Top Banner
lS E-T R-81-2? 濾. .8婆 A MOBILE ROBOT WITH SONIC SE AND ITS UNDERSTANDING OF A SI by Yutaka Kanayama Shin’ichi Yuta Jun’ichi ljjima February 27, 1981
25

AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

Oct 11, 2020

Download

Documents

dariahiddleston
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

lS E-T R-81-2?

    濾.     .8婆

   A MOBILE ROBOT WITH SONIC SENSORS

AND ITS UNDERSTANDING OF A SIMPLE WORLD

                by

            Yutaka Kanayama

             Shin’ichi Yuta

             Jun’ichi ljjima

            February 27, 1981

Page 2: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

ISE-TR-81-22

   A Mobile Robot with Sonic Sensors

and its Understanding of a Simple World

Yutaka Kanayama     ロ                   ロ

   Un■vers■’ヒy of Tsukuba, Sakura, 工baraki 305 Japan

Shin’ichi Yuta

   Un■▽ers■’ヒy of Tsukuba, Sakura, 工baraki 305 Japan

Jun’ichi 工ijima

   The University of Electrocommunications, Chofu, Tokyo 182 Japan

Abstract:     This paper discusses how a mobile robot with .sSl!RgEELEe S.一9.一.PA.S....一.beSpme...;..一S.一9g.S..

can l。。k at the w・rld and翼旦典9叫噸$the difference between…璽聖.

and pthg,.r,.obJ.e..cts. This intelligent self-contained robot, named

Yamabico 3・l was de▽eloped under a proゴect whose purpose was, in

part to construct a robot which knows its position in the world.

The robot has 璽乱~三≡P∫⊇9卑pqtgr sys’ヒems for the brain and the legs,

and a cornmunication system between the two. Zt has .ESQuy supersQnic

eyesr a neck, and ba七teries. The leg sys・ヒさm consists of two driving

wheels with DC mo・ヒors and ensures smoo’ヒh movement・

     Second, a method for the robot to recognize a’simple world is

described. A simple method is proposed for n.oise eliminat二ion・

Then the robo七1s images of a simple world are described and a method

by which the robσヒ can recognize鵡1≦血虹_y愚.↓1β is proposed. 象黛旦t

』q巴鰭鰻無tting is used f・r the discriminaセi・n・This pr・cequre

giveS theゑ工£短喚£9...㌻9..the嚇ll and the罵ζgc襟9要...9≠.やe..wall with

respect ・ヒ○ ・ヒhat of the robot i s pa Uh・  Fina:Llyr a robo’ヒ con’ヒrol

me. Phod for ssF-t一.F一.ha....;. gh.一t.rgp..v. ero{.n,.t ..a.long q wa」..1 is proposed. The software

for all of this is a part of the basic brain monitor of Yamabico 3.1.

Page 3: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

 0. 工n・ヒroduction

     Mob±lity is an extremeiy ±mpodr-tant ability for intelligent.

robots’ because ’ヒhey cannot perform・ヒheir ・t二asks :fully without it.

Many mobile robo七s which use TV cameras for understanding thei:r

environment have been reported [1][2][3][4]. Scene analysis is

needed for an inteUigent automaton to recognize objects and get

『ヒhree-dimensional information.

     Supersonic sensing is another tool for understanding”ヒhe world

[5]. This paper describes the method, abilities and iimitations of

supersonic・ sensing by mobile robots. Smali power consumpt.ion, good

cost/perfOrmance ratios, light weighdヒ and ease of acquiring distance

information are advan・ヒages of this met二hod.

     Yamabico 3.1 is a self-cOntained intelligent robot.from the

Yamabico family and has functions very similar ・ヒO those of Yamabico

3 [6]. The name r roughly equivalent to ”Echo”, was ’chosen to

highlight its sonic sensing method. Since the angular precision

of sonic devices is limited, the images obtained are fuzzY. An

algori『ヒhm for deriving reliable information from noisy data is

presented later j n this paper.

     工n this analysis, the environment of the robot is restricted

to a hallway su:rrounded by plana:r waユls. Worlds of 七his type are

well adap七ed to being unders’ヒood by a robot二wi七h four eyes. This

is the first step foward describing the world by using simple eyes.

