½× ¤å, °ê¥ß¥xÆW¤j¾Ç¸ê°T¤uµ{¾Ç¬ã¨s©Ò. ¤¤µØ¥Á°ê¤K¤Q¤C¦~¤»¤ë [16] Àd§J©ú, «Øºc©ó¤À´²¦¡ª«¥ó¤Wªº¦h-«§U ²zµ{¦¡¦X§@¬[ºc. ºÓ¤h½×¤å, °ê¥ß¥xÆW ¤j¾Ç¸ê°T¤uµ{¾Ç¬ã¨s©Ò. ¤¤µØ¥Á°ê¤K¤Q ¤K¦~¤»¤ë [17] ²»Êºû, À³¥Î©ó¤À´²¦¡ºô¸ôÀô¹Ò¤§´¼¼z «¬¦Lªí¾÷¥N²zµ{¦¡. ºÓ¤h½×¤å, °ê¥ß¥x ÆW¤j¾Ç¸ê°T¤uµ{¾Ç¬ã¨s©Ò. ¤¤µØ¥Á°ê¤K ¤Q¤»¦~¤»¤ë
[18] Hsu, J. Y. J and Yih, W.-T. Template-based information mining from HTML documents. In Proceedings of AAAI-97, pages 256-262, 1997.
[19] Hsu,J. Y. J. A multi-agent framework for intranet service integration. In Toru Ishida, editor, Proceedings of Pacific Rim International Workshop on Multi-Agents, pages 26-37, November 1998.
[20] Cowie, J. and Lehnert, W. Information extraction. Communications of the ACM, 39¡]1¡^:80-91, 1996.
[21] Hao, X., Wang, J. T. L., and Ng, P. A. Information extraction from the structured part of office documents. Information Sciences, 91:245-274, 1996
[22] Kushmerick, N., Weld, D., and Doorenbos, R. Wrapper induction for information extraction. In Proceedings of Fifteenth International Joint Conference on Artificial Intelligence ¡]IJCAI-97 ¡^, pages 729-735, 1997.
[23] Bradshaw, J. M. editor. Software Agents. AAAI/MIT Press, Menlo Park, CA, 1997. [24] Sycara, K. P. Multiagent systems. AI
