Realization of Simultaneity in Bilateral Control System ...

IEEJ Journal of Industry ApplicationsVol.4 No.3 pp.253–261 DOI: 10.1541/ieejjia.4.253

Paper

Realization of Simultaneity in Bilateral Control Systemunder Communication Delay

Satoshi Nishimura∗ Student Member, Seiichiro Katsura∗ Senior Member

(Manuscript received March 12, 2014, revised Sep. 14, 2014)

The realization of a new communication medium that realizes the transmission of haptic information between dis-tant places is required. A bilateral control system is an effective technique that can share tactile sensation between twosystems. However, the performance of the bilateral control tends to destabilize, and the haptic information deterioratesunder a communication delay because haptic information has the bilateral information flow property. This is becausethe control goals cannot be achieved in real time owing to the delay time. Therefore, a novel method is proposedto realize simultaneity in a bilateral control system under a communication delay. The proposed control system isdesigned to realize the control goal equations for bilateral control regardless of the delay time and is designed symmet-rically. Buffering the force information of the system resolves the interference between the modal space caused by thecommunication delay. The entire control system is stabilized by using a phase-lag compensator that has the equivalentmeaning of acceleration response feedback with a high pass filter. The validity of the proposed method is confirmedby experiments.

Keywords: Acceleration Control, Bilateral Control System, Communication Delay, Simultaneity, Modal Space, Phase-Lag Com-pensator

1. Introduction

Communication method between remote places has been awide interest for human, and there has been a lot of tech-niques practicalized that realize a smooth communicationwith distant points. Some of the examples are telephone, tele-vision, and the Internet, and they cleared the way to have acommunication without any limitations of time and space.The medium mentioned above removed the communicationrestriction of space by sending the information on-line andit realized people to communicate with other people whois in the remote places. Moreover, the medium enabled toovercome the communication restraint of time by recordingand preserving the information, realizing the communicationwithout considering the time axis. However, the informa-tion of the present medium that enables communication be-tween remote place are only audio and visual information,and there is no medium that deals with haptic information.Realization of the haptic information transmission permits todramatically enhance the transparency of the communicationbetween the remote place; therefore, the practicalization ofmedia that transmits haptic information is expected.

There is an academic field that handles a haptic informa-tion of the real world and it is called real-world haptics.In the academic field, there is a control system called a bi-lateral control that enables force sensation transmission be-tween two systems. A bilateral control realizes force sen-sation transmission by fulfilling two control goals, that aresynchronization of the position and the realization of the law

∗ Department of System Design Engineering, Keio University3-14-1, Hiyoshi, Kohoku-ku, Yokohama 223-8522, Japan

of action and reaction between the two systems. There hasbeen many bilateral control systems proposed (1)–(3), and thecontrol system that is based on acceleration control (4) is themost effective way of sending vivid haptic sensation to re-mote places. Therefore, it can be said that in order to estab-lish a brand new media that transmits haptic information, abilateral control system is one effective tool to realize.

As for the media establishment perspective, it is needed toovercome two problems; that are the limitations of time andspace. Haptic information is now able to be transmitted re-gardless of time restriction by using the motion-copying sys-tem (5). Motion-copying system is a control system that al-lows to preserve the tactile information and reproduce it inany time it is needed. However, as for the limitation of thespace, there is yet to be a problem. It is difficult to over-come because haptic information has bilateral informationflow unlike visual and audio information. Haptic informa-tion is transmitted between the two objects following the lawof action and reaction, and the law has the bilateral informa-tion flow property. When assuming to communicate hapticinformation between remote places, there inevitably occurstime delay in the communication links. Since a haptic infor-mation has bilateral property, a time delay is contained insidethe control system loop, destabilizing the whole control sys-tem and deteriorating the haptic information. A time delayis said to be one of the biggest problem because of its char-acteristic phase response (6), and there are several studies thatanalyze the performance of a bilateral control system under atime delay (7).

