Implementation of Internet-Based Personal Robot with Internet Control Architecturekhhan.com/pdf/ICRA2001.pdf · 2003. 7. 1. · embodiment [6] and an intelligent telerobot [7] were

Implementation of Internet-Based Personal Robot with InternetControl Architecture

Kuk-Hyun Han, Shin Kim, Yong-Jae Kim, Seung-Eun Lee and Jong-Hwan KimDepartment of Electrical Engineering and Computer Science,

Korea Advanced Institute of Science and Technology (KAIST),

Kusong-dong, Yusong-gu, Taejon, 305-701, Republic of Korea

{khhan, skim, yjkim, selee and johkim}@vivaldi.kaist.ac.kr

http://vivaldi.kaist.ac.kr

Abstract

This paper describes the implementation of aninternet-based personal robot with novel direct inter-net control architecture which is insensitive to the in-herent internet time delay. The personal robot can becontrolled by using a simulator provided at a local site.However, a large internet time delay may make somecontrol inputs distorted. Moreover, since it is affectedby the number of the internet nodes and loads, this de-lay is variable and unpredictable. The proposed controlarchitecture guarantees that the personal robot can re-duce the path error and the time difference between avirtual robot at the local site and a real robot at the re-mote site. Simulations and experimental results in thereal internet environment demonstrate the effective-ness and applicability of the internet-based personalrobot with the proposed internet control architecture.

1 Introduction

Recently it has been observed that many re-searchers take interest in internet robotics because ofthe merits of internet which enables users to access anysystems on the network cheaply. The robot arm con-trol system [1] through a Web browser was designed,and TeleGarden system [2] and Mars Pathfinder [3]were developed. The sensor-based mobile robot sys-tem [4] which can be controlled by using a Webbrowser and the internet-based supervisory architec-ture [5] were reported. The concept of a personal tele-embodiment [6] and an intelligent telerobot [7] wereintroduced recently. Most of them have the supervi-sory control scheme which enables users to issue highlevel commands. The internet time delay is variableand unpredictable so that the design of a direct con-

trol scheme which enables users to control the motionof the robot continuously may not be easy. The di-rect control scheme [8] on the internet was proposed,but the modeling of the internet time delay was notadequate.

This paper describes the implementation of aninternet-based personal robot with novel internet con-trol architecture which guarantees that the personalrobot can reduce the path error and the time differ-ence between the actions of a virtual robot at the localsite and a real robot at the remote site. An internetuser can control the real robot using a simulator pro-vided at the local site, and can have information on thereal environment at the remote site since the simula-tor has a virtual environment. The path error and thetime difference in the internet-based personal robotsystem are caused by the unpredictable internet timedelay and the difference between the real environmentof the remote site and the virtual environment of thelocal site. It is not easy to model internet time delay,hence a control architecture that is insensitive to timedelay, is needed. Main components of the proposedinternet control architecture are a command filter torecover the information loss of control commands, apath generator and a path-following controller to re-duce the time difference between the real robot andthe virtual robot. The difference between the realrobot and the virtual robot model of the simulatorcan be overcome by a posture estimator. The problemcaused by the difference between the two environmentscan be solved by applying a virtual environment su-pervisor to the control architecture. We have alreadyproposed the control architecture with the virtual en-vironment supervisor [9], and it will not be mentionedin this paper. The graphic user interface(GUI) imple-mented with Java and the practical applications of theinternet-based personal robot were described in [10].

This paper is organized as follows. In Section 2, thedeveloped internet-based personal robot system, themodeling of a mobile robot, and the characteristics ofinternet time delay are described. In Section 3, thenovel internet control architecture is designed step bystep, which is insensitive to internet time delay. Sec-tion 4 presents the simulation results of the proposedinternet control architecture. Real experiments withthe developed internet-based personal robot are pro-vided to show the effectiveness and applicability of theproposed internet control architecture. Concluding re-marks follow in Section 5.

