Embed Size (px)
Citation preview
Publ ished by the Ad vanced Robo tic Systems Int ernational an d p ro lite ra tu r Ve rla g
p lV pro lite ra tu r Verlag Ro bert Mayer-Scho lz
Mammend o rf
German y
Abs tra cting and no n-p rof it use of the ma te rial is permitted w ith cred it to th e so ur ce . Sta tements and
opinions ex p ressed in the chap te rs ar e these o f the indi vidua l con tribu to rs and no t necessa rily those of
the edi to rs o r p ub lisher. No respons ibility is accepted for th e acc u racy o f in fo rm atio n con ta ine d in the
pub lishe d a rticles. Publ ishe r assu mes no responsibility liability fo r an y d am age o r injury to pe rsons o r
property a ris ing out o f the use of any mat e rials, inst ructions, me thods o r id eas co nta ined ins ide . Afte r
th is wo rk has been pub lished by the A dvan ced Robotic Sys tem s Int ernati ona l, au tho rs have the right to
re publish it, in w ho le o r part, in any p ubl ica tion of w hich they a re an au tho r o r ed ito r, and the make
o the r persona l use of th e work
© 2007 Advanced Robo tic Sys tems Int ernation al
ww w .a rs-jou rna l.co rn
Addit ion a l cop ies can be obta ined from:
pll b1ica tione-a rs-jou rn a I.com
First published M ar ch 2007
Printed in Croa tia
A ca tal og reco rd for th is book is ava ilable from the Ge rma n Lib ra ry .
M obile Robo ts, Per cept ion & Naviga tio n, Edi te d by Sascha Kolski
p. em.
ISBN 3-86611 -283-1
1. Mobile Robotics . 2. Perception. 3. Sas cha Kolsk i
Mobile Robots Perception & Navigation
Edited by Sascha Kolski
pro Iiteratur Verlag
v
Preface
Mobil e Robot ics is an active research area where researchers from all over the world find new technologies to im pr ove mobile robots int elligence and areas of applica tion . Today robo ts navi gate au tono mous ly in office env ironme nts as we ll as outdoors. They show their abi lity to besid e mec ha nica l and electronic barri ers in build ing mobile platfor ms, perceiving the environme n t and d ecid ing on how to act in a given situa tion are crucial p robl ems. In this book we focu sed on these two areas of mobile robo tics, Perception and Naviga tion .
Perception includes a ll means of collecting informa tion abo ut the robo t itse lf and it's envi ronment. To make ro bot s move in their sur rounding and int eract with their enviro nment in a reasonable way, it is cruc ial to und erstand the actual situa tion the robot faces.
Robots use se nsors to measu re properties of the env iro nment and interpre t these measureme nts to ga ther kno w ledge needed for save interaction . Senso rs used in the wo rk described in the articles in thi s boo k include compu ter vision, ra nge find ers, so nars and tactile sensors and the way those sens ors can be used to allow the robot the perception of it' s envi ronment an d enabling it to safe ly accomplishing it' s task. The re is also a number of contribu tions that show how measu rements fro m different sensors can be comb ined to ga the r more re liable and accura te inform ati on as a single senso r could provide, th is is especially efficient when sensors are complementary on their stre ngths and weaknesses.
As for man y robo t tasks mobi lity is an important issu e, robots have to navigate their env ironments in a sa fe and reasonab le way. Naviga tion describes, in the field of mob ile robotics, techniques that allow a robot to use informat ion it has ga the red abou t the env ironment to reach goa ls that are given a pr iory or derived from a higher level task descript ion in an effectiv e and effic ient way .
The main ques tion of navigation is how to ge t from where we are to where we want to be . Researchers wo rk on that ques tion since the early days of mob ile robo tics and have developed many so lu tions to the pr oblem con sidering different robo t env ironments. Those include indoor env iro nmen ts, as well is in mu ch large r scale ou tdoor environments and un der water navigat ion .