Expe「’men七s ス’th m。「e c。mplex w。「lqs will be rep。rted in subsequent

papers.

     .Sec’t二ion l describes the hardware and sof・ヒware of the Yamabico

3.1. robot. ln Section 2, some results of basic experi‘ments with

superson■c eyes、are presented.  These resul七s are needed in

designing the wall recognition algorithm presen・ヒed in Section 3.

                             一1一

Page 4: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

Section 4 is devoted

along a waU.

to a meセhod by which Yamabico can walk straight

1. Amobile Robo・ヒ Yamabico.3.1

     Yamabico 3.i was constructed to serve as a tool to develop the

world-understanding concep七 used by mobile robots.  (See Figure l)

One of the main purposes of this project is to construct a robot

which is aware of i’ヒs position in the world, wi’ヒh or without a map。

The Dresults reported in this paper are the first step towards that

goal.

1.1 Hardware

     The architecture of Yamabico・3.]r is shown in :Figure 2.  工’ヒs

main features are as follows:

(1) rヒ consists of a bra=Ln, tWo 工egs, a neck, four supersonic eyes,

    and batteries, carried in one body.

(2) 工’ヒ con’t:ains 七wo microcompUt;er systems for the brain and the leg

    、system, which are hand-shaked ’ヒhrough a communication por’ヒ..

    Each of the system consists o£ an MC6802 8.bit microprocessor,

    ROM, RAM and monitor.

(3)The entire b・dy can be mQved by七w。 driv斗ng wheels and tw。

    swiVel-ca s{二〇:r wheels.  The dri▽ing wheels are con・ヒ:ζo二Lled by a

    mic:roprocessor and DC mo『ヒors ’ヒ。 ensure smooth movement [7].

(4)・・The eye system is mdun’ヒed on ’ヒhe neckr which ca;ロ acotate around

    the central axis.

(5) The eye system consis七s・of four transmi『ヒter-recei▽er pairs Which

    are moun’ヒed so tha’ヒ adjacent eyes are separated gQ。 in’a2imuth.

    Each transmi’ヒter emits 40KHz waves horizontaユly ∂r〕d the

    correspondinq.receiver ge七s an echo if an obゴec●ヒ exists in j.ts

                               - 2 一一

Page 5: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

    direction。 The time interval is proportional ・ヒ。 ・ヒhe dis七ance

    to the object. This eye can detect an object as far as 3m

    distant。  The horns are now being rebuil’ヒ and 七he sensing

    ability wili be impxoved.

(6) The energy supply is from NiCD batteries.

!.2 Software

     Yamabico is a multipr。cess・r system and its s。ftware sys七em is

devided into two parts; the brain monitor and the leg moni・ヒor. The

leg system is a slave of the brain system. The communication between

the七wo is done by an interrupt procedure through the co㎜uni6ation

port and there is a leg command system (See Figure 3).

     The m・st imp。rtan七brain m。nit。r functi。hs are,

(1)

(2)

(3)

(4)

(5)

(1)

(2)

(3)

(4)

SupP。rting c。ncurrent pr・cesses and・nterrupt handling・

SupPorting basic 工/O functions indluding console ones, such as

   ロ                                                    コ

son■q eye operat■on.

Sending commands to the leg system, such as a turn co㎜and.

Suppor・ヒing basic arithmetic functions.

Execu’t=ion of 七he user I s proqrams.

 The mos’ヒ impor七ant leg monitor func’ヒions are:

Supporting concurrent processes and interrupt handling。

SupP。rtinq the basic leg c。㎜andr・

Sending commands to the brain system.

Suppor’ヒing basic arithmetic func『ヒions.

a:.m一一g!in1igslSg!2ELES3sE一一gU!}g-SgnE!s2LSgpsgizsCharactensbics of the Sonzc Sensors

     工n this section, basic experimental da・ヒa of the sonic sensors

of Yamabico 3・l are presented. Assume the robo・ヒ is mo▽ing straigh・ヒ

’ヒ。 the right along a wall (See Figure 4).  工f Yamabico uses 七he

                               ・一 3 一

Page 6: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

lef七 eye ・ヒ。 measure the distance ’ヒ。 ●ヒhe wall at fixed intervals,

we obtain a sequence of pointst Pot Pl, …, Pi. Assume that the

path of the robo七 is straiqh’ヒ, and take a coordinat二e system such

・ヒhat:the x axis coincides with the pa’ヒh.  Then the sequence is

exp「essed as(x・’y・)’●’ (xi’yi)’whe「e each xゴis。btained

by using a leg command that tells the walking distance, and each

yゴis。btained by using an eye function・.

