The 22nd Annual Conference of the Japanese Society for Artificial Intelligence, 2008
3H2-2
! "$#%&
Automatic Generation of Control Programs for Cleaning Robots Using Simulated Annealing Programming
')(+*),
∗1 Mitsunori MIKI
-).0/)1
∗2 Yuki MATSUI
2)3+4)5
∗3 Tomoyuki HIROYASU
∗1687)9;:)<0=?>8<)@
Department of Science and Engineering, Doshisha University
∗2687)9):)<+:)<)A?>8<)B)C)D
Graduate School of Engineering, Doshisha University
∗3687)9):)<0E)F?G8D;<)@
Department of Life and Medical Sciences, Doshisha University
When precise movement is demanded of a robot, the control program must be optimized regarding not only program structure, but also the degree of movement. However, only the program structure was optimized with the conventional Simulated Annealing Programming(SAP). Therefore, we propose the method that optimized degree of movement as well using Simulated Annealing(SA). The control program generated by the proposed method for a cleaning robot performed well.
1.
HJILK)MNPOPQSRUTSV8WYXPZ\[L]_^\`PaSb_ZP[\cPdSeLfgX\hLR\iPj
k ]Pl\mPn\o_pLqPrPs\t_u8vg[\wPx\aLbLvy[PhSRPi\jLz_m\i
{P|8}~_
u\LbL
L
QP P
{PLP\L
eLsS_e
P r\_P L bhLRPi\j
k
]_ PX\ P bPS _\P
e_ _ b__X8v[PhLRiPjLzUmPiUXPPP guQ¡¢
XP£P¤\¥
SP¦
zoL§gpP¨SV8©YªP«\p8¬_m\i_hSRPi\jLz_m\i
(Simulated Annealing Programming:SAP)[S® 2007]X\£
¤L]_ZLt_uPLb_
SAP Q ¦ zUoL§yp\¨V¯©Yª\«\p8¬m\i
(Simulated An- nealig ° SA)]_±P²P³ {´Lµ bPs_r_¶P· gPP LL¸ Qº¹ Z£¤LrU»PuP¼m_½¾pL¿UÀPÁr_»ÂPbUÃPÄÅ {Æ~_ uP
b
[L® 2007]¯ gÇ8 gQPÈPÉLe_[ ]_ÊPË ~_ bPR_TLV¯WÌX hLRiPj
k
L QºhLRPij
k
XP²³L]UÍPÎPÏ gQ ~LÐ r_[ X
ÑPÒ
L]_ÓÔPaLbUÕPÊ {
b_ºÖL
×
£¤
L
QØPÙR_TV
WYXPZP[P^P`PhLRPiPj
k
]_Ú ¸gÛSÜ Q
SAP ]_ÝPPuPhLRPiPj
k
]vg[wPxaLbUÞLrUQº[ X ÑPÒ
PXßPàe_ÍÎPÏ]_ab_
Lr_áPPuPâPãPaLb_
2.
äæåçéèêæëæìíî;ï8êñðò)óæô;õó)öæå ò)ó(SAP)
SAP ±\²P³L]_÷8P gu ´ PªPøLùL¬gú k
SA _û_ü
r_QºwPxPýþPQºÿPþ
Q
Q¿_p8¬UmPiUÇ
Ð
xb_
r_ßPàL] Æ a
1.
wPxPýPþ
1r_wxPýPþXPü] Æ aº
1XLsLUrXP÷PÇ
Ð
jPm k r_p)©g]8 gQ
~
_p)©g]8
aLbP±S]PÙPaLb_LrPÙ8 yLr_QLjPm
k
r_wPx8 gP±L]8 gQP÷L]_wPxPab_
2. ÿPþ
Q
\XP÷\X
E
\÷\X
E0
yX!
∆E(=E0−E)Q"#$ Ñ% j&_pLq T
]'Lr_QP÷
(*) ¹
: +*,.-*/ QLû10*2*354*3*456*7*4 £ ¤*8 Q 9;:=<=9
®?>=@=A
¢B :=C
1-3Q 0774-65-6924Q [email protected]
Ԛᰴ⸃
ԙ Ԙߩ⸃
ㆬᛯὐ
㕖⚳┵⸥ภ
⚳┵⸥ภ
㒰
ㇱಽᧁࠍᝌ
1
wPxPýPþ
L]_ÿPþPaLb_ÇDPÇ\X ]_Z8 _X r E
(1)
r Æ a
Metropolis'FL]_ÝPLb . PAC=
G
1 if ∆E≤0
exp(−∆ET ) otherwaise (1)
E
(1)X PAC
ÿPþHI L¸ QP÷ {JKLNMO
wPx ~_ P Ò QRSP ÿPþ8 gQ JTLMO wPx ~
P Ò HIULr_ÿPþPaLb_
3. ¿_p¬_mPi
ÛVÆ
1.