Besid e the ques tion of global navigation, how to ge t from A to B navigation in mob ile robotics has loca l aspec ts. Depend ing on the architecture of a mobile robo t (d iffere ntial d rive, car like, submarine, plain, etc.) the robo t's possibl e actions are cons trai ned no t on ly by the ro
VI
bats' env ironme n t but by its dynamics. Robot motion planning takes these dyn amics int o account to choose feasible actions and thus ensure a sa fe moti on.
Th is book ·gives a w ide overview ove r di fferent na vigation techniques describing both naviga tion techniques d ealing with loca l an d contro l aspec ts of navigat ion as well es those handling glo bal navigation aspec ts of a single robot and eve n for a gro up of robots.
As not only thi s book shows, mobile robot ics is a livin g and exciting field of research combining man y di fferent ideas and approach es to build mechatroni cal sys tems ab le to in terac t with their environ ment.
Edi tor Sascha Kolski
VII
Contents
Perception
1. Robot egomotion from the deformation of active contours Gu illem Alen ya and Ca rme Torras
01
2. Visually Guided Robotics Using Conformal Geometric Computing Eduard o Bayr o-Cor rochano, Luis Eduardo Falcon-Morales and Julio Zamora-Esquivel
19
3. One approach To The Fusion Of Inertial Navigation And Dynamic Vision : Stev ica Gr aovac
45
4. Sonar Sensor Interpretation and Infrared Image Fusion for Mobile Robotics ............. Mark Hinders, Wen Gao and William Fehlman
69
5. Obstacle Detection Based on Fusion Between Stereovision and 20 Laser Scanner Raphael Lab ayr ad e, Dominique Gruye r, Cyril Royere, Mathias Perrollaz and Didier Aub ert
91
6. Optical Th ree-axis Tactile Sensor Ohka, M.
111
7. Vis ion Based Tactile Sensor Using Transparent Elastic Fingertip for Dexterous Handling 137 Gora Obinata, Dutta Ashi sh, Norinao Watanabe and Nobuhiko Mor iyama
8. Accurate color classification and segmentation for mobile robots. ....... 149 Raziel Alvarez, Erik Millan, Alejand ro Aceves-Lope z and Ricard o Swa in-Oropeza
9. Intelligent Global Vision for Teams of Mobile Robots 165 Jacky Baltes and John Ande rson
10. Contour Extraction and Compression-Selected Topics 187 And rzej Dziech
VIII
Navigation
11. Comparative Analysis of Mobile Robot Localization Methods Based On Proprioceptive and Exteroceptive Sensors................................. Gianluca Ip politi, Leop oldo [etto, Sauro Longhi and Andrea Monteriu
215
12. Composite Models for Mobile Robot Offline Path Planning Ellip s Maseh ian and M. R. Ami n-Nase ri
237
13. Global Navigation of Assistant Robots using Partially Observable Markov Decision Processes....................................................... Ma ria Elen a Lop ez, RafaelBarca, Lu is Migue l Berga sa, Manuel Ocana an d Maria Soledad Escudero
263
14. Robust Autonomous Navigation and World Representation in Outdoor Environments Favio Masson, Ju an Nieto, Jose Gu ivant and Eduardo Nebo t
299
15. Unified Dynamics-based Motion Planning Algorithm for Autonomous Underwater Vehicle-Manipulator Systems (UVMS) Taru n K. Pod de r and Nilanjan Sarkar
321
16. Optimal Velocity Planning of Wheeled Mobile Robots on Specific Paths in Static and Dynamic Environments Maria Pr ado
357
17. Autonomous Navigation of Indoor Mobile Robot Using Global Ultrasonic System Soo-Yeong Yi and Byoung-Woo k Choi
383 ·
18. Distance Feedback Travel Aid Haptic Display Design Hide yasu Sumiya
395
19. Efficient Data Association Approach to Simultaneous Localization and Map Building : Sen Zhang, Lihu a Xie and Mar tin David Adams
413
20. A Generalized Robot Path Planning Approach Without The Cspace Calculation Yong ji Wan g, Ma tthe w Car tmell, QingWan g and Qi uming Tao
433
21. A pursuit-rendezvous approach for robotic tracking ......... Feth i Belkhouc he an d Bou medien e Belkhouche
461