2 Internet-based Personal Robot

2.1 System description

The internet-based personal robot (IPR), a kind ofservice robot, can be used for a person’s convenientlife in a house/office or any indoor environment. Ithas a personal computer (PC) as a main part, and itcan obtain information about environmental changesby using vision cameras, sonar sensors, etc. Actua-tors enable the robot to move and to carry out somephysical work. It has a wireless LAN system for theinternet remote control. A user can control the IPRusing a simulator provided at a local site. It has theintelligence to gather the data from the sensors and toprocess them to decide its action.

The overall system consists of computers at the lo-cal sites, internet, wireless LAN system and the IPR.Users can access the IPR located at the remote sitevia internet using a computer at the local site. Thewireless LAN system connects the IPR to the internet.

At the local site to control the IPR, a remote con-trol architecture of the IPR system should be designedconsidering the inherent internet time delay. A basicremote control architecture of the IPR system consid-ered in this paper is described in the following. A usercontrols a virtual robot in a simulator provided at thelocal site, and the virtual robot uses a virtual environ-ment for obstacle avoidance. The command signalsgiven by the user are sent to the IPR at the remotesite via the internet. The IPR moves like the virtualrobot, and avoids obstacles using sensor information.The posture of the virtual robot in the simulator canbe updated by feedback of the IPR posture informa-tion through the internet.

The developed IPR has a square body of size45cm × 52cm × 75cm as shown in Figure 1. Theweight is about 70Kg. It is a 4-wheeled drive with

two fixed wheels and two auxiliary off-centered ori-entable wheels. It consists of a personal computer(Pentium II 333Mhz), a wireless LAN (Samsung Mag-icWave, 2Mbps), a head with two vision color cameras,a 12.1 inch TFT monitor, a speaker, a microphone,sonar sensors (10 pairs), a 12V 100Ah battery (5hr80Ah), and two AC servo motors (LG Industrial Sys-tems, 200W). Two cameras of the head part can rotatearound a vertical and a horizontal axis under the com-mand of the three DC motors. The IPR is connectedto the internet through the wireless LAN, and it worksas a server. The user can connect to the IPR using aWeb browser or a TCP/IP application anywhere andalso the user can give motion commands to the IPR.

CCD cameras

TFT LCD

Sonar sensors

Actuators

Power system

PC

Microphone

CCD cameras

TFT LCD

Sonar sensors

Actuators

Power system

PC

Microphone

Figure 1: The IPR hardware

2.2 Modeling of the IPR

The modeling of IPR is needed for the implementa-tion of the simulator. Two fixed and two auxiliary off-centered orientable wheeled mobile robots with non-slipping and pure rolling are considered. The velocityvector u = [v ω]T consists of the translational ve-locity of the center of driving wheel axis and the ro-tational velocity with respect to the center of drivingwheel axis. The velocity vector u and a posture vec-tor Pc = [xc yc θc]T are associated with the robotkinematics as follows:

Pc =

xc

yc

θc

=

cos θc −h sin θc

sin θc h cos θc

0 1

[vω

]= J(θc) u (1)

u =[vω

]=

[12

12

1L − 1

L

] [vR

vL

](2)

where vR is the right wheel velocity, vL is the left wheelvelocity, h is the displacement from the center of robotto the wheel axis, and L is the distance between thetwo wheels.

2.3 Internet time delay

A large internet time delay may make some con-trol inputs distorted. Moreover, since it is affectedby the number of the internet nodes and loads, thisdelay is variable and unpredictable. Figure 2 showsthe influence of the internet time delay on the con-trol information. The received data at the remote sitewas distorted severely, and the information of the sinefunction was almost lost, when the test function usedwas y(t) = 5 sin(0.2πt) + 5.

0 10 20 30 40 50 600

1

2

3

4

5

6

7

8

9

10

time (sec)

Figure 2: Influence of the internet time delay

3 Internet Control Architecture

A user can control the IPR at the remote sitethrough internet using a simulator provided at the lo-cal site. The user regards the status of the virtual IPRat the local site as that of the real IPR at the remotesite. Since the user cannot recognize the environmentof the remote site, it is expected that the real IPRmoves as the virtual IPR does. However, because oftime delay we have to compensate for the path errorand the time difference between the real IPR and thevirtual IPR, which increase as time goes on.

