Extended Suary pp.1224 1230 Perforance Analyi of a Three-Channel Control Architecture for Bilateral Teleoperation with Tie Delay Ryogo Kubo Meber (Keio Univerity, kubo@u.d.keio.ac.jp) Noriko Iiyaa Student Meber (Keio Univerity, noriko@u.d.keio.ac.jp) Kenji Natori Student Meber (Keio Univerity, natori@u.d.keio.ac.jp) Kouhei Ohnihi Senior Meber (Keio Univerity, ohnihi@d.keio.ac.jp) Hirotaka Furukawa Non-eber (NTT DoCoMo, Inc., furukawahiro@nttdocoo.co.jp) Keyword: bilateral control, acceleration control, tie delay, tranparency, three-channel architecture, haptic Thi paper preent a novel three-channel control architecture for bilateral teleoperation with/without tie delay. In concrete ter, thi yte ha two traniion channel of poition and force inforation fro the ater ide to the lave ide and one traniion channel of force inforation fro the lave ide to the ater ide. The ater controller of the propoed three-channel teleoperation yte doe not include a poition controller, i.e. only force control i ipleented in the ater ide, in order to iprove operationality in the ater ide. Fig. 1 how a general four-channel architecture for bilateral teleoperation. In Fig. 1, θ re, θ re,ˆτ ext and ˆτ ext are the poition of the ater robot, the poition of the lave robot, the etiated external torque exerted on the ater robot and the etiated external torque exerted on the lave robot, repectively. The external torque exerted on the robot ˆτ ext and ˆτ ext i etiated by uing not a force enor but the reaction torque oberver (RTOB). T 1 denote delay tie fro the ater ide to the lave ide and T 2 denote delay tie fro the lave ide to the ater ide. C 1, C 4, C and C are poition control paraeter, and C 2, C 3, C 5 and C 6 are force control paraeter. In the propoed three-channel control architecture, control paraeter are et a (1) (3). Fig. 2 how experiental reult with tie delay in the cae uing the propoed three-channel architecture. Virtual and randolyfluctuating counication tie delay (50 [] T 1, T 2 150 []) i inerted into counication channel. In Fig. 2(a), it i hown that the poition repone of the ater and lave robot alot perfectly tracked each other. In addition, a hown in Fig. 2(b), the operator felt little anipulating force in free otion. The validity of the propoed ethod wa confired by experiental reult. C 1 = C = C p () (1) C 4 = C = 0 (2) C 2 = C 3 = C 5 = C 6 = C f (3) Therefore, ater and lave acceleration reference value are calculated a (4) and (5) Fig. 1. General four-channel architecture = C f (ˆτ ext + ˆτ ext e T 2 ), (4) θ re f =C p ()(θ re e T 1 θ re ) C f (ˆτ ext e T 1 + ˆτ ext ), (5) where C p () = K p + K v and C f = K f are a poition controller and a force controller, repectively. K p, K v and K f denote poition feedback gain, velocity feedback gain and force feedback gain, repectively. (a) Poition repone (b) Force repone Fig. 2. Experiental reult (propoed three-channel architecture) 6
Perforance Analyi of a Three-Channel Control Architecture for Bilateral Teleoperation with Tie Delay Paper Ryogo Kubo Noriko Iiyaa Kenji Natori Kouhei Ohnihi Hirotaka Furukawa Meber Student Meber Student Meber Senior Meber Non-eber Bilateral control i one of the control ethod of teleoperation yte. Huan operator can feel reaction force fro reote environent by ean of thi control chee. Thi paper preent a novel control architecture for bilateral teleoperation with/without tie delay. The propoed bilateral control yte ha three counication channel between ater and lave robot. In concrete ter, thi yte ha two traniion channel of poition and force inforation fro the ater ide to the lave ide and one traniion channel of force inforation fro the lave ide to the ater ide. The ater controller of the propoed three-channel teleoperation yte doe not include a poition controller, i.e. only force control i ipleented in the ater ide, in order to iprove operationality in the ater ide. The three-channel controller with tie delay a well a without tie delay give better perforance (higher tranparency) than other conventional controller uch a four-channel controller and o on. In the propoed controller, odel of a lave robot and counication tie delay are not required differently fro conventional ethod, and robut acceleration control i achieved by uing the diturbance oberver (DOB). Hybrid atrice are utilized to analyze four-channel and three-channel control yte. Traniion characteritic of force and poition inforation between ater and lave robot are clarified in the analyi. The validity of the propoed ethod i confired by experiental reult. Keyword: bilateral control, acceleration control, tie delay, tranparency, three-channel architecture, haptic 1. Introduction Bilateral teleoperation robot have been utilized in pace, water, nuclear power plant and o on. In thi yte, the huan operator anipulate the ater robot, and then the lave robot track the ater robot. If the lave robot ha