22. Sensor-based Global Planning for Mobile Manipulators Navigation using Voronoi Diagram and Fast Marching S. Garrido , D. Blanco, M.L. Munoz, L. Moreno and M. Abd er rahim
479
IX
23. Effective method for autonomous Simultaneous Localization and Map building in unknown indoor environments............................ Y.L. Ip, A.B. Rad, and Y.K. Wong
497
24. Motion planning and reconfiguration for systems of multiple objects ........................ Adrian Dumi trescu
523
25. Symbolic trajectory description in mobile robotics Pradel Gilbert and Cal eanu Ca talin-Dan iel
543
26. Robot Mapping and Navigation by Fusing Sensory Information Maki K. Habib
571
27. Intelligent Control of AC Induction Motors Hosein Marzi
595
28. Optimal Path Planning of Multiple Mobile Robots for Sample Collection on a Planetary Surface j .C. Cardema and P.K.e. Wang
605
29. Multi Robotic Conflict Resolution by Cooperative Velocity and Direction Control. Satis h Pedduri and K Madhava Krishna
637
30. Robot Collaboration For Simultaneous Map Building and Localization M. Oussalah and D. Wilson
667
17
Autonomous Navigation of Indoor Mobile Robot Using Global
Ultrasonic System
Soo-Yeong Yi, Byoung-Wook Cho i
Dept. of Electrical Engineering, Seoul National University of Teclmology,
Republic of Korea
1. Introduction One of distinctive features of the recent intelligent autonomous robo t from the conventional indu strial robot is the mobility, which makes the robot overcome the limited workspace and expands it arbitrary . For au tonomous navigation in workspace, a m obile robo t should be able to figure out where it is and w hat direction it moves towards, which is called the self-localization (Singh & Keller, 1991). The self-localization capability is the most basic requirement for mobile robots, since it is the basis of the on-line trajectory planning and contro l. The trajectory error of a dead -reckoning navigation, which re lies only on the in ternal sensor such as the odometer or the enco de r, grows with time a nd d istan ce. Therefore, an externa l sensor is necessary in order to local ize the position of the robot in the workspace and to compensa te it for the trajectory erro r. Among the severa l alterna tives, the ul trasonic sensor is regarded as the most cos t-effective ex ternal sensor and it is widely used for gener al purposes (Kuc & Siege l, 1987). The methods of the se lf-loca liza tion using the external sensors can be classified in to two gro ups: local methods and globa l methods. (1) In the local method, a mobile robot makes a local object map using the relat ive dis tance data from the env iro nme n ta l objects and matches the local map with a globa l map da tabase. As a result, the mobi le robo t figures ou t its own posi tion in the wor kspace. (2) In the global method , the mobi le robo t computes its position directly in the global coordina tes using the d istan ces from some referen ce position s in the workspace. The local method has so me advantages in that collision-avoida nce mo tion and mapreco nstruction for the tran sformed env ironme nt are mad e possibl e by using the distance sensors as well as the self-loca liza tion system. However, this req ui res massive com pu tations in terms of the local map-making and the mat ching processes w ith the globa l map datab ase. In the extre me case, the robot has to stop movin g momentarily in order to obtain the necessary en vironmental informat ion (Leonar d & Durrant-Wh yte, 1992) (Ko et al., 1996). On the other hand, in the globa l method , the local map-making and the match ing processes are avoidable and the self-loca lization is com putationa lly efficient and fast (Leonard & Durrante-Whyte, 1991) (Hernande z et al., 2003). A globa l ul trasoni c sys tem presented in th is cha p ter is a kind of a pseudo-lite system wi th a
384 Mobile Robots , Perception & Navigation