     The following figures show how a simple would is seen by the

robot wit二h supersonic eyes. Fi gure 5 shows a result for an une▽en

wall.  Each dot in ’ヒhe figure represents an object which caused an

echo. No・ヒe tha’ヒ at the transient area TA, the obser▽ed ▽alues do

no’ヒchange ab:rup’ヒly. We no●ヒice some points which lie much closer 七〇

七he r・b・t than adゴacentか・in七s d。. We call them n。ise datar which

are supposed to be caused by environmental sonic noise. , All noise

data are smaller than true da‘La. Zn a usgal laboratory environment r

the pr・babili七y。f n。ise data gccurrence is ab・ut l署・

     Figure 6 and 7 describe how the eye sees a plastic sylinder・

and a person standing in fron’ヒ of a s’ヒraigh・ヒ wall. The broken

ve:rtical lines in the figures represent 七he absence of echoes at

・ヒhose poin・ヒs. ]=n this case we st:ipulate that y.=0.                                               )

     We are interested in the angular precision of sonic sensing.

工f ’ヒhe eye looks at a wall at an incidence angle of 450, then ’ヒhe

sonic recei▽er accepts no echoes a’ヒ all.  Figure 8 shows responses

which are.o:btained by scanning ’ヒhe wall when ’ヒhe neck is ro.ヒa’ヒing.

The data tell us that七he eye gets an ech。 when it l。。ks’ ≠煤@the wail

a’ヒ an incidence angle less 『ヒhan 200.

     When the robot is sta’ヒionary and a person goes by beヒween ・ヒhe

robot and a’@wall, the eye perceives an image like that shown. in

Figure 9.

                               一 4 一

Page 7: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

     The preceding results say tha七measured values sometimes

contain noise, ’ヒhat false ▽alues are always smaller than the true

ones’ and tha『ヒ under certain condi’ヒions ・ヒhe eye misses echoes.

Let us propose a simple method of elimina・ヒing noise.

     Assume ’ヒhat the robot is moving and・coordinates (x.’ y,) have                                                        コ  ⊃

been obse:rved.  Le{二 (x, y) be the results of the next observation.

「pmos七cases’there is a parame七erδsuch that the difference

ly 一yゴビδナ・n-the・ther hand’if y・yi一δ,七hen y might p・s』ib・y

be noise data・ Also, if y=0, (・ヒhat is, :no echo has come), ・ヒhere

i串 a slight possibility 七hat the eye will catch an echo in 七he

second trial・  工n the lat’ヒe:r two’cases’ new coordina・ヒes (x曾, y雪) are

immediately obse「ved and are taken as(xゴ+・’yゴ+、)・That is・

          (Xj+・・ Yj+・)一{:1∴llhl誌ll岬y≠0’

Thenr ’ヒhis method makes not more ・ヒhan two measurements for one point.

One reason    ’ヒhe robot does not二make many obser▽a・ヒions is ・ヒha七

’ヒhe robot is mo▽ing be・ヒween sonic emissions.

     :F■gure lO shows the result of noise elimination in the same

   コen▽■ronment as ’ヒhat of Figure 5. This function is implemented in

the basic eye subrout;ines of the brain monitor.

pt. Recognizzng Walls

     エtis necessary f。r a m。biie r。b。t t。 kn。w the distance fr。m

itself t。 the nearest wall・工七is’h。wever, n。t apPr・priate y・ge七

QniY one value and accep・ヒ i・ヒ as a 1・ヒrue distance’ because random

十  Typically, δ=4cm.

一 5 一

Page 8: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

  ノ

noise may override a signal and. because the eye may pick up a value

in a tranrient a「ea between tw。 walls・The following l.ine.趾ting

procequre is employed after noise elimin.ation. This is

a sort of one. 一dirnensional image processing.