contact with the environent, the reaction force fro the environent i tranitted to the ater ide. Therefore, the huan operator can feel reaction force fro the reote environent a if he/he touche the environent directly (1) (2). By uing thi chee, he/he can work afely and effectively. Tranparency i one of the evaluation indice of bilateral teleoperation yte (3). High tranparency ean that the force control and the poition control are achieved perfectly both in the ater ide and in the lave ide. In fact, the force control ha to be achieved with accuracy to atch ater and lave force repone, and the poition control ha to be alo achieved with accuracy to atch ater and lave poition repone. In bilateral control yte, however, decopoition of a bilateral control yte into two coponent, i.e. a force control yte and a poition control yte, i a ain Departent of Syte Deign Engineering, Keio Univerity 3-14-1, Hiyohi, Kouhoku-ku, Yokohaa 223-8522 Counication Media Reearch Group, Reearch Laboratorie, NTT DoCoMo, Inc. 3-5, Hikarino-oka, Yokouka 239-8536 proble. Thee two controller hould be deigned independently for ideal bilateral control. For above purpoe, oe reearcher have utilized the diturbance oberver (DOB) (4) and ode decopoition (5). In addition, traniion characteritic of force and poition inforation between ater and lave robot are clarified by uing hybrid atrice in the cae without tie delay (6). In teleoperation yte, counication tie delay between ater and lave robot i alo a eriou proble. Particularly, the ue of paivity-baed forali i very coon to deign bilateral controller for teleoperation yte with tie delay. Nieeyer et al. (7) defined wave variable baed on cattering theory (8), and deigned a paivitybaed bilateral controller. The Sith predictor i often utilized to copenate tie delay effect in bilateral teleoperation (9). However, thi ethod require a precie lave odel and a tie delay odel. Munir et al. (10) propoed a wavebaed controller with the Sith predictor. In thi ethod, the Kalan filter and an energy regulator were utilized to cope with uncertaintie of the lave odel and unpredictable and varying counication tie delay. The counication diturbance oberver (CDOB) wa propoed to etiate unpredictable and varying counication tie delay a network diturbance (ND) and to copenate tie delay effect (11) (12). However, the perforance i greatly affected by order rearrangeent of tranitted data due to network condition, c 2007 The Intitute of Electrical Engineer of Japan. 1224
Three-Channel Architecture for Bilateral Control with Tie Delay ince CDOB require lave odel including integral eleent. Meanwhile, there are a lot of bilateral control architecture with/without tie delay (13). The ot general architecture i the four-channel architecture which ha four counication channel (3). The controller utilize the inforation of the ater/lave poition and force. In addition, oe reearcher have been propoed three-channel architecture (14). For exaple, Hahtrudi-Zaad et al. (15) propoed two clae of architecture, the operator-force-copenated (OFC) architecture, which doe not have a traniion channel of force inforation fro the lave ide to the ater ide, and the environent-force-copenated (EFC) threechannel control architecture, which doe not have a traniion channel of force inforation fro the ater ide to the lave ide. The three-channel architecture propoed by Fleer et al. doe not have a traniion channel of poition inforation fro the lave ide to the ater ide, and it ha local poition feedback (16). However, ideal perforance (high tranparency) cannot be achieved, ince thee control chee are not baed on robut acceleration control (5) (6). In thi paper, a novel three-channel architecture for bilateral teleoperation i deigned baed on acceleration control. The propoed yte ha three counication channel between ater and lave robot. In concrete ter, thi yte ha two traniion channel of poition and force inforation fro the ater ide to the lave ide and one traniion channel of force inforation fro the lave ide to the ater ide. The ater controller of the propoed yte doe not include a poition controller, i.e. only force control i ipleented in the ater ide, in order to iprove operationality in the ater ide. The three-channel controller with tie delay a well a without tie delay give better perforance (higher tranparency) than other conventional controller uch a two-channel and four-channel controller. In the propoed controller, odel of a lave robot and counication tie delay are not required differently fro conventional ethod, and robut acceleration control i achieved by uing DOB. Hybrid atrice are utilized to analyze four-channel and three-channel control yte. Traniion characteritic of force and poition inforation between ater and lave robot are clarified in the analyi. The validity of the propoed ethod i confired by experiental reult. The copoition of thi paper i decribed a