There has been several methods proposed before to com-pensate the time delay problem. A bilateral control utiliz-ing μ synthesis was considered in (8). A bilateral control

c© 2015 The Institute of Electrical Engineers of Japan. 253

Realization of Simultaneity in Bilateral Telecommunication System（Satoshi Nishimura et al.）

implementing passivity method is a well known techniquethat ultimately stabilize the whole control system (9) (10). Real-time network protocol stack (RTNP) was proposed (11) to re-solve the problems inherent in a multitasking operating sys-tem. Smith predictor that is a famous method for solvingthe time delay problem was utilized (12). Communication dis-turbance observer (CDOB) introduced the new concept ofnetwork disturbance, which view a time delay as a distur-bance (13), and improved the operational force in a bilateralcontrol (14). The concept of frequency domain damping (FDD)was proposed (15) to enhance the stability performance. Forcecontroller gain was designed to be variable (16) to realize goodposition tracking performance. There is another method thatutilizes visual information to deal with a communication de-lay (17). Although there are many conventional methods, thereis no method that tried to construct the control system thatseems there is no time delay between the systems. Conven-tional methods presuppose the presence of time delay; there-fore, the performance of the bilateral control system whenthere is no time delay cannot be regained. One of the biggestproblem for a bilateral control system under a communica-tion delay is that the control goals cannot be realized in thesame time axis because of the time delay. The time axis ofthe input for each system is different from each other, causingserious error inside the control system.

Therefore, in this paper, a novel control method that real-izes position synchronization between the two systems with-out the effect of the communication delay is proposed. Thetime axis of the input for each system is corrected by buffer-ing the force information of the own system. The motivationof the paper is to regain the performance that was lost becauseof the time delay. The proposed method utilizes bufferedforce information to regain the performance. A phase-lagcompensator is used for the stabilization. The control systemis analyzed by modal space analysis and by hybrid parame-ter. The control system is also designed symmetrically. Themerit of the symmetrical property is that it is able to operatefrom both systems because there is no distinction between thesystems.

The proposed method requires model of the delay timein order to realize simultaneity property. The situation thatis assumed in this paper is wired communication, not wire-less communication. The variation of the delay time is muchsmaller in wired communication compared to wireless com-munication; therefore, in this paper, it is assumed that timedelay variation is too small to be neglected. However, thereis also a situation that time delay variation is large and can-not be neglected. The proposed method also can be applied totime-varying delay situation by utilizing time stamp or othermethods that are proposed by other authors (19). By measuringand updating a delay time model, the method can be appliedto time-varying delay to realize simultaneity property.

The validation of the proposed method is confirmedthrough experiments. Since the orientation of the control cri-teria is different from any other conventional methods, a bi-lateral control system with a phase-lag compensator only isutilized as the conventional method in the experiments. Itis known that the stability can be enhanced by the use of aphase-lag compensator (16).

The paper is organized as follows. In section 2, a bilateral

control system is explained. The proposed control systemis written in section 3, and both performance enhancementmethod and stabilization method are proposed. Some exper-iments are conducted to verify the validity of the proposedmethod in section 4. This paper is summarized in the lastsection.

2. Bilateral Control System

2.1 Bilateral Control System without Time DelayAs mentioned in the previous section, a bilateral control

is a control system that realizes force sensation transmissionbetween the two systems. The two systems are called mastersystem and slave system respectively. Master system is a sys-tem that human operates, and slave system is a system thatcontacts with the environment. The control goals for eachsystem are shown as follows:

F̂extown + F̂ext

otr = 0 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (1)

Xresown − Xres

otr = 0 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (2)

where F̂ext，Xres，©own，and©otr stand for the estimated ex-ternal force, position response, the value regarding the ownsystem, and the value of the other system. When two systemsare placed near to each other, the information from the othersystem will instantly arrive. The estimated external force isobtained by using the reaction force observer, and the posi-tion response is obtained by the encoder in a bilateral con-trol system. Each system is implemented with disturbanceobserver; therefore, every disturbance inside the cut-off fre-quency region of disturbance observer is rejected to 0, real-izing ideal acceleration control system. In this paper, it isassumed that ideal acceleration control system is achievedby the disturbance observer. Two systems synchronizes as iftwo systems are connected with each other, and force infor-mation are transmitted to the other system in a real time. Forsimplicity, F̂ext and Xres is expressed as F and X respectivelyhereafter.