     We h4ve to determ」ng whether or not the mobUe robot is facing

a planar wal:L.  For tha’ヒ purpose, we ’ヒake the last 2:N+l data

(XnvN, YmN)r ’”r (XN, YN) and make a least squares tit to get the

equation. y=ax+b.w±th respect.to the robot path. (See Figure ll).

Assume that f・r allゴwith-N≦」≦N’yゴ≠0・That is’at each 2N+l

point, the eye accepts an echo.

     The coefficients a and b and the sqm, of squares of residuals

S are giyen by the foUowing.equations.

          (iXj2)a+(lxゴ)b.一lxゴyゴ

          (iXゴ)a+(!’)や=lyゴ

          S=lyゴ2-aΣxゴyゴbΣyゴ

                      コ      ]              コ

where all ’summation is taken from 一N to N.

     Let x-N,…, xN be n・rma・iz6d as-N△x)…’N△x respective・yナ・

That is, the’x coordinate is local ahd moved to the forward direction

of the robot as a new current vaiue (x, Y) comes. Let K denote’

争(N+・)(2N+・). Then la;・band S a士e e▽a・uated by七he f。・・。wingl

J

          a千丈IXゴyゴ・

          b 一 sSi.FTiiYj

           ly」∵妾(IX」y」)2一 sStT・(lyゴ)2

Tberef・re.七he・n・y r’ ?曹浮奄窒?高?獅煤@is t・c・mputeΣyゴ’Σx」yゴandΣy」2

at each step. The p’窒盾№窒≠香@for this evaluation in real tirne i. s

十  Typically N=8 and △x=・lcm.

                              一 6 一一

Page 9: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

shown ±n Figure 12. Zn this programr the variables ”sumy”, ”sumxy”

and ”sumyy” denote the above summations. z[!. ar1] is an array

representing the sequence y-N to yN. The variable ”sig”, ”dx” a’ 獅

lt xO 1’represent the number of nonzero values in ’ヒhe array 乞, △x and

nAx respectively. [Phe 2N+1 data under consideratiop a.re recognized

as on a.st二:誓aight wall if sig=2N+l and SくSmr where Sm is a

parametet。f this meth。d+.

     This algorit:hm has been coded and execut二ed for the situat二ion

shown in Figure 5. The result is presented in Figure 13, where the

bold parts of the lines are the region that the robot recognized as

a wall.

     Tf there exists any difference in the distances between walls

and 『ヒhe robot, then t二he robot recognizes the shape of walls. Thus,

the robot can draw a map of a woscld which the robot had not known

before.

4. Control of Straight-Line Movement

     TO walk straight is an impo:r’ヒan’t二function for mobile robo七s.

工t is, howev6r, impossible to do so through a long Pa・ヒh wi・ヒhout

help of eyes, because of imbalances between ’ヒhe ’two mot二〇r sys七ems

and irregularities in 七he floor surface.  Therefore it is necessary

for the robot to utilize its eye during its walk.

     A first-order pMedictipn of the track of the robot is the

basis for ’ヒhis method.  SupPose that we ha▽e Yamabico walk along a

long straight wall. 1ts four eyes and the algorithm described ±n

Section 3 can be used for that purpose.

t Typ±cally Sin=IOocm2.

一 7 一

Page 10: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

     LL l is a line on which the mobile robo’ヒ is to walk, and is

away from the wali by a distance c. (See F±gure 14). When the

rgbot is at Pl:

(1) By using ’ヒhe algori’ヒhm.shown is Section 3, the robot gets the

    distance b and. the slope a at Pl and predicts its position P2

    which is beyond Pl by a distance e.’ Let b2 be a distance from

    the wall to the robot.

(2) 1f c-t S b2 S c+t, then no special operatio.n occurs. Otherwis’

    the directi5n of the robot is changed by Ae= liLitli E-C such that new

    expected point Pi is on LIJ’. This adjustment of direction can

    be done smmothly by one of the leg commands of Yamabico 3.1 [7].

(3) The above procedure (1) and (2) is repeated once ih a distance

    r along the route. t, e and r are important parameters of this

    procedure. Their typical values are 5cm, 2m and lm.