2.
XW ]X ¸ZY Z8ULU]UªP«p
¬Umi][#Q\ ] ª«Pp¬UmPi]UZtU^Q$ Ñ%
j&UpLq
Tk
]_ ~` aLb_ab^PX$
Ñ
Tk+1
E
(2)
r Æ acdePª«p8¬UmiP]UÝPuf aLbUeU»Q
γ
¿_p¬_mPiI L b_
Tk+1=γTk (0.8≤γ <1) (2)
3.
gih=j?k×
£P¤
L
QºZP[PcdPXvg[wPxlLÂ
L
e ` Qº[
X ÑÒ
LnmopqLerPÀ\Á8\ guPØ\ÙLRTLV8WYXPZ\[PcPdS] ´
PQ ]UØPÙPÀPÁZ[sPºØPÙPÀPÁ Q
1t XR_TV8W {
u8Ðg
PZ[vd
Qxwy_pPø;©Zz]_ÄIUr{
¸Z|
Ue
`}
[PaLbLU] } U8aLbUÀPÁ L b_~ g X{
¸gL b_
• wy_pø)©
2r Æ a 10 10¾PX 6áPXwypø)©g]_ÝPLb_
• R_TV¯WÌX~
R_TLV8W P X ]L guP» ¸ Q }Z XPXPâ
{ pq8gaLb_R_TV8W X ]L
2r Æ gLsL
r_Qwy_p\ø)©ÌX gQ MP_ÛMP gab_e_»PQ
wy pø;© X
1¾X\ > X ~ RUTV WÌXabU
1
The 22nd Annual Conference of the Japanese Society for Artificial Intelligence, 2008
C D E
F G H
ࡠࡏ࠶࠻
㓚ኂ‛
ო
⒖േน⢻ߥ㗔ၞ
2 wypø)©YXPü
• ZP[PcPd
ZP[PcPd8P guPQ
1Q"# 2r
Æ
a
V
]_ÝPLb_PPQLR_TLV8W
(
[pqLe_¾PXd
)
3v [pq LL¸ Q V ] 1áPZPabL_r ZP[vdL]
1 aLbP_X8gaLb_
1 V
if obstacle ahead 2 "!# $
%'&)("*+,.-/02143
156798
:#;<01<3
2576=">2?@'$
progn2 2 "!# $
3
1A8
3
2CBD
(
=">"?5@'$
progn3 3 "!# $
3
1A8
3
2A8 3 356EB6D ( =>2?@'$
progn2r 2 "!# $
3
1A8 3 2567 1:1BFG#H
I6J)K9L
=">"?5@9$
2 V
4#
RIGHT 45M'N5"O.PQ$
LEFT 45M'R5"O.PQ$
MOVE S4T#UWVEBXYZ4[
%\
?5@9$
%&#("*"+',.-/01
8]#B9^.H'_"`?@'$
• L
[pqLe5abLr_PaLb{ ¸ | gabPXc Ò r
st_uPË
bP_XgaLb_ g
{
t_u\Q
E
Q E
(3)r Æ a } UeddLrPsPtuË b_Le_»PQ
Se
[p
qLe5abXfgL]UQ
Sr
R_TV8W { [8 gfgL]_Ö
h
PÆ
a
E= (1.0−Sr
Se
)∗100 (3)
4.
ikjklkmknpoqlkrksktV
eLfUr_ÝP
Ðg
b_[
X ÑPÒ
(u vv
Ñ Q
[w
x eLf
){ mopqLeUrPÀPÁr_»PPuQN_p)©yLrU[ X Ñ
Ò
XPÓÔL]UaLbL Qs ¸ ÀPÁr_Î ghLRiPj k Xz{
{]|L
b_ÖL
Q}~
PL
SAPrPs
¸
hLRPiPj k
XP²³PXÍPÎÏL]UZPQ}~^ P [ X ÑPÒ PXÓPÔ]_Z
}~PL]5PaLb_[ X ÑPÒ PXÓPÔLr
SAP _û_ü
X
1 }~ L b SA]_ÝLbUP XP[ X ÑPÒ
PXÓPÔ
]
SA/c Simulated Annealing/continuousness [s
r_ßPàL] Æ a
1
±P²P³PXPÍPÎPÏ
SAP]_ÝPPuPQ±P²³PXPÍPÎPÏ]_Z8vd {
}~vdPX
R% r58 g8 }~L]58 QPX÷
]5}~ { Ðg PÍ K ÷8gaLb_
2
[ X ÑPÒ PXPÍPÎPÏ
1X}~LrPs ¸ { Ðg \Í K ÷ S b_±\²P³L]'Sr_[ X ÑPÒ \XPÍPÎPÏL]_Z¯vd { }~vdPX
(100−R)%r5 g }~]5PaLbUULe}W
Q r Æ a
2-1
wPxPýPþ
mopqLe % j&pLq\]\_á uPX p)©yr_»\
uPQ
%
j&_pqP]Pjm k rmo8 QP÷]_w
xPaLb_
2-2
ÿPþ
Q
{ Ðy P÷L]n¯ gQ
Metropolis
'F
E
(1) ]_ÝPPuPÿPþ QP»Ls# ]_Z
5.