well-kn own CPS like stru ctu re (Haih an g et al., 1997). For the se lf-localization of an indoor mobil e robot, the global ultrasonic sys tem con sists of four or more ultrasonic gene rators fixed at referenc e position s in global coord inates and two receivers mounted on the mobile robot. Based on the distance measurements between the ultrasoni c generators and the receiver s and the appropriate data fusion algorithm for the d ist an ce data, it is possible to compu te the position of the mobile rob ot in global coord inates. Wh en seve ral ultrasonic gene ra tors are used under a system, the following probl ems should be taken into con sid eration;
(1) Cross talk between ultrasonic signals (2) Identificat ion of each ultrason ic signa l (3) Synchroniza tion between each ultrasonic genera tor and the receiver to cou nt the
TOF (Time-Of-Flight) of ultrasonic Sign al In orde r to solve the above problems, a small-sized RF (Radio Frequenc y) module is adde d to th e ultrasonic se nso rs and the RF ca lling signa l is tran smitt ed from the ultrasoni c receiv er side, i.e . the mobil e robot. By using th is config ura tion, the robot is able to contro l triggering time and seque nce of u ltrasonic sig na l ge ne ration as we ll as to sy nchronize the ultrasoni c se nsors, so that to localize its ow n position in the global coo rd inates . In thi s cha p ter, we propose a global ultrasonic sys tem and adopt the EKF (Extend ed Kalman Filter) algorithm designed for the self-locali zation. The performance of the au tonomo us navigation syst em based on the self-loca liza tion is ver ified throu gh ext ensive experiments.
2. A global ultrasonic system
The overall structure of the global ultrasonic system is depi cted in Fig. 1. The ultrasonic generators are fixed at known position s, T =[x. v, z. ]I , i = I ... 4 in the wo rk space, e.g. at
I I' .. PI' ~
each corn er of the ceiling. Usin g the front and the rear ultrasonic sens ors situated at PI .an d
p,. ' the mobi le robo t receives the ultrasonic signal and comp utes the distances by counting
the TO F of the sig na l. It is conveniently assumed in Fig. 1 that the number of ult rason ic genera tors is four, which can be increased as needed in conside ra tion of the wo rks pace size and objects in the immediate environme nt. In order to avoi d cross-talk bet ween the ultrason ic signals and to synchronize the ultrasonic receivers with the generators, the RF receivers, RX - RX , and the RF transmitter, TX , are adde d to the ultrason ic ahenera tors
1 4 . <
and the ultrasonic receivers, respec tively. By using the RF cha nnel, the mobile robot seque ntia lly activ ates each one of the ultrason ic genera tors in success ive time slots. Ass uming that the delivery time for the RF calling signal is negligibl e, the ultrasonic signa l genera tion occurs simultaneously with the RF calling signal transmission and it is possibl e to syn chroni ze the ultrasonic genera tors with the receivers. In Fig . 1, h _ h . deno te the
. (. 1 !.4
d istan ce between T - T4
and P . The di stan ces, h - h , be tween T - T and p are 1 . f ,..1 ,..4 1 4 r
omi tted for brevity. The positions of the ultrason ic receivers on the robot, P _ [y. ". ~. -I' I , f - " ( ' I ( ' ~ ( c
and P, =[xn
V,., Z,. ] I, wi th respect to the center position of the mobile rob ot, P =[x, y, z..J{r
can be described as follows:
385 Autonomous Navigat ion of Indoor Mobile Robot Using Global Ultrasonic System
X+ /COSOl [X-/eaSB] (1) PI = Y + ~: inB Pr = Y - ~:in B
[
whe re I represents the distance bet ween the center po sition of the mobile robot and the ultrason ic receiver , and B denotes the heading angle of the mobile robot. It is ass umed that the moving surface is flat, so that the Z compo nent of the pos ition vec tors is cons tant as Zc
in (1).
tJ, ax: "X'rJ T · ' r ,@ . " ~~~.-®
/ / ' ',
/
x
Fig. 1. A global ultrasonic system.