In this section, a novel internet control architec-ture is designed step by step to minimize the effectof internet time delay. The proposed architecture iscompleted in three design steps, and its effectivenessis verified through simulations and experiments.

The internet control architecture-I consists of a userinterface, simulator, virtual environment and postureestimator, which can be devised from the basic con-cept as shown in Figure 3. In the figure, ur(i) is theith control command [vr(i) ωr(i)]

T from a user, udr(i)

the ith control command passed through the internet,Pc(i) the ith robot posture, Pd

c(i) the ith robot pos-ture passed through the internet, Pc(i) the ith esti-mated posture, and Ps

c(i) the ith posture of the virtualrobot. In order to correct the posture error betweenthe virtual robot and the real robot, the real robot

generates feedback signals such as posture informationof the real robot, to the simulator. The architecture-Ican be considered as a basic structure.

)(ˆ iPc

)(iur

)(iudr)(iPc

UserInterface

PostureEstimator

Personal RobotSimulator

VirtualEnvironment

Internet

Personal Robot

Environment

Local Site Remote Site

)(iur

)(iPdc

)(iPsc

)(ˆ iPc

)(iur

)(iudr)(iPc

UserInterface

PostureEstimator


VirtualEnvironment

Internet

Personal Robot

Environment


)(iur

)(iPdc

)(iPsc

Figure 3: Internet control architecture-I (Step 1)

User Interface which can be implemented by Java,C++, etc., enables a user to control a remote IPR.Posture estimator estimates the current posture of thevirtual IPR based on the feedback information of thereal IPR. Personal robot simulator is the same as thevirtual mobile robot at the local site. Virtual envi-ronment has the information of the real environmentso that it enables the virtual robot to avoid obstacles.Personal robot is the same as the real mobile robot atthe remote site. Environment is a circumstance wherethe real IPR is working.

The internet control architecture-I has a weak pointthat the information loss of control commands in-creases when internet time delay occurs. The postureestimator can recover the information loss eventually,but the time required for the recovery becomes toolong. The architecture which can get rid of the causeof the information loss is needed. Figure 4 shows theinternet control architecture-II, where a command fil-ter is introduced. The command filter can recover theinformation loss of control commands caused by theinternet time delay. It means that the filter reducesthe path error between the real robot and the virtualrobot. The function of the command filter is shownin Figure 5. Command signals received at the sametime after the internet time delay Td are regeneratedwith the sampling time T in the command filter. Thecommand filter consists of two modules such as a com-mand queue and a command generator. The commandfilter and the two modules can be defined by DEVS(Discrete Event Systems Specifications) formalism [9].The command filter receives a control command, andstores it in the command queue. The command gener-

)(ˆ iPc

)(iur

UserInterface

PostureEstimator


VirtualEnvironment

Internet

Personal Robot

Environment


)(iur

)(kPdc

)(iPsc )(iud

r

)(kur)(kPc

CommandFilter

)(ˆ iPc

)(iur

UserInterface

PostureEstimator


VirtualEnvironment

Internet

Personal Robot

Environment


)(iur

)(kPdc

)(iPsc )(iud

r

)(kur)(kPc

CommandFilter

Figure 4: Internet control architecture-II (Step 2)

ator pulls out the command from the command queueand outputs it at each sampling time T .

ru

tT

t

dT

tdT T

Internet

CommandFilter

ru

ru

Figure 5: Function of command filter

The internet control architecture-II can recover theinformation loss of control commands, though internettime delay exists, but it still has the serious problemthat the time difference between the real robot andthe virtual robot increases, as internet time delay Td isaccumulated in the command filter. In order to solvethis problem, the internet control architecture-III isfinally designed.

