成果報告書
Contents
1.2.
3.
4.
5. fod DIC
6. Supervised learning
7. AC DC 16
8. KOL e WWW FAX
9. 60
10. 15
11. EPSON NT 2700 Foe Ethernet 1 OFF
12. a7 Ne ae 2 2
13. 1 0
14. 1 0 0 0 20000 1 5 0
15. 8 200 5 10 12 13 14 33 5 ate CS 2 vyt
16. 38 2 17 18 1 0 0 ae 3 0 0 0
17. 41 1 N HK
18. 2 2 1 boll oma MAHON R ill
19. RNN SUG
20. E 0
21. 7 pn 32 8 1 0 4
22. 4 2005 12 20006 3 ATR A TR
23. 2006 4 23 2006 1500 lt
24. 55 P2M A 2002 OL 79 DOERR
25. RT 2000 4
26. eee Vol 36 No 2 pp 28 37 2002 Mae AA j 7 EE IT f 1 2 52 3 4
27. 11 30 12 3 2 3 152 40 881 11 30 17 787 12 1 22 845 4
28. Ek 53
29. Web CSCW push pu11 web EK web cama cutee aan
30. CE TOTO 2004 FR LOT VIR OCW SCE CHS web ZEA
31. Sit Al
32. MY 1 2 5 r _ od 36 12 5 RT 2 RT
33. 1 TA Java 2 Runtime Environment SE v1 4 2 EARMEN PRE EPER DRAA 6 SUN ME ROSYSTEMS INC Hit ESELI eis JAVA TM 2 RUNTIME ENVIRONMENT J2RE STANDARD EDITION VERSION 1 42 SUN MICROSYSTEMS ING UAF Sunta t BERN aI F ARE SU MEHE AAM FHL MAFO orro RRR LSS CRRA gS ERE LS ORE ES LE ADD gt
34. 2 0 0 5 AS ar
35. 66 2 1 2004 10 15 16 2004 12 10 34 BS 2 3 4 2004 11 11 2005 06 10 2005 06 10 2 2005 06 16 EX
36. TRO 7 UNA Ce GeO CVC ep Ce ies 39 2005 12 2006 3 PC AV
37. 4 23 2006 1 1 40 980 0 NHK CATV
38. 43 4 6 12 12 8 6 8 14 8 Gibbs sampling Q learning 4 6 2 4 6 6
39. 1 CT WFT a 2
40. 61
41. 8 58
42. CAT V ROA Riis CATV NHK
43. KD Os
44. 3 PR 9 19 OMHE 6 D
45. 6 Pyar DT OTHIECEV 2 VeEBBICOREBRETIRO UR yy hE LUG RIES EAT 2005 1 2 200 54F2A15 16E
46. NHK NBS CATV LCV LCV 120 CATV LCV NHK 42 2005 6 LCV 120 120 6 21 16
47. NT in 7 1 0 2 0 0 5 6 ITR Oc een oes 7 l 8 5
48. TV
49. 3 ag 5 aa ee ee ome oe a i a OE TN A a BSUS CLOVER W amp 9 DESIGN a 2005 2 18 19
50. 0N 2 3 3 Ethernet EPSON NT 2700 Ethernet Ethernet 1 Ethernet 2 3 3
51. ies 1 50 6
52. CAD CAD CAD
53. 1 2 2 WCE es 6 8 1 1 We ere a 6 4 1
54. 44 5 1 2 2 mina Leeks 1 A
55. 2 5 2 1 5 2 2 47 2 5 2 3 5 2 4 X 5 2 5 48 BRT e ae 7
56. MIE ARK BP 2003 Vol 36 No 2 pp 28 37 2002
57. 2 Q learning Actor Critic 2 Actor Critic Q learning 2 Q learning l G
58. Window PAUSE Pause Push OK Button to Resume OK OK 13 Files LoadParamFi le 3 3 SaveParamF ile ae System Start se ee Stop ost tt tees Out amp amp da TimeSteps 10 50000 10steps 50000steps PAUSE Pause ee ee 14 5
59. 1 6 2005 AE 6 a 200 5 2 00 6 2 0 0 5 ee Cae
60. oa 2 4 1 18 PIXIE 24 2 Psr GOD 2 4 3 MRR M2X3L A B 2 4 4 19
61. web 54 web
62. X 2 4 5 ERX 2 4 7 6 2 4 6 1 a 20 RR HEA
63. NHK CATV 5 CATV 90 NHK 63 2 1
64. X 2 1 3 OCT2 12 FEF 1 2 6 1 2 KPORECHS BP iis A 8 5 B X 2 1 4 13 OTEA SHAE 1 2
65. CAT V ICDA Pe CATV 62 NHK
66. sees JEH WE Pe HS ae RBS 394 0021 4 11 O ARE EAM BRE RA KAR MEER 1 1 15 ZBE 65 2 3 FR h
67. JRE Java 10 4 O f C Documents and Settings SD07 UDPQ Agent FE CE RT SATANA Yo APH ay 9 2 PARR Wr Fr PELA C Documents and Settings SD07 UDPGAgent vy Eee NetagentMTQLnsnajaya NodeDataclass a NodeData java 2 QFuncclsss A COPA Web A QFunc java fe reall BAK felreall 52bat PS 1 Q1 bat reall offline BAK 1 O2 bat FS reall offline bat 5 O1 bat real4of
68. nent ee OR gt Ja 56 CAD
69. Actor Critic 4 5 4