3. The EKF for the self-localization and the autonomous navigation algorithm
The position vector in the x - y plane, r =[x, YJ" together wi th the head ing angle, 0, of a
mobile robot ha vin g differential wheels, follows the sta te equation (2) in the di screte-time domain (Fox et al., 1997):
Xk + TUk COSBk] [Yk +TuksinOk
. xk +Pk cosBksin(Tcod
[ ~k+ l ]= - Pk sinOdl -cos(Tcod)
xk + Pk cosBksin(TU\)
J k+1
(2) +Pksin Ok(l -cos(Tcok ))
0k+l =Bk -Tro,
whe re the subscrip t k is the time index, T , denotes the sampling inter val , uk and co arek
the linear and the angu lar ve locities of the rob ot respectively, and - !:!L represents the Pk
OJk
rad ius of rotation . The positi on vector and the heading an gle of the mobile robot are aug me nted so as to become P = [x,Y, o] I , which is referred to as the robot posture. The bold
and normal symbols represent the vector and the scalar variables, resp ectively.
386 Mobile Robots, Perception & Navigation
As a consequence of (1) and (2), the state equation for the ultrasonic receivers on the robot can be d escribed as follows:
rf.k+' = f j (rj k· uk' qk )
Xj ,k+TUkCOS()k +ql,k ]
[Yj,k + TUk sin ()k + 'h:« (3-1)
x/, k - (I cos()k + Pksin()k)(1 - cos( Twk))
+(-I sin()k + Pk cos()k )sin(T wk) +ql,k
YO +(-/ sin()k +Pkcos()k)(I-cos(Twk))
+ (I cos ()k +Pksin ()k )sin(T Wk)+q2,k
r-i -: " j,,(rr.k' uk' qk)
Xr'k +TUkcos()k +ql'k ] [Yr,k +TUk sin ()k +q2,k (3-2)
xr,k-(lcos()k+ Pk sin ()k )(I -cos(Twk)) =
+(-/ sin()k +Pkcos()k) sin(Twk)+ql k
Yr,k +( -/sin ()k + Pk cos()k)(l- cos(Twk))
+(1cos()k+Pk sin ()k) sin(Twk) +q2.k
where r - [x y ] ' and r = [x )1]' represent the positions of the front and rear f - . / ' .r r r , r
ultrasonic receivers respectively and q =[q q r is the Gaussian random noise with , k I,k' 2,k
zero mean and Q varia nce . The ' measurement equation at the ultrasonic receivers can be
modeled as follows:
Zj ,k=: h/,i(rj ,k' vk) (4-1) = {(xO -xi +(Yr,k - yi +(zc- z}}1/2+Vk
Zr.k =h,) rr,k' Vk) (4-2) = {(x - x-)2+(y -y.)2+(z - z. )2}1 /2+vr,k I rk I e , k
where the measurement noise, V is assumed to be Gaussia ~1 wi th zero mean and G k,
variance, and the subscrip t, i , denotes one of the ultrasonic genera tors, T - T", which is 1
called by the mobile robot a t tim e k .
From the state Eq. (3-1) and the measurement Eq. (4-1), it is possible to ge t the followin g set of equati ons constituting the EKF estimation for the front ultrasonic receiver position:
rr .k+1 = f r(rr,k' uk' 0) (5)
V;.k +1 =Ar.k Vr.k A ' r.k + Q. . . . .
387 Autonomous Navigation of Indoor Mobile Robot Using Global Ultrasonic System
Kf .k = Vf,kHr / (Hr ,kVj,k H r / + cr l
(6) "!» = (/ - K f,kHf .k) Vf,k
r r .k = r f,rtKf .k(Zr .k - hr)rj,k ' 0))
wher e K . is the kalman filter gain, r- k and i f k represents the a priori and a posteriori f~ f "
es tima tions for r , , res pec tively, and V - and V represent the a priori and a posteriori f .k f .k . ,k r
error covariance matrices, respectively, as defined in (7).
r..= E [(rf,k - rj,k )(rf,k - rj,k) '] (7)
Vf,k = E[(r/ .k-r/ .k)(r/ .k -r/.k) ' ]
where EO denotes the expecta tion of the corres ponding random va riables. The Jacobian
matrices, A and H , in (6) ar e given as follows: I .k f .k
(8)
(9)
Yr ,k - Yi] Dr·. •1
wh er e D . . is defined by the following Eq. (10). / ./
(10)
The EKF es timation, rr.k' for the rear ultrasonic receiver position is similar and om itted here
for the sake of brevity. From r and r the posture estimation for the mobile robot can be described as follow s:
/ ,k r,k I
(11)
Assuming th at the estimation error covariances for positions of the front and the rear ultrasonic receiver are the same, the error covariances of the posture es timation are give n in (12) as shown in Fig. 2.