The internet control architecture-III guaranteesthat the path error and the time difference between thereal IPR at the remote site and the virtual IPR at thelocal site can be reduced. The proposed architectureincludes a path generator and a path-following con-troller. The path generator restores the moving path ofthe virtual robot. The path-following controller guar-antees that the real robot follows the generated path.The time difference between the real robot and the vir-tual robot can be reduced by the path generator andthe path-following controller. As the control input of

the real robot is separated from the control commandpassed through the internet by the two modules, thecommand generator in the command filter can be mod-ified by replacing sampling time T with the processingtime Tp which is shorter than T . The processing timeis the computing interval for generating a path seg-ment for one control command.

)(ˆ iPc

)(iur

UserInterface

PostureEstimator


VirtualEnvironment

Internet

Personal Robot

Environment


)(iur

)( jPdc

)(iPsc )(iud

r

)(kur

CommandFilter

Path-FollowingController

PathGenerator

)(sPS

ucP)( jPc

)(ˆ iPc

)(iur

UserInterface

PostureEstimator


VirtualEnvironment

Internet

Personal Robot

Environment


)(iur

)( jPdc

)(iPsc )(iud

r

)(kur

CommandFilter

Path-FollowingController

PathGenerator

)(sPS

ucP)( jPc

Figure 6: Internet control architecture-III (Step 3)

Figure 6 shows the internet control architecture-III,where PS(s) is the moving path of the virtual robot, uis the control input of the path-following controller, Pc

is the current posture of the real robot, and Pc(j) isthe jth robot posture to be fed back to the simulator.

In this paper, the path-following controller is imple-mented with the uni-vector field navigation method[11]. The uni-vector field makes the mobile robot con-verge to a desired path.

4 Simulations and Experiments

4.1 Simulation results

Simulations were performed in the real internet en-vironment. The physical distance between the localsite and the remote site was about 300Km in Korea.Figure 7 shows the path error and the time differencebetween the two robots.

It should be noted that the architecture-I had cu-mulative errors. Because of this, it might be in-convenient for the user to control the robot in thesimulation environment by the architecture-I. By thearchitecture-II the path error could be reduced, butthe time difference increased continuously. However,by the architecture-III, the time difference as well asthe path error were very small, although the internettime delay was quite variable.

0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)

Local siteRemote site


0 5 10 15 20 25 30 35 400

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber

0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)


0 5 10 15 20 25 30 35 40 450

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber



0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)


0 5 10 15 20 25 30 35 400

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber

��

��

��

0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)


0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)




0 5 10 15 20 25 30 35 400

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber



0 5 10 15 20 25 30 35 400

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber

0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)


0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)



0 5 10 15 20 25 30 35 40 450

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber


0 5 10 15 20 25 30 35 40 450

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber




0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)



0 50 100 150 200 250 300

20

40

60

80

100

120

140

X (cm)

Y (

cm)


0 5 10 15 20 25 30 35 400

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber



0 5 10 15 20 25 30 35 400

10

20

30

40

50

60

70

80

90

time (sec)

even

t num

ber

��

��

��

Figure 7: Simulation results. The path error andthe time difference. (a),(b) architecture-I. (c),(d)architecture-II. (e),(f) architecture-III.

4.2 Experimental results

Experiment was performed with the developed IPRsystem equipped with an overhead CCD camera forglobal positioning.

In this experiment, the proposed internet controlarchitecture was implemented as a TCP/IP applica-tion version, and the physical distance between the lo-cal site and the remote site was about 300Km, whichwas the same condition as that of computer simula-tions. The Posture Estimator was removed from thearchitecture so as to find out the exact characteris-tics of each control structure. Figure 8 shows thepath error and the time difference between the tworobots. In the experimental results of the internetcontrol architecture-I, the path error was caused bythe information loss of control commands. The infor-mation loss of control commands made the real robotpath different from the virtual robot path. In the re-sults by the internet control architecture-II, the timedifference between the actions of the two robots in-

creased continuously, since internet time delay accu-mulated as time goes on. In the results by the internetcontrol architecture-III, the path error and the timedifference were quite small. The experimental resultsdemonstrated the effectiveness and the applicability ofthe proposed internet control architecture-III as thesimulation results did.