70. Sa te Ve 2 28 6 8 6 19 SMALE COT ETRE 1 0
71. 2004 6 2004 7 10
72. JRE 19 rel Java 2 Runtime Environment SE v1 4 2 le es Eo 4 ERELTI w RLS ety i ee al InstaliShted T Java 2 Runtime Environment SE vl 4 7 Java 2 Runtime Environment SE y1 42 04 istalliShied Gath Java 2 Ruri Enron SE v14 M e Le Ue 2
73. 200 004 10 1 ST ROE NN 2004 11 2004 12 1 26 2004 12 10 11 6 6
74. Q amp A 1 3 3 6V 2 Ethernet 5V 3 Ethernet 10Base T Ethernet 4 2 5 EPSON NT 2700 a 1 DVD DVD CD ROM CD
75. 2 MM 1 2 3 4 2005
76. 17 6 CD ROM CE E RTV BACANA YD OQR O B PRR Br PELAO gt Ex JDEL UDPOAgent 2006 05 02 14 36 ED 5 pdf 33KB Adobe Acrobat 70 2004 06 19 1529 jre 1 4 2 04 windows i596 238 KB gt 2006 05 02 14 40 8 COPA RE Web were SI ES EL aoe 737 E 0 g O GQ 3 269 KB F J RE CD ROM JRE1 4 2 CD ROM NO J RE j2re 1_4 2 04 windows i586 p if
77. D 6 1 2 2 D
78. DIY CAD Web 2 0 7 2 AREICHR 6G LER D 6
79. 1 b 1 2 4 2 8 CT ICT CAD NO ICT
80. ol Vol 36 No 2 pp 15 27 2002 Vol 36 No 2 pp 38 53 2002
81. Beeb Bots Microsoft Internet Explorer IVE RED RTV YD APH RS O 2 JORR wi Jt http www shisaku jp filehtml pid 1 Googe HH OIRR EFoodee oo 2 iS SVR Rel ZA IL oT eR Poa samman 8 71MB 9 BFA 2004 11 26 22 58 3 Ver aa fE04112 pir i in 2 aes c TE ANTENA 041125 pdf page dr 478 1KB 1 ee 1 25 Ea Les 041124pdf 2 6886KB 98711796 T 041123pdf pg ale 598 8KB 1 ahi 23 a TE be It 48828041 123 pat AIDE RIE p e al Kna g ARA Uas C 8 misumi BELTpdf z ageing me oe E belt 041121 pdf pe 188MB 1 e A a
82. Rite eee Fu ee ae L 1 5 1 2 Xi 5 1 3 46 2 DARIE 2 a
83. aS Eaton S 2 5 2 1 SM IAS CU s 2 TOAT P LA a 2 i ER
84. 2 T 112C A 1 EE NIE IE tei P TE 30 0 20 0 10 0 0 0 5 10 15 20 2 ee ee 80 0 70 0 60 0 E 50 0 o 40 0
85. a OAD ATO eit fi Ei 1 N HK
86. 2 OCT1 2 1 1 2 8 4 2 11 2 ice 1 2 4 swing gt X 2 1 2 OCT1 e A AAL ERE 1 2 2
87. A 1 SASHA 13 30 15 30
88. Boe y 3 di 10 00 17 00 gt BEF 44 10 1 10 12 20 5 30 5 g L i 40 3 50 7 60 4 60 1 6 EEJ EFJ
89. 29 6 29 6 2 9 1 6 29 Ok 12 45 14 15 90 2 3 3 7 3 4 30 7 1 6 in 10 1 2 0 0 5 E Y Ri ae
90. 10 29 30 10 29 3 0 30 5 0 0 NHK UA Cec
91. CD R CD RW 2 3 4 Ethernet 10 100Base T Ethernet 5 6 LAN 11g 11b Ethernet LAN e Pentiume Ill 4 Microsoft Windows 2000 Windows XP 512MB RAM 1GB 1GB
92. 5 SE Fo a b RA Summary 1 Outline of the Project Reinforcement Learning is a machine learning framework so as to acquire a particular function autonomously through trial and error In this project we have developed walking robots with 6 limbs through collaboration of the industrial accumulation and several universities with coordination of the Okaya Chamber of Commerce and Industry The developed robots were demonstrated successf
93. Ans RARE PEEK 1 aa EZAR 2 1 POM 1 2 2 4 8 21 2 4 10 24 11 be jarani f DE ih i i ia ul i E EER 6 aN 22 p ap fet 4h 5 U 9 X 2 5 1
94. 30 0 20 0 10 0 0 0 5 105 155 205 Sy 1 2 2 W a 5 10 15 20 3 3 A 17 0O8 2005 3 1 2 amp amp WN
95. 2 EB X 2 5 3 24 6 AFT YY E an 2 6 1 25 3
96. 12 MiG ey a bp Java LAN 1 amry UMS Nm 3 1 3 1 10 2 3 Ex 1 1
97. 2 0 0 4 2003 12 25