Mobile Robots , Perception & Navigation 388
i : V .k = p E [(r -rk)(r-l\ l2]
= V/ .k ( = Vr •k ) (12)
V =E [(e -ek)2] e.:
V . =: tan " '---1Ji.. . I
Eq. (12) implies that the es tima tion for heading an gle becomes more accurate accord ing to the di stan ce between the tw o ultrasonic receivers. Based on the self-loca liza tion given in (11), a simple contro l inpu t, v and OJ , to dri ve the mobil e robot toward the given goal position ,
k k
-can be written as (13).
y
%
Fig. 2. Error covari anc es of the posture es timation.
. (13)
OJ,. =ke(e,l.k' - ek.) , e =tan- 1Yg - Yk > d .k " _ ~.
rs "k
where c and k are positiv e cons tants. The mobile robot adjusts its heading ang le toward e the intended positi on and moves with the constant velocity as d epi cted in Fig. 3.
y
,• I
I , I
,
, d. = t~ ----.-~,'~ - 1 y~ - y; " , Xg- Xk
I , I ,
I _
", .. ~;
1,-' ,
x
Fig. 3. Naviga tion con trol.
389 Autonomous Navigation of Indoor Mobile Robot Using Global Ultrasonic System
4. Experiments and Discussions
In order to ver ify the performance of the EKF based se lf-localization and autonomous navigation syst em using the global ultrasoni c system, a simple expe rime ntal se t-up was es tablished as shown in Fig. 4, which has dimension of 1,500 mm and 1,500 mm in width and length respectively and 2,500 mm in height. The ultrasonic generators installed with the RF receivers are fixed near the four corners of the ceiling and their po sit ion s are descr ibed in (14).
Fig. 4. Exp erimental set-up .
T, =[10.0 ,10 .0 , 2360.0r
(14)T2 =[1427.0,5.0, 2370.0r
T3 =[1423.0 ,1445.0,2357.0]'
T4 =[0.0 , 1380.0 , 2370.0]'
At first, a pr eliminary experime nt was carried out for the ultrasonic calibra tion and the result is presented in Fig. 5.
2800 : !
2600 ~
2400
12200 ~
t: IE IID 2000 ~ o c:: ro en l 800 l '0 i Cii CJ 0:: 1600 ~
1400 ~ i
1200 ~
! 1000 '
i • data 1- linear
! 800 i
3000 3500 4000 4500 5000 5500 6000 6500 7000 7500 8000
TOF (time'of night)
Fig. 5. Real di stance with respect to ultrasonic TOF.
390 Mobile Robots, Perception & Navigation
The linea r equation relating the ultrasonic TOF to the real distance is give n in (15), as obtained from the least-square method, and the var iance of the measu rement noise is specified as (16).
D = 0.34533·T - 57.224 (15)
G = 1.8 (16)
where D(mm) represents the real d istance between the ultrasonic ge ne rator and the receiver
and T(fl sec) is the ultrasonic TOF.
Fig. 6 shows the resu lts of the self-loca lization expe rime nt, in whic h the robot is moved manually from the initial posture, (x, y, B) = (600, 600, 0) to the goal posture,
(900, 900, J( / 2) at 45 sec. The initi al value of the posture estimation is set arbitrarily as
(650, 650, 0). The di stanc e and the heading ang le ar e described by mm and rad.,
respecti vely. As sho wn in Fig. 6, the position error s in the xand y axes are less than 25 111m
in the steady-state. Since the distance between the cen ter po siti on of the robot and the ultrasonic receiver is d esigned as 1=75 mm , the es tima tion error of the heading angle in (12)
becomes tan- 1(25/7 5) "" 0.32 rad . as shown in Fig. 6 (c).