Reference inputs of each control goal are generated sepa-rately by using the concept of modal decoupling (18). The ref-erence input regarding the force information is generated inthe common modal space, and the reference input regardingthe position information is generated in the differential modalspace. Two modal spaces does not interfere when there is nocommunication delay inside the control system. The acceler-ation references for each mode space are shown as follows:

s2Xre fdi f = −Cp(Xm − Xs) · · · · · · · · · · · · · · · · · · · · · · · · · (3)

s2Xre fcom = −C f (Fm + Fs) · · · · · · · · · · · · · · · · · · · · · · · · (4)

where Cp and C f stand for position controller and force con-troller respectively. Position controller is a PD controller andthe force controller is a P controller. The final accelerationreferences for each system are obtained as shown in the fol-lowing:

s2Xre fm = −1

2Cp(Xm − Xs) − 1

2C f (Fm + Fs) · · · · · · (5)

s2Xre fs = −1

2Cp(Xs − Xm) − 1

2C f (Fs + Fm). · · · · · · (6)

The hybrid matrix mathematically describes the control

254 IEEJ Journal IA, Vol.4, No.3, 2015


goals of the bilateral control. The matrix shows the rela-tion between the force and position information of each mas-ter and slave system. The definition of the hybrid matrix isshown in the following:[

Fm

Xm

]= H

[Xs

Fs

]

=

[H11 H12

H21 H22

] [Xs

Fs

]. · · · · · · · · · · · · · · · · (7)

The hybrid matrix for usual bilateral control is obtained asfollows:

H =

⎡⎢⎢⎢⎢⎣ − 2s2

C f−1

1 0

⎤⎥⎥⎥⎥⎦ . · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (8)

H12 shows the simultaneity property, and H21 is the indexfor realization of the law of action and reaction. The controlgoals are achieved when H12 = −1 and H21 = 1，as shown in(1) and (2).2.2 Bilateral Control System under Time Delay In

order to use a bilateral control system as the communicationtool for the haptic information, the communication betweenthe remote places needs to be assumed. Since there is a timedelay between the systems, the information from the othersystem is delayed. The further two places are, the biggerthe delay time becomes; deteriorating the haptic informationand destabilizing the control system. The control goals whenthere is time delay are modified as follows:

Fown + e−T sFotr = 0 · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (9)

Xown − e−T sXotr = 0. · · · · · · · · · · · · · · · · · · · · · · · · · · · (10)

It is obvious that the control goals are different from (1) and(2).

The acceleration references for each system are shown asfollows:

s2Xre fm =−1

2Cp(Xm−e−T sXs)− 1

2C f (Fm+e−T sFs) · · · · · (11)

s2Xre fs =−1

2Cp(Xs−e−T sXm)− 1

2C f (Fs+e−T sFm) · · · · · (12)

where T stands for the communication delay time. As shownin (11) and (12), the information from the other system is de-layed due to the communication delay. The block diagram ofthe bilateral control under time delay is shown in Fig. 1. Thechange regarding the control goals indicates that the bilateralcontrol system under time delay cannon have the same per-formance as the bilateral control system without time delay.The effect is analyzed in the modal space.

The communication delay causes interferences betweenthe differential and the common modal space, and the inter-ference deteriorates the haptic information and destabilizesthe whole control system. The acceleration references foreach modal space are shown as follows:

s2Xre fdi f = −

12

(1 + e−T s)Cp(Xm − Xs)

− 12

(1 − e−T s)C f (Fm − Fs) · · · · · · · · · · · · (13)

Fig. 1. Block diagram of the bilateral control under timedelay

Fig. 2. Block diagram of the conventional differentialmodal space

Fig. 3. Block diagram of the conventional commonmodal space

s2Xre f = − 12

(1 + e−T s)C f (Fm + Fs)

− 12

(1 − e−T s)Cp(Xm + Xs). · · · · · · · · · · · (14)

The block diagrams of each mode space are shown inFigs. 2 and 3. As for the differential modal space, there is adisturbance from the differential mode of the force informa-tion; therefore, simultaneity in bilateral control under timedelay cannot be realized. As for the common modal space,there is a disturbance caused from the differential mode ofthe position information; therefore, the law of action and re-action is not realized in the conventional bilateral control un-der communication delay. The block diagrams visually showthat the control goals of bilateral control cannot be achievedunder time delay.