5. Concluding Remarks

     The four-way eyes work 21stLx..m.y well in recognizing worlds where

弓↓よ一・唄工S~鎚一脚allel・r.一Lゑ七工独隙鳳e鉱一The c。st階

perfoDrmance of supersonic eyes is good. Least square fitting is

prXac’ヒical for recOgnizing walls in real time. The leg func・ヒions

mentiQned in this paper are necessary fOr a robo・ヒ without any map

of ’ヒhe world and also for a robo『ヒ tha・ヒ has a Inap of the world.

一 8 一

Page 11: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

Acknowledgement

     The authors thank Professor J. W. Higgins of Tsukuba University

for his useful suggestions and oritical reading of the manuscript                                                               ,プ

and aiso thank Mr. Shingo Tokizono for his experimental work for the

charac『ヒeristics of sonic.eyes.

一 9 一

Page 12: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

                                                        ノ

References

[1] Nilsson, J. J., A.Mobile Automation: An Application of

    Ar・ヒificial 工ntelligence Techniques, Proc・ of lst 工JCAI’

    pp.・ 509-520, 1969.

[2] Dobrotin,’B. and R. Lewis, A Practical Manipulator System,

    Proc. of 6th 工JCA工, pp. 723-732, 1979.

[3】Th。mps。n, A. M., The Na▽igati・n Sy Stem・f the」PL R。b。七,

    Proc. of sth 工JCA工, PP. 749-757, 1977・

[41Giralt, G.,R. Sobek and R. Cha・ヒila, A Mul’ヒi-Le▽el Planning

    and Naviqation Sys七em for a MobUe Robo’ヒ’ Proc・ of 6’ヒh 工JCA工’

    pp. 335-337, 1979.

[5] Coles, L. S.,A. M. Robb, P. L・ Sinclai:r, M・H。 Smi’ヒh and

    R. R. Sobek, Decision Analysis for an Experimen’ヒal Robot wi’ヒh

    an Unreliable Sensors, Proc. of 4th IJCAZ, pp. 749-757, 1975.

[6] K:anayama, Y. et.a’ヒ.’ A Self-Con’ヒained Robo’ヒ ”Yamabiko”, Proc.

    3rd USA-JAPAN Comput二er Conference, PP. 246-250, 1979・

[7] 工iゴima, J. e・ヒ.al., Locomotion Control System for Mobile Robo’ヒs’

    Proc. of 7th IJCA工, 1981.

一 IO 一

Page 13: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

Figure Z Yamabico 3.1

brain.._一_へ

reiprocessor Imemory

: =一

console t;6i颪㎜i;瀟、

interface

bus

supersonictransmitter/receiver ×4

step motordriver

lrefgrencei

   コ       コ

pos■t■onde・ヒector

eye+neck

processor 匡璽]ド司bus

dc motorcontroUer

:婁:副匿轟壷

left  、つ

legs

    rot二ary   I-glggf2ggc o d e.F..j

right         ”

Figure 2 The architecture of Yamabico 3.1

一 ll 一

Page 14: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

Brain

Leg

userls program

monitor

.CO㎜an

monitor、

Figure 3 The brain and the leg system

y

wa11/////.1ん!/宏/

’隊

. ヴ

/ん/)∴堕冊’・!/’

・ /一’ //

.一

robot pathX

robot二

Figure 4 Walking along a wall

一 12 一

Page 15: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

150((rm)

IOO

50

坐、

   ee  ee

  no echo   \

 サ   び ロ             の    ロ

「 ゆ f  o φ!.「㌔  ,ρ94’   o             ■   ●

   胆

  9   ◎ 

.  ●o .

          wa 11 ・一

二1

.”∵…‘、御∴㌔“ゼ’◎∴ P…畠∵..….

..

         の コ  の      の●        . θ●■o●OG● ■ ■    σ●・ψG・  ■曽● ●●09 ●9  ●

TA:

     N

    noise da・ヒa____う ・

Txansient Area

  o

Figure’ T.