.- 'est jm at~ ' 00 • 0 acilial r:' :65<>
800
~ e .,g 750)
~ X 700
650 \
"-- .-.--L 600 - - --- _-- - • .:V- ko": J
~ 5')O!:-~, 0:---2O ~~ O--:~~ ;:----:~--:Y.l . O:--:--,~ OO 70--,OO 90
' 00
65"
E 600
~ .g 750 . ~
c a,
700 >
65<>
600
55"
Timetsec.)
(a) Position estimation in x axis.
eslimaled - --- --------_ ..-------- --.. - -. -- _ _ 0 actual ,
,,,, ,
I' \...--
0 '0 3Q 40 so eo 70 80 90 " Time( sec.)
(b) Position estima tion in y axis.
391 Autonomous Navigation of Indoor Mobile Robot Using Global Ultrason ic System
)
;r '=" ';~i~';rl'~- i L~~. ~_·...~~~.r...........I
2
I
(- -- ------ ---0 -- -._----- . - - .. -. --_._.. _._.
.,
.,
.a ,Ii) 20 )('l ~ , ..0 e-o , J ~ :,1')
Timc(scc.}
(c) Estima tion for heading 0 ang le. Fig. 6. The self-localiza tion of the mobile robot.
The autonomous navi gation system using the global ultrasonic system is compared to the dead -reckoning navi gation system on the straight line connect ing the initi al posture, (650, 650, Jr/4) , and the goal posture, (900,900, Jr/4 ), in the workspace.
(a) Position in x and y axis
(b) Heading an gle
392 Mobile Robots, Perception & Navigation
y (900.900)
x'Y,I---------------.
(c) Trajectory in x- y plane
Fig. 7. The d ead-reckoning navigati on .
Fig. 7 shows the res ults in the case of the dead -reckoning navigati on , in which the mobi le ro bot cannot reach its goa l posture, due to the uncert ainties in the sta te equa tion. In Fig. 7 (c), the dotted polygon s represent the desired postures of the mobile robot with respect to tim e. The results of the autonomo us na vigati on system based on the self-localizat ion using the g loba l ultrasonic sys tem are prese nte d in Fig. 8 for the same in itial and goal postures. As shown in thi s figure, the mobile robo t reac hes the goa l posture, overcoming the
u ncertainties in the state equa tion, and the head in g ang le at the final position is aro und % as desired . It sho u ld be noted that the posture da ta in Figs. 7 and 8 are ob tained by using the global ultrasonic system also, thus these values may be d ifferent from the actual postures to some degree.
950r-- --.-- ----.-- - .--- --.-- ----.-- - .--- --.-- ---,
0;::
.sc
c .g 'iii c a.
600 '-_ ....1.-_ ---'--_ ---'__-,'-_---'--_---''-_....1.-_--' o 10 15 25 30 3S ' 0
(a) Position in x and y axis
393 Autonomou s Navigat ion of Indoor Mobile Robot Using Global Ultrasonic System
,"
(b) Heading angle
(900 .900>,
, (650.650 ) .~.
·l::..i ii--- - - - - - - - - - - - ......
(c) Trajectory in x - y plane
Fig. 8. Na vigation with global ultrasonic sys tem.
The size of the ultrasoni c region in the work space is dependant on the beam- width of the ultrasoni c genera tor. In the case of a general ultrasonic ran gin g system, in whi ch both the signa l generator and the receiver are lumped together, an ultrasonic generator with a narrow beam-width is preferable in orde r to avoid the ambiguity and to enhance the measurem ent accuracy . On the oth er hand, the proposed globa l u ltrasonic system, whi ch has a di stributed signa l generator, requires the use of a wide beam-width genera tor, in order to expand the ultrasonic region in the work space.