5 Conclusions

An internet-based personal robot with novelinternet control architecture was developed. Theproposed architecture was insensitive to internet timedelay and guaranteed that the path error and thetime difference between a real IPR and a virtualIPR could be reduced. Simulations and experimentalresults in a real internet environment demonstratedthe effectiveness and the applicability of the proposedinternet control architecture.

Acknowledgements

The authors would like to thank the support givenby MIC(Ministry of Information and Communica-tion), Korea, to the development of the internet-basedpersonal robot.

References

[1] K. Taylor and B. Dalton, “Issues in InternetTelerobotics,” in Int. Conf. on Field and ServiceRobotics, Dec. 1997.

[2] C. Sutter and J. Wiegley, “Desktop Teleoperationvia the World Wide Web,” in Proc. IEEE Int.Conf. Robot. Automat., pp. 654-659, May 1995.

[3] R. Volpe, J. Balaram, T. Ohm and R. Ivlev,“The Rocky 7 Mars Rover Prototype,” in Proc.IEEE/RSJ Int. Conf. on Intelligent Robots andSystems, pp. 1558-1564, Nov. 1996.

[4] T. M. Chen and R. C. Luo, “Remote Supervi-sory Control of An Autonomous Mobile RobotVia World Wide Web,” in Proc. IEEE Int. Sym-posium on Industrial Electronics, vol. 1, pp. ss60-ss64, July 1997.

[5] K. Brady and T. J. Tarn, “Internet-Based Re-mote Teleoperation,” in Proc. IEEE Int. Conf.Robot. Automat., pp. 65-70, May 1998.

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

��

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 50 100 150 200 250 3000

50

100

150

200

250

X [cm]

Y [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm]

��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

300

Time[msec]

Pos

ition

X [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm] ��

��

0 0.5 1 1.5 2 2.5 3 3.5 4

x 104

0

50

100

150

200

250

Time[msec]

Pos

ition

Y [

cm]

��

��

��

Figure 8: Experimental results. The path error and the time difference. (a),(b),(c) architecture-I. (d),(e),(f)architecture-II. (g),(h),(i) architecture-III.

[6] E. Paulos and J. Canny, “Designing PersonalTele-embodyment,” in Proc. IEEE Int. Conf.Robot. Automat., pp. 3173-3178, May 1998.

[7] E.P.L. Aude, G.H.M.B. Caneiro, H. Serdeira,J.T.C. Silveira, M.F. Martins and E.P. Lopes,“CONTROLAB MUFA: A Multi-Level FusionArchitecture for Intelligent Navigation of a Teler-obot,” in Proc. IEEE Int. Conf. Robot. Automat.,pp. 465-472, May 1999.

[8] R. Oboe and P. Fiorini, “A Design and ControlEnvironment for Internet-Based Telerobotics,”Int. Journal of Robotics Research, vol. 17, no. 4,pp. 433-449, Apr. 1998.

[9] K.-H. Han, S. Kim, Y.-J. Kim and J.-H. Kim,“Internet Control Architecture for Internet-BasedPersonal Robot,” Autonomous Robots Journal,Kluwer Academic Publishers, vol. 10, no. 2, pp.135-147, Mar. 2001.

[10] K.-H. Han, S. Kim, Y.-J. Kim and J.-H. Kim,“Internet-Based Personal Robot System usingMap-Based Localization,” in Proc. the 32nd Int.Sym. on Robotics, Apr. 2001.

[11] J.-H. Kim, K.-C. Kim, D.-H. Kim, Y.-J. Kimand P. Vadakkepat, “Path Planning and Role Se-lection Mechanism for Soccer Robots,” in Proc.IEEE Int. Conf. Robot. Automat., pp. 3216-3221,May 1998.

Implementation of Internet-Based Personal Robot with Internet Control Architecturekhhan.com/pdf/ICRA2001.pdf · 2003. 7. 1. · embodiment [6] and an intelligent telerobot [7] were

Documents