           50 IOO                                  (em)・

Uneven waユls scanned by a moving eye

      一 13 一

Page 16: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

150(cm)

                      1                      :

100t        ・1          ,.」.。.。・。...・。・.・9.;.∴.1。…・,..,∴.・e.。・ρ9;

   り              のののワ  コ /!1’〆∵’●,。

       9.●・

?E「㌔・f・∴ら・●●。・・「,・. /

50マ

y

 l

 I

 l

 t

 l

 竃

 1

 1

 し冷\/L,ol.... .・ll:ll.1)f

 X.L ”e“e,“e s.・

  Nu.’V”’

一一一一一一一一一一一

。 50 100 (cm)

一) X

Flgure 6 A理工よ鳳ζL黛蔓翌愚孕pe¢.1. 1 frQnt

   as’ seen by a moving eyeof a waU,

一 14 一

Page 17: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

no echoes

100tl

   翫蝋畑ノ.噌‘..1.1謡%儲・

  1  き

  l

  I  l

  i

50も  i

    lll    川    田    ”1    川    川    駐l    lit     ●

・・.・G纏

t一’ 撃奄

    iii

  川      ll口 Il  lll      l1口 ll  lll     旧I ll

淵/争1編ンii輯に塑互 lt lI

 魏■ 1口l tl tl ll }l

e 1!lI

 l日l ll日 tl lI

 I口l IH l H ll

 日Il 亀lll lt lt

Il

ll

ll

uull

y

the posi『ヒion of feet

。 50 100 (cm)

一〉 x

Figure        弼ぜ      ノ                   し

7全艦9蒙~躍手瀦,fgggt of a wallr

一 15 一

Page 18: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

//ノ/、奪

1,。讐

//苓/、

1co 9.0 80

iXO X /ア

÷L

           ’

!//㍗菅・謡r据’ダ/ブ

 .・ 60

^ soi’ / /, ,i

/46 ,/ / /

  150    ×

 160  ×

書70\

measuredvalues’

the eye      Xa1

 つ

wall ×一

4g cm

      1/・se

      t / tc     オ     ”i  1 IO

      i

      !畠

Figure 8 A wall, looked at by a:ro七a’ting sonic eye

一 16 一

Page 19: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

150(㎝)

an image of a

{100 l l

〆1・漏怖1・〆・伽唖ノ1.,’w/

/      ノ

Dr6●幽鱒‘●’

1

1

 ●,

50

o

Figure 9.

iooo

Walking pe’rson,

person

a wall』↓ソ  ’ ノノ//

             け ゆユ         

  \      2000               (mse6)

as seen by a fixed eye

一 17 一一

Page 20: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

 Y

150a(em)

IOO

50

   ら                      ●  唇 』

    りつ          の  の       り       ゆ   つ  り           りぼ    ら        ロ  らの う

9。● u… , 。・P弓.●。 ●・●.●/

                /一一

.”j二〆弾.ll%『〆:●轡ンγ9グ,

む                         つ   サほ     の

。9G●働,。鵬.●● B● 秩D。.∂●・○・●.● ●f・・○●!

 9

t

L一一一一一一ewMin

  o

Figure 10.

        __i                     し

           50 ・ IOO                                      X                                 (em)

Uneven waUs after noise elimination            爆・9t,_.一...』』t;.“,._・_・…v・.’…転く...ジ7融1㌧こ,國』∴一… 「、r・、一く…』.・・一・・・・・・・…一・・

一 18 一

Page 21: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

y

y=?

● 8畢

1曹

x-N XN

y= ax +b

x

Figure li Least squares fi・ヒti ng of 2N+l observa’ヒions

.一@19 一

Page 22: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

(a) declaration

      const maxm瓢 101ラ

            コ                                        ロ

                          ロ      Var S■q, m, n, コ :■n’ヒeger;

           s㎜y’sumxy, sumyy, y, dx, xO:realア

           z:array[1・・maxm】Of real,

(b) initialiZatiOn

      begin sumy:=0;sumxy:=0アsumyy==0ア

            sig := 0フ

           m:・=2★n+1ア

            xO :・= n★・dx,

     end f.rゴ’:=・’里嘩Z[II:10

(c) updating with new y

     begin sumyyドs㎜yy+sqr(y)一sqr(z[m])l

           sumxy:=sumxy+・xO*y+xO *・ z・[m]一(sumy-z[m])★dx;

           SUmy:=SUmy+y-Z[m]ア

           主皇y≠Othen sig:=sig+1ア

           主皇z[m]≠0牟hen串ig・ニs⊥9-1ラ

           forゴ:=mdownto 2 do z[」]:=z[」_1]ア

           Z.[1]=y

     end

:Figure l2  An a工go:rithm for leas七 squares fittj.ng’