The hybrid matrix shows that both control goals are notachieved mathematically as shown in the following:



Hcon =

[Hcon

11 Hcon12

Hcon21 Hcon

22

]· · · · · · · · · · · · · · · · · · · · · · · (15)

⎧⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎪⎩

Hcon11 =

(2s2 +Cp)2 − (e−T sCp)2

2e−T sC f (s2 + Cp)

Hcon12 =

2s2 + (1 + e−2T s)Cp

2e−T s(s2 + Cp)

Hcon21 =

2s2 + (1 + e−2T s)Cp

2e−T s(s2 + Cp)

Hcon22 =

(1 − e−2T s)C f

2e−T s(s2 + Cp)

· · · · · · · · (16)

where ©con stands for the value of the conventional method.Both elements H12 and H21 are not the desired values, that iscaused by the interferences of the modal space.

3. The Proposed Method

A control system that realizes simultaneity in a bilateralcontrol under time delay is proposed in this section.3.1 Removing the Disturbances in the Modal SpaceFigure 2 describes that the disturbance from the differential

mode of the force information is the main reason for the inter-ruption of the simultaneity realization. Moreover, Fig. 3 in-dicates that position difference causes the interference in theforce common modal space. Therefore, removing the distur-bance in the position differential modal space has influenceto not only the position differential modal space but also tothe force common modal space.

The disturbance is derived because the time axis of theforce input information is different between the two systemsdue to the communication delay. Even though a bilateral con-trol has two acceleration inputs which are force-derived andposition-derived acceleration references, the pure input forthe control system that is from outside the system is the forceinformation. Force information is pure control input to movethe actuators; therefore, when the time axis of this force inputis different from each other, the position response obviouslybecomes different from each other, making it impossible torealize simultaneity in bilateral control.

Therefore, in this paper, the own force information is de-layed in order to correct the time axis of both systems. Bycorrecting the time axis, the force differential disturbance re-garding the position differential modal space becomes 0 asshown in the following:

(e−T s − e−T s) ·C f (Fm − Fs) = 0 ·C f (Fm − Fs)

= 0. · · · · · · · · · · · · · · · · (17)

The physical effect regarding the use of buffered force in-formation is the shift of the control goals. The control goalsof the proposed method are shown as follows:

e−T sFown + e−T sFotr = 0 · · · · · · · · · · · · · · · · · · · · · · · (18)

Xown − e−T sXotr = 0. · · · · · · · · · · · · · · · · · · · · · · · · · · · (19)

The control goal regarding force information is modified sothat the time axis in the force control space becomes thesame.

The proposed acceleration references for each system areshown as follows:

Fig. 4. Block diagram of the proposed bilateral controlsystem under time delay

Fig. 5. Block diagram of the proposed differentialmodal space

Fig. 6. Block diagram of the proposed common modalspace

s2Xre fm = − 1

2Cp(Xm − Ppe−T sXs)

− 12

Pf C f (e−T sFm + e−T sFs) · · · · · · · · · · · (20)

s2Xre fs = − 1

2Cp(Xs − Ppe−T sXm)

− 12

Pf C f (e−T sFs + e−T sFm) · · · · · · · · · · · (21)

where Pp and Pf stand for phase-lag compensator for the po-sition control and force control respectively. The explanationof the phase-lag compensator is written later. The block dia-gram of the proposed control system is shown in Fig. 4.

The acceleration references for each modal spaces areshown as follows:

s2Xre fdi f = −

12

(1 + Ppe−T s)Cp(Xm − Xs) · · · · · · · · · · (22)

s2Xre f = −Pf e−T sC f (Fm + Fs)

−12

(1 − Ppe−T s)Cp(Xm + Xs). · · · · · · · · · (23)

Figures 5 and 6 indicate the block diagrams of position



differential modal space and force common modal space re-spectively. Comparing with the conventional block diagramsof the modal space, there are no disturbances in the proposedmethod.