98. a C WINDOWS system32 cmd e Y NetAgentMTISnsnaWindow mB Files System TimeSteps Func PAUSE C Documents and Sett ings amp a NetAgentMT Snsnallindow 1 chase 12711 prm hostname 192 168 1 108 TCP IP port number 567 hostname 192 168 1 108 IP port number 567 ParamFileName 4 stateDim 10 act ionDim 6 NumOtRewardVariables earnl nest ep discountRate ambdaP DU3 4 192 168 1 108 5677 IlearningRate learningRateP Finished ig cx CH WINDOWS syste 2 NetagentMTISnsnaWi Java J RE Java J2SE Java Runtime Environment JRE JRE Java Java
99. 77 CRIFLE ERAS 59 FECHA
100. NHK Vi Aa FAL TEDS EAR TRO CWO CE CRED 6 b ARC BNC RMR 78 ESAE 6 fe
101. CoC HMOs aS Teva el Uae AT ee CT ASAT Vas aS eye ve OS 1 1 DBR 1 Pwo Ive Ades oes 2005
102. FeO 1 2 1 MAINT D o 1 6 15 _ CETE 2 RORACA AREIG CHR EE A D BASE CREMEREOT MELETUMSORHGA oe AAR D 2 con ener k aoe X
103. Fo TCP IP RO IP IP IP 192 168 1 120 lt 192 168 1 254 255 255 255 0 SORDM OCP IPIOS DART Ret RED Cato IP SESS ane Asse cit TP IP FEL AMD FANU D ONS h MP ELA EB BHA ed ONS OF rE ee Bae DNS ONS H Yia Stine wl 21
104. Os 2 B DO RG 1 ETEB A BDIH 5 1 1 45 2 1 2 1
105. 5 0 6 2005 3 OAR 2 3 8 2005 4 6 At 20054264 2005 12 2006 3 2005 6 2006 3
106. 0 TARP HT A
107. 16 1 7 ESCEA A AER gt Q 6 64 2 2 1 E
108. 1 0 1 bat 1 02 bat 4 0 1 bat 4 02 bat 4 0 3 bat _6 0 1 bat 6 02 bat 6 0 3 bat 3 Window Files LoadParamFile 1 4 6 prm prm prm 3 TimeSteps 200 steps 1000 steps 4 Window System Start
109. D IS OA a 8
110. 3 l 2 5 2 7 BRONTE YY TNE Ze 2 X 5 2 8 3 tA BRL TWA hE 2 3 5 2 9 90 3 ODHT aX
111. a Co
112. 1 EPSON NT 2700 OFF 3 EPSON NT 2700 gt gt 3 gt Web http studious jp
113. EPSON gt 44 2 o O O OOR ia TO O N p AIBO gt J NA o Sap E o AK Fa HAA 7 _ 2 o 3 8 O ORU 8 aR E COW RF RE CST EX DRE ARAL Bx e DBE CREM TOER Of OBE TJE 0 J KA S nt amr ke a mk mk N mk mk NAINI 1 GIINI NO1 OO JJ 10 1G Hh J oo O O1 amp Co Roy SIS IS S EA Pas as
114. 10 10 20 0304 E 40k G5ott 60 60 BUA D Ver 1 O OU A N O VRE EG EE EEEE REEE EREEEETEE RAF AVREMBTA rrna 1 AC Ethernet EPSON NT 2700 CD ROM
115. SW i 4 EPSON NT 2700 5 ON 6 EPSON NT 2700 EPSON NT 2700
116. 0 2 4 0 1 4 02 4 03 6 0 1 6 02 6 03 bat bat bat bat bat bat bat bat 11 java UDPAgentWindow 192 168 1 100 5677 IP TP IP java UDPAgentWindow 192 168 1 100 5677 4 Drm I P 12 1 2 J a v a
117. 1 6 7 04001511 0 1 pdf 6 1 8 3 1 1 DP BG stat eet eeee eet ee ea eee eo we es i aa eoan pee as Owen ee eee aS on HIE faaan e e e ee eeeooccooo coo 6 co oN 2 2 ee 1 2 3 4 1 2 2005 02 18 19 3
118. RAD ee on 34 10 19 20 DTF Bee EERE 5 DTE Vile 2005 10 DTF 6 0 0 DTE
119. ml R3B Av gt JRE Java com 20 JURE Duke JRE http java com ja download help testvm xml CD ROM UDPQAgent IP T bao
120. 49 NO
121. im POMA 2 3 2 2 X 2 3 2 83 1 02 1 4 Z004 11 2 A X 2 3 2 4 6 17 2 3 3 2 3 4 Lees beer ol 2 3 5 1 6 1 1 6 4 2 0
122. 1 1 4 2 1 5 1 1 2153 14 2 6 itis
123. 2 3 1 DREAMER 1 1 oe 46 yee ua BUY HL te 75 4 2 mA y HRAT A pa isi SUS304 t 3 0 al SRF yta Rem Y DMI _ EB SARA Y Hor EE RATSA A amp 3 4 234 AML EA GAY E Lorian 754 AML a E et N mate F uim AE Ful St 3 0 L it 3 07 LORY Ett 2 3 2 1 16 10 2
124. 25 jpe lite ia71kg 1 2004 11 17 B QR push flow pull stock web Push Pull Flow Stock Web Cant E e vee V IP Web