一 20 一

Page 23: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

s

30000(crn2)

20000

10000

Y

IOO‘

・一一一 x× 一一一 s

50

ite‘ D一 c鱒● 恃芙`●...・...の。.。・q●・.・。の、・…

   e  一一 一 一  一    一e 一   一       一     一  e  e e 一  一 一t 一 一 一e 一 一t一 一i .  一 一一 b  e “ e一一一t e一   一 一 一   一

t一

Pt   メk

 x   x

X x x pt

  x ゴ

  潭

x

sum of sguaresof residuals

E

1;.’・.

i・’・._・...・,。・・・…。.・’・’・●’一・。

o

妙勾!1・

メ       メ

 燕ン. x)cy’x

メ ’♪〆

x        x

Figure 13.

50

A resuLt of wall recognition

Lt一一一一一)t)ee‘,“‘,e(ncF)eee,c,eoec,eE”・・一

100    (cm) X

一 21 一

Page 24: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

L

/wa l // /

////.1 /

/ / !/ ! / / /

        ●盾b唐?窒魔≠煤FLon lo2

C

一一一i置 t

_._一黶Q_ 一 一 一    一

�レt

?・き茎a竃貧er・b・t輩

l         P21         1{←一一e 一一一一’一’一一一一一一一 1

               ロ       ロ

pl:current pos■t■on

P2: predicted po『i七ion

P>lm。dified predic七ed p。siti。n

                                               のmodifica・ヒion of the walking・d■rec●ヒュon

v

Figure 14

一 22 一

Page 25: AND ITS UNDERSTANDING OF A SIMPLE WORLD...A simple method is proposed for n.oise eliminat二ion Then the robo七1s images of a simple world are described and a method by which the

.I NSTITし「「E OF ELECTRON I CS AND INFORfvlATION SC I ENCE

            UN I VERS I-rv OF TSUKUBA

   SAKURA-iVl,UR/A:,, N i I HAR I-GUN, I BARAKI JAPAN

RE[PORT DOCUIViErvrAT I ON PAGEREPORT NUMBER

    工SE-TR・一81-22

工工田L:E

A Mobile Robot with Sonic’Sensorsand i’ヒs Underst:anding of a Simple World

AUTHOR(s)

   Yutaka Kanayama      Un■vers■ty of Tsukubar Sakura, 工baraki 305 Japan

Shin’ichi Yuta   Unlversi’ty of Tsukuba, Sakura, 1baraki 305 Japan

Jun冒ichi 工iゴima   The University of Electrocommunications,   Chofu, Tokyo 182 Japan

REPORT DATE

February 27, 1981

地域IN CATEGORY

Ar’ヒificial 工ntelligence

NUMBER OF PAGES

22

CR CATEGOR工ES

3・6r 5・1

KEY WORDS

Intelligent Robots, Mobile Robots, World-Understanding

Sonic Sensors, Leas’ヒ Squares Fitting

1t

ABSTRACT        An ■ntelliqent self-contained.:robot, named Ya皿abico 3.l wasdevel・ped under a pr。ゴect wh・se purp。se was, in part t。 c。nstructa.r・b・t which kn。ws its p。si・ti。n in the w。rld. The r・bQt ha$tw。

m■c Mocomputer systems for七he brain and the legs, and a co㎜un↓一catエ。n$ystem between the tw・・It has f。ur supers。nic eyes,. a濫sa融a齢器モs=誌言u婁ぎ:七9:。:舗彰謡t亨w・dr’v’ng

  Second’ a method for the robot to recognize a simp工e wqrld isdescribed1・Asimple meth。d is pr。P・sed f。r n・ise eユiminati。n。Then the robot l s images of a simple world ate descri:bed and ameth。d by which the r。b。t can rec・qnize straight wa「 撃撃刀@is

P「oposed・ :Least squares fitting is used fo:r the discriminatiOn.This pr・cedure qi▽es七he distance七・the wall and the directi。n:f脇tWこき搬h翻ε§c茎。茎Os鵠9豊{蓋9蹴£臨蹴・wa互ln呈1・y’

P「oposed。 The sof’ヒwa:re for all o.f this is a par・ヒ of the baSiCbra■n monitor of Yamabico 3.1。

SUPPLETv[ENTAR¥ NOTES