The control goal regarding position information is notchanged in the proposed method; however, the validity of theproposed method is shown by analyzing the system in hybridmatrix. The hybrid matrix shows the effect of the proposedmethod from the mathematical point of view

Hpro =

⎡⎢⎢⎢⎢⎢⎣ −2s2+(1−Ppe−T s)Cp

Pf e−T sC f−1

1 0

⎤⎥⎥⎥⎥⎥⎦ . · · · · · · · · · · · · · · (24)

As shown in the previous section, not only simultaneity butalso the law of action and reaction cannot be realized whencommunication delay exists. However, H21 of the proposedmethod shows that simultaneity is achieved regardless of thetime delay. Moreover, H12 of the proposed method indicatesthe law of action and reaction is also realized without anyinfluence of the communication delay.3.2 The Effect of the Phase-Lag Compensator Fig-

ures 5 and 6 indicate that there is time delay element insidethe control loop. Each control loop needs to be stabilized torealize haptic transmission through bilateral control. To sta-bilize the system, a phase-lag compensator is utilized in thispaper. At first, the physical meaning of the phase-lag com-pensator, which is the designing of equivalent mass of themotor, is explained.

The equation of the phase compensator is written in (25)

P =s + g

1α

s + g. · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (25)

when α is smaller than 1, it becomes the phase-lag compen-sator. The transfer function of the block diagram shown inFig. 7 is written in (26)

G =s + gM

(1 + 1KM M )s + gM

. · · · · · · · · · · · · · · · · · · · · · · · · · (26)

Comparing (25) and (26), 1α

is same as 1+ 1KM M . As long as

KM is positive value, the coefficient of s is bigger than 1. Fig-ure 7 shows the block diagram of the acceleration responsefeedback with a high pass filter, and the equivalent transfor-mation becomes as same as the phase-lag compensator thatis shown in Fig. 8. There are two parameters that can be de-signed in the phase-lag compensator which are KM and gM.

Fig. 7. Acceleration response feedback with high passfilter

Fig. 8. Equivalent form of Fig. 7

The equivalent mass of the motor is determined by the gainKM and the frequency from which region the equivalent mo-tor mass is changed is decided by the cut-off frequency gM .The phase-lag compensator can change the equivalent massheavier than actual mass over the frequency region of gM .3.3 Phase-lag Compensator Design The phase-lag

compensator parameters of Pp and Pf are designed in thesubsection. Stability of the whole control loop is assured bystabilizing both position differential modal space and forcecommon modal space. Open loop transfer functions of eachcontrol loop are shown in the following:

Gopendi f =

(1 + Ppe−T s

)Cp

2s2· · · · · · · · · · · · · · · · · · · · · · · (27)

Gopencom =

Pf e−T sC f (Zh + Ze)

2s2 +(1 − Ppe−T s

)Cp

. · · · · · · · · · · · · · · · · (28)

Pp is first designed to stabilize the position differential modalspace, and the performance of simultaneity property is deter-mined. Parameters of Pp is designed not to cross (−1, 0) inNyquist plot. The effect of Pp is shown in Fig. 9. The fig-ure indicates that the control loop is stabilized by utilizingPp. The parameters of Pp is determined also by consideringthe effect of modeling error of delay time since the effect ofmodeling error appears in position differential mode as a dis-turbance. Cut-off frequency gp determines the frequency areawhere simultaneity property is realized.

At next, parameters of Pf are designed to stabilize the forcecommon modal space, and equivalent mass of the whole sys-tem is determined. The design criterion is the same as that ofPp. Cut-off frequency g f determines from which frequency

Fig. 9. Nyquist plot regarding position differentialmodal space

Fig. 10. Nyquist plot regarding force common modalspace



Table 1. Control parameters

Parameter Description Value

T Delay time 0.1 sKp Position gain 900.0Kv Velocity gain 60.0C f Force gain 1.0gdis Cut-off frequency of DOB 1000.0 rad/sgreac Cut-off frequency of RFOB 1000.0 rad/sαp Phase-lag compensator gain 0.4gp Cut-off frequency of Pp 17.5 rad/sα f Phase-lag compensator gain 0.01g f Cut-off frequency of Pf 25.0 rad/sKh Stiffness of human 1000Dh Viscosity of human 100Ke Stiffness of environment 2500De Viscosity of environment 100

the mass of the system becomes large. As same as positiondifferential modal space, Fig. 10 shows that the use of thephase-lag compensator can stabilize the modal space. Theparameters for Nyquist plotting are shown in Table 1.3.4 Performance Analysis The transmitted impe-

dance in the proposed system is analyzed in the following.The transfer function from the master force and the masterposition is shown as follows:

Fm=

(2s2+(1−Ppe−T s)Cp

Pf e−T sC f+Ze

)Xm. · · · · · · · · · · · · · (29)

The transfer function shows the performance of the systemwhen the slave system is contacting the environment. Trans-mitted impedance is analyzed through the index operational-ity Po and reproducibility Pr

(4). In ideal situation, Po = 0 andPr = 1 is desired. In the proposed method, operationality andreproducibility are defined as follows:

Po =2s2 + (1 − Ppe−T s)Cp

Pf e−T sC f· · · · · · · · · · · · · · · · · · · · (30)

Pr = 1. · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · · (31)

Transfer function of the proposed method can realize idealreproducibility even though there is time delay between thesystems.

The bode plots for each index are shown in Figs. 11 and 12.The parameters that are used for bode plotting are shown inTable 1. As for the comparison, conventional bilateral controlwithout any compensation, a bilateral control with a phase-lag compensator, and 4ch bilateral control with FDD (15) areshown. The results show that though the operationality be-comes large in the proposed method, the reproducibility hasthe best performance in the three methods.

Analytical results show that even though proposed methodhas relatively large operational force, the method can trans-mit the exact impedance of the environment.

4. Experiments

In this section, some experiments are conducted to verifythe validity of the proposed method.

Three motions, which are free motion, contact motion to asoft environment, and contact motion to a hard environment,are compared between the conventional method and the pro-posed method. Conventional method is the system which aphase-lag compensator only is implemented. Since there was

Fig. 11. Bode plot regarding operationality

Fig. 12. Bode plot regarding reproducibility

Fig. 13. Experimental setup

no conventional method that has the same design criteria asthe proposed method, the validity of the proposed bufferingmethod was evaluated by comparing the proposed methodwith the method without buffering method. Experimental pa-rameters are same as the one shown in Table 1. Cut-off fre-quency of DOB gdis and cut-off frequency of RFOB greac wasset as high as possible in order to eliminate the effect of forcedifferential mode and to transmit large bandwidth force sen-sation.

At first, free motion is compared. Experimental results areshown in Figs. 14 and 15. Blue lines show the error. In theconventional method, there are errors in the position responseand some severe vibrations, and this is because time axis ofthe force input is different between the master and slave sys-tems. On the contrary, the proposed method has no error oroscillations in the position response.

The enlarged view of the free motions are shown in Figs. 16and 17. The difference between the conventional and the



Fig. 14. Experimental results of the free motion andcontact motion with soft environment in the conventionalmethod

Fig. 15. Experimental results of the free motion andcontact motion with soft environment in the proposedmethod

proposed method can be seen in the figures. In the con-ventional method, since the human force information arrivesfaster to the master system than to the slave system, the tim-ing that the master system actuates becomes faster than theslave system, preventing the bilateral control system to real-ize simultaneity property. The difference also causes positionresponse overshoot that can be explained in the same rea-son. On the other hand, the proposed system has simultane-ity property since the timing that force information to arriveis corrected. It can be observed that there is no position re-sponse overshoot in the proposed method.

At next, contact motion to soft environment is compared.Experimental results are the gray region of Figs. 14 and 15.Gray zones show that slave system contacted to the envi-ronment. The conventional method shows some position er-ror when contacting the environment, however the proposedmethod shows no position error in the contact motion. Theforce response error is is due to the operational force.

At last, the contact motion to hard environment is com-pared. Experimental results are shown in Figs. 18 and 19. Pf

is conservatively designed in the experiment. The conven-tional method has large position error when contacting the

Fig. 16. Enlarged view of Fig. 14

Fig. 17. Enlarged view of Fig. 15

hard environment and force response of the slave system isoscillating because of the position error caused by the inter-ference of the modal space. On the other hand, position errorin the proposed method is much smaller than in the conven-tional method. The oscillation of the force response is atten-uated in the proposed method, however, there are still someoscillation because of the position error.

The errors in each experiment are compared by usingRMSE in the following. The errors during contact motion arecalculated within the grey region. Position error was evalu-ated in Table 2, while force error was listed in Table 3. Theresults show that both errors were reduced in the proposedmethod.