5. Conclu sions
In this cha p ter, the global ultrasonic sys tem with an EKF algorithm is present ed for the selflocalizati on of an ind oor mobile robot. Also, the performanc e of the au tonomous navigation based on the se lf-localiza tion system is thus verifi ed through various experiments. The global ultrasonic system consists of four or more ultrasonic generators fixed at known posit ion s in the work space, two receivers mounted on the mobile robot, and RF modules added to the ultrasonic sensors. By controlling the ultrasonic signal generation through the RF cha nne l, the robot can synchronize and measure the distance between the ultrasonic generators and receivers, thereby estimating its own position and heading an gle . It is shown
394 Mobile Robots, Perception & Navigation
throu gh expe rimen ts that the estima tion errors are less than 25 mm in terms of the posi tion and less than 0.32 rod . in terms of the heading ang le. Since the estimation error of the heading angle is dependant on the distan ce between the two ul trasonic receivers on the ro bo t, it is possible to obtain a more accurate estima tion for the heading angle by increasing thi s d istance. The globa l ultrasonic sys tem has the follow ing sa lient features: (1) simple and efficient sta te es tima tion since the process of local map-making and ma tch ing wi th the global map d atabase is avoidable due to the GPS- like natu re of the sys tem, (2) active cuing of the u ltrasonic generation time an d sequence through the RF channel. and (3) robustness agai nst signal noise, since the ultrasonic receiver on the mo bile robot processes the signal received d irectly from the ge nera tor, instead of through an indirect reflected sig nal. In this cha p ter, it is ass umed that an ideal env ironme nt exis ts wi tho u t an y objec ts in the workspace. Env ironmenta l objects may result in an area of relative obscuri ty, which the ultrasoni c sig na ls canno t reach . It is possible to overcome the problems associated wi th environments containing obs tacles by increasing the number of ultrasoni c gene ra tors in the work space as needed. Th is enhance me nt is currentl y being studied .
6. References
Fox, D.; Burgard, W. & Thrun, S. (1997). The dynamic window ap proach to collision avoida nce, IEE E Robotics and Automation Magazine, Vol.d, No .1, March, pp .23-33, ISSN:1070-9932
H aih an g, S; Muhe, G. & Kezhong, H . (1997). An integra ted GPS/ CEPS position es tima tion sys tem for ou tdoor mobile robot, Proceedings of IEE E International Conference on Intelligent Processing Systems, Beijing, China, October, pp.28-31
He rnandez, S.; Torres, J.M, Morales, c.A. & Acos ta, L. (2003). A new low cos t sys tem for autonomo us robot heading and posit ion localization in a closed area, Autonomous Robots, Vol. 15, pp . 99-110, ISSN:0929-5593
Kleema n, L. (1992). Op tima l Estima tion of Position and Head ing for Mobile Robots Usin g Ultrasonic Beacon s and Dead-reckoning, Proceedings of IEEE Conference on Robotics and Au tomations, Nice, France, May, pp .2582-2587
Ko, J.; Kim, W. & Chung, M. (1996). A Method of Acous tic Lan dmar k Extrac tion for Mobile Robot Naviga tion, IEE E Transaction on Robotics and Automation, Vo1.l2, No.6, pp .478-485, ISSN :1552-3098
Leonard , J.; Durrant-Whyte, H. (1991). Mobile Robot Localization by Tracking Geometr ic Beacon s, IEE E Transaction on Robotics and Automation, Vol.7, No.3, pp.376-382, ISSN:1552-3098
Leona rd , J. & Durrant-Whyte, H. (1992). Directed sonar sensing for mob ile robo t navigation, Kluwer Academic Publishers, ISBN:0792392426
R. Kuc & SiegeL M.W. (1987). Ph ysically based simulation mod el for acous tic sensor robot navigat ion . IEEE Transaction on Pattern Analysis and. Machine Intelligence, Vo1.9, No.6, pp .766-777, ISSN :0162-8828
Sing h, S. & Keller , P. (1991). Obstacle detect ion for high speed au to nomous navigation . Proceedings of IEEE International Conferenceon Robotics and Automation, pp.2798-2805