kk× £\¤ S Q
3 P yPØ\ÙPÀ\ÁLr¯ gu\Q\ULe
SAP»sn#PQ 4
8 yPPS]_ÎPÝ8 yQL]_ZLt
P$ Ñ ePo_pPø ¿_p8¬_mPiP]_ÝPPP
SAP
X}~Lr
ÝPP % j&_pLqP]5
3 r Æ ae_»PQ 3 XPÍ$ Ñ
Í3 JT ]
50%5L b$ Ñ QPÍ$ Ñ 1¿_p8¬_m\i5 ] r_Í_ JT ]
1
v 5L b$ Ñ L b_
PPQPLr_»PuPQ
SArPsb % j&UpLq_ÓPÔr5 ¸ µ b5c Ò
R
75% SAPX}~vd 24,000Q SA/cX}
~vd
8,000 P gP[ X
ÑPÒ
PXPÓÔ¡¢8 guPQ
u v
v Ñ
{RIGHT,LEFT} 5£ 90
Ñ Q [w
x
{MOVE} R
TLV8WYX3 P~ X
0.5£ 2.0¤ XP8 gP
3 $ Ñ% j&_pq
¥§¦'¨
.© ª
total steps 32,000
cooling num 32
max temperature 144 min temperature 0.01
PX«¬L]
30 ZPXPÍ
K
PX®¯QÍ3PQÍ_
]
3r Æ aLe_»PQ 35°±{ PQ² ±P{ wy_p
ø)©g] Æ gQ _ ~ e³Lf K L b_
㪪㪘㪧 㪪㪘㪧㪂㪪㪘㪆㪺
㪇 㪌 㪈㪇 㪈㪌 㪉㪇 㪉㪌
䊐䉞䊷䊦䊄
⹏ ଔ ୯
㪸㫍㪼㫉㪸㪾㪼 㫄㪸㫏 㫄㫀㫅
C D E F G H
3 wy_pPø)©gr_»LÂPbPX
3Ç Ð Q SAPX´X}~Ls ¸ PPX LP{ ®¯PQÍ
_¯_\r Kµ
e÷L]¶{PSL
{·
ÇtPL
r\s
¸
QLR
TV8WYXZP[eLfUr_»LÂb_²³PXÍPÎÏlLÂ L e ` Q[ X
ÑPÒ
L_Í\ÎPÏPaSbPS QSs ¸ ÀPÁSr_ί gP÷\X}~ {¸
b_
6.
¹pº8K× £P¤ L QPØPÙLRTLV¯WÌX\ZP[P^\`PhSRPiPj
k
]_Ú ¸gÛ
Ü Q
SAP ]_ÝPPuPhLRPiPj
k
]Lvg[PwPxPaLb_ÞLrQP[ X Ñ
Ò
PXßPàe_ÍÎPÏ]_ab_LrUáPuPâãL] gP§«¬
Ç Ð QPX L{Kµ e5«¬L]{LbL {PL P»^X
¼PÁ8P uPQ
SA/crPsPbU[
X ÑPÒ
PXPÍPÎÏL]_Z8Uq5½8z
mPiPr_áPPuPQPâPãL]_Z8gÕPÊ {L bL¾ µLÐ b_
¿kÀkÁkÂ
[L® 2007] ÃWĶÅ#Æ284ÇÈÊÉ"Ë8ÌÍÎ2Ï8'ÐÑÓÒ"Ô ,Õ5Ö×2Ø7ÙU
Ú4Û'¶Ü6ÝÞ4<ßà2áâQãäS'Þ
¦
Ö'Ý'Þ ,åæ4çè"éê4ëÏ'ì.8<í"è îï'ðñWò7ó
ß ,Vol.48, No.SIG15(TOM18),pp.88-102,2007
2