Actual network contains jitter problem, and the effect ofjitter can be reduced by designing the parameters of phase-lag compensator conservatively. The overall results showedthe validity of the proposed method.

5. Conclusions

This paper proposed a novel method to realize simultane-ity in a bilateral control system under a communication delay.The proposed control system could realize the control goalsof a bilateral control regardless of a time delay. By bufferingthe force information in order to correct time axis of the forceinput, the interferences between the modal spaces became 0.



Fig. 18. Experimental results of the contact motion withhard environment in the conventional method

Fig. 19. Experimental results of the contact motion withhard environment in the proposed method

Table 2. Position response error

Conventional method Proposed method

Free motion 1.12 × 10−3 [m] 7.29 × 10−5 [m]Contact motion tosoft environment

9.24 × 10−4 [m] 1.56 × 10−4 [m]

Contact motion tohard environment

9.39 × 10−4 [m] 5.52 × 10−4 [m]

Table 3. Force response error

Conventional method Proposed method

Contact motion tosoft environment

1.71 [N] 1.70 [N]

Contact motion tohard environment

1.01 [N] 0.85 [N]

The proposed control system was designed symmetrically sothat human could operate from both systems.

The proposed method was stabilized by using the phase-lagcompensator. The paper indicated that the phase-lag compen-sator in the force control has the equivalent effect of acceler-ation response feedback with a high pass filter.

The experimental results showed the validity of the pro-posed method. The position error reduced to almost 0 in theproposed method, and the force response was also improvedbecause of the modal decoupling.

AcknowledgmentThis work was partially supported by JSPS KAKENHI.

References

( 1 ) B. Hannaford: “A Design Framework for Teleoperators with KinestheticFeedback”, IEEE Trans. on Robotics and Automation, Vol.9, No.5, pp.426–434 (1989)

( 2 ) B. Hannaford and R. Anderson: “Experimental and Simulation Studies ofHard Contact in Force Reflecting Teleoperation”, Proc. of the 1988 IEEEInternational Conference on Robotics and Automation, ICRA 1988, pp.584–589 (1988)

( 3 ) K. Hashtrudi–Zaad and S.E. Salcudean: “Bilateral Parallel Force/PositionTeleoperation Control”, Journal of Robotic Systems, Vol.19, No.4, pp.155–167 (2002)

( 4 ) W. Iida and K. Ohnishi: “Reproducibility and Operationality in BilateralTeleoperation”, Proc. of the 8th IEEE International Workshop on AdvancedMotion Control, AMC 2004, pp.217–222 (2004)

( 5 ) Y. Yokokura, S. Katsura, and K. Ohishi: “Stability Analysis and Experimen-tal Validation of a Motion-Copying System”, IEEE Trans. on Industrial Elec-tronics, Vol.56, No.10, pp.3906–3913 (2009)

( 6 ) J.P. Richard: “Time-Delay Systems: an Overview of Some Recent Advancesand Open Problems”, Automatica, Vol.39, No.10, pp.1667–1694 (2003)

( 7 ) K. Natori, S. Katsura, T. Tsuji, and K. Ohnishi: “Analysis of Bilateral Sys-tems with Time Delay”, Proc. of the 14th IEEE International Symposium onIndustrial Electronics, ISIE 2005, pp.1511–1516 (2005)

( 8 ) G.M.H. Leung, B.A. Francis, and J. Apkarian: “Bilateral Controller for Tele-operators with Time Delay via μ-Synthesis”, IEEE Trans. on Robotics andAutomation, Vol.11, No.1, pp.105–116 (1995)

( 9 ) R.J. Anderson and M.W. Spong: “Bilateral Control of Teleoperation withTime Delay”, IEEE Trans. on Automatic Control, Vol.34, No.5, pp.494–501(1989)

(10) M. Franken, S. Stramigioli, S. Misra, C. Secchi, and A. Macchelli: “BilateralTelemanipulation With Time Delays: A Two-Layer Approach CombiningPassivity and Transparency”, IEEE Trans. on Robotics, Vol.27, No.4, pp.741–756 (2011)

(11) Y. Uchimura and T. Yakoh: “Bilateral Robot System on the Real-TimeNetwork Structure”, IEEE Trans. on Industrial Electronics, Vol.51, No.5,pp.940–946 (2004)

