Exploiting Interaction with Environment to Solve Some of Hard Classication Problems and Its Experimental Demonstration

Conventional Arti cial Intelligence and cognitive science have been argueing the emergence of intelligence by completely separating the system from its environment. In this paper, we will intensively investigate and analyze the role of the system-environment interaction through the comparison between the information processing and embodied cognitive approaches. By taking account of this interaction, we will experimentally show some of the hard classi cation problem su ering from the large input space and the ambiguities due to the perceptual aliasing problem can be greately simpli ed. To demonstrate this concept, we will apply to a garbage-collecting robot as a practical example. %-!<%o!<%I!’<+N’0\F0%m%\%C%H!$<1JL!$4D6-$H$NAj8_:nMQ (automonous mobile robot, distiction, classi cation, interaction with environment) 1 $O$8$a$K <+N’0\F0%m%\%C%H$K$H$C$F!$<1JL$OM?$($i$l$?%?%9 %/$ro$K:$Fq$JLdBj$H$J$k!% $3$l$^$G$N8EE5E*$J?M9)CNG=!JAI!K$dG’CN2J3X$NJ, Ln$K$*$$$F$O!$<1JLLdBj$O%7%s%\%kA‘:n$d9bGE@$,CV$+$l$F$-$?!%$3$NMM$J]%f%K%C%H$X$N%^%C%T%s%0 $K$h$C$F%b%G%k2=$5$l$k!%K\O@J8$G$O!$$3$N%"%W%m!< %A$r>pJs=hM}E*%"%W%m!<%A!J information prosessing approach!K$H8F$V$3$H$K$9$k!%$3$Ne$G!$CNG= $NAOH/$r9M$($F$$$k!%$7$+$7$J$,$i!$%m%\%C%H$O%3%s %H%m!<%i$@$1$G$J$/!$%"%/%A%e%(!<%?$dJ*M}E*$Jo$K4JC12=$5$l$k$3$H$r<($7 $? [2][3]!%H‘$i$O!$?HBN@-$d%7%9%F%‘$H4D6-$H$NAj8_ :nMQ$O$I$A$i$b!$CNG=$NH/8=$K=EMW$JF/$-$r$9$k$3$H $ro$K:$Fq$J$?$a!$L@3N$+$D8z2LE*$JCNG=$N9=C[K! $O8=CJ3,$G$OB8:_$7$F$$$J$$!% $3$NLdBj$r2r$/0l$D$NM-NO$J2A4X ?t$N@_7WK!Ey$K!$0MA3LdBjE@$,B8:_$9$k$b$N$N!$$3$N %"%W%m!<%A$K$h$C$F!$G’CN.$5$J1_Cl!K$rJI$r6hJL $9$kpJs=hM}E*l9g!$IOe5-$N $h$&$J7k2L$,F@$i$l$?$N$+$N9M;!$O9T$J$C$F$$$J$$!% $=$3$G!$K\8&5f$G$O!$$"$$$^$$@-$,K\e=R$N9M;!$K$*$$$F$O!$ action $H behavior $H$$$&35G0$rJ,N%$7$F9M$($kI,MW @-$,$"$k$3$H$r<($9!%$5$i$K!$K\Z $ $k$?$a!$%4%_<}=8%m%\%C%HLdBj$X$N1~MQ$bJ;$;$F;n $_ k!% 2 7B 5cm $N1_Cl!K$H!$$=$l0J30 $NJ*BN!JD>7B 4cm $N1_Cl!$JI!K$N<1JL$re$2$k!% K\8&5f$G$OGE@$rCV $$$F$$$k$?$a!$7B 55mm!$9b$5$O 30mm $N 1_Cl9=B$$rM-$7$F$$$k!%$^$?!$8!CNHO0O$, 0mm $+$i 30mm $^$G$N!$8 $D$N@V30@~6a@\%;%s%5!J6 $D$OA0B&!$ ;D$j$O8e$mB&!K$,Hw$(IU$1$i$l$F$*$j!$$3$l$i$N%;%s %5$O8!CN5wN%$K1~$8$F!$0 $+$i 1023 $^$G$N@0?tCM$r JV$9!%$^$?!$D>N.%b!<%?$GFHN)$K6nF0$5$l$k:81& 2 $D $N$l$N>l=j$K!$J*BN !JBg!?>.1_Cl$b$7$/$O!$JI!K$rCV$-!$IR %;%s%5$NCM$r 5-O?$9$k!%@:EY$r>e$2$k$?$a$K!$$=$l$>$l$N>l=j$G!$ $=$l$>$l$NJ*BN$KBP$7 30 2sB,Dj$r$7$F!$J?6QCM$HI8 =‘JP:9$r5a$a$?!%8e$K9T$J$&%7%_%e%l!<%7%g%s$G$N%; %s%5$NCM$O!$$3$NJ?6QCM$HI8=‘JP:9$r4p$K:n$i$l$k!% Figure 1: Experimental setup for the collection of