125. PO 12 2005 10 14 PR 2005 10 15 FF 2005 10 80 Tee e CATV 2005 02 09 and 4001 51 1 0 67 1 2 92005 02 18 19 3 68 2 2 2006 2 15 22 A 4 No6 2 2 1 2
126. Pag Ua Cau Ca NN 604 m 28 FUNAC EPSON E AIBO O ASIMO ORR BSR DAA Mm D ERA BLA DER m omi E Dal D ERR
127. fline6dofbat real6bat real6 151 bat real6_152 BaK real6 152bat 3 TEA at real6offline BAK j fe lreal6offline bat a SD04FourLeg6dof BAK AgentMTQLnsnaCanvas class fe SD04FourLeefidof bat lavebat SendPacketclass CalStatistics class SendPacket java J CalStatistics java StateNode class config14sBatxt i StateNode java J config18s12a txt taskOListtxt Y confiel8s12aXx txt UDP AgentWindaw 1 class GenerateNodesUsingEpisode BAK UDPAgentWindow class GenerateNodesUsingEpisode class H UDPAgentWindow java Generate NodesUsine Episode java UDPnetlOclient class J1 lOboardSimjava a UDPnetlOclient java iikken10 bat BOSS Lett S jikken10_251 bat s localbat MT19937class JA MT19937 jaya NetAgentMTQLnsna class 397 KB o 63 1 f C Documents and S me 1447 gt as IP 1 0 1 1
128. have improved the robots in the following points for more effective application of reinforcement learning algorithms to the robots Improvement of the reward sensor introducing a slider mechanism Improvement of durability of the robots by a developing the heat sinks for servo motors and b introduction of the pre tension mechanism by torsion springs for limbs QAM 1 reinforcement learning
129. ibbs 3 Java PC LAN 1 3 1 LAN UDP
130. tw exe 1 http java com 2 3 Windows 4 YZ 5 18 eet ca eo ane tno ome jre 1_5_0_xx windows i586 p exe 7 8 jre 1 5 0_xx windows i586 p exe 1 5M 9
131. ully in the 2005 World EXPO AICHI JAPAN As well as the developed algorithm of reinforcement learning the project aims at contribution to the society through development of the process of trial production through collaboration of industrial accumulation and university researchers 2 Development Issues In this project we addressed the following issues as targets of development 1 Development of the reinforcement learning algorithms Under coordination of the Okaya Chamber of Commerce and Industry advanced reinforcement learning algorithms for acquisition of walking function by robots in real time has been developed by researchers in Kyushu University Tokyo Institute of Technology Kyoto University Nagano National Collage of Technology and Tokyo University of Science Suwa 2 Development of 6 limb robots With assistance of several factories in the Suwa Okaya area the Industry Network Corporation has been developed 6 limb robots as a test bed for study of reinforcement learning 3 Investigation for Promotion of the Results The Okaya Chamber of Commerce and Industry has investigated the promotion of the result and related issues by holding several committee meetings 4 Demonstration of the Robots The developed robots were named The Reinforcement Learning Robots STUDIOUS and were demonstrated in the 2005 World EXPO AICHI JAPAN 5 Improvement of Robots Based on the findings obtained in the demonstration in the EXPO we
Download Pdf Manuals
Related Search