(12) A.C. Smith and K. Van Hashtrudi-Zaad: “Neural Network-based Teleopera-tion using Smith Predictors”, Proc. of the 2005 IEEE International Confer-ence on Mechatronics and Automation, ICMA 2005, Vol.3 pp.1654–1659(2005)

(13) K. Natori, R. Oboe, and K. Ohnishi: “Robust Time Delayed Control Systemswith Communication Disturbance Observer”, Proc. of the 33rd Annual Con-ference of IEEE Industrial Electronics Society, IECON 2007, pp.316–321(2007)

(14) K. Natori, T. Tsuji, K. Ohnishi, A. Hace, and K. Jezernik: “Time-Delay Com-pensation by Communication Disturbance Observer for Bilateral Teleoper-ation Under Time-Varying Delay”, IEEE Trans. on Industrial Electronics,Vol.57, No.3, pp.1050–1062 (2010)

(15) A. Suzuki and K. Ohnishi: “Frequency-Domain Damping Design for Time-Delayed Bilateral Teleoperation System Based on Modal Space Analysis”,IEEE Trans. on Industrial Electronics, Vol.60, No.1, pp.177–190 (2013)

(16) H. Morimitsu, S. Katsura, and M. Tomizuka: “Design of Force Compensatorwith Variable Gain for Bilateral Control System under Time Delay”, Proc.of the 22nd IEEE International Symposium on Industrial Electronics, ISIE2013, pp.1–6 (2013)

(17) D. Yashiro, K. Natori, and K. Ohnishi: “A Synchronization Method of Visualand Tactile Senses for Bilateral Teleoperation over Network”, IEEJ Trans.IA, Vol.129-D, No.6, pp.585–592 (2009)

(18) S. Katsura, W. Iida, and K. Ohnishi: “Medical Mechatronics—An Appli-cation to Haptic Forceps”, IFAC Annual Reviews in Control, Vol.29, No.2,pp.237–245 (2005)

(19) X. Xu, Z. Xiong, X. Sheng, J. Wu, and X. Zhu: “A New Time Synchroniza-tion Method for Reducing Quantization Error Accumulation Over Real-TimeNetworks: Theory and Experiments”, IEEE Trans. on Industrial Electronics,Vol.9, No.3, pp.1659–1669 (2013)



Satoshi Nishimura (Student Member) received his B.E. degree insystem design engineering and M.E. degree in in-tegrated design engineering from Keio University,Yokohama, Japan, in 2013, 2014, respectively, wherehe is currently working toward the Ph.D. degree. Hisresearch interests include real-world haptics and mo-tion control. He is a Student Member of IEEJ, as wellas IEEE, The Japan Society of Mechanical Engineers(JSME), and Robotics Society of Japan (RSJ).

Seiichiro Katsura (Senior Member) received the B.E. degree in sys-tem design engineering and the M.E. and Ph.D. de-grees in integrated design engineering from Keio Uni-versity, Yokohama, Japan, in 2001, 2002 and 2004,respectively. From 2003 to 2005, he was a ResearchFellow of the Japan Society for the Promotion of Sci-ence. From 2005 to 2008, he worked at NagaokaUniversity of Technology, Nagaoka, Niigata, Japan.Since 2008, he has been with Keio University, Yoko-hama, Japan. His research interests include human

support, super embodiment, real–world haptics, systems energy conversion,and industrial electronics. Prof. Katsura received the Best Paper Award fromthe Institute of Electrical Engineers of Japan (IEEJ) in 2003, the Dr. YasujiroNiwa Outstanding Paper Award in 2004, The European Power Electron-ics and Drives–Power Electronics and Motion Control Conference, EPE–PEMC’08 Best Paper Award in 2008, and the IEEE Industrial ElectronicsSociety Best Conference Paper Award in 2012. He is a Senior Member ofIEEJ, as well as a Member of the IEEE, EPE, The Society of Instrumentand Control Engineers (SICE), The Japan Society of Mechanical Engineers(JSME), The Japan Society for Precision Engineering (JSPE), Robotics So-ciety of Japan (RSJ), The Institute of Electronics, Information and Com-munication Engineers (IEICE), and The Japan Society of Computer AidedSurgery (JSCAS).


Realization of Simultaneity in Bilateral Control System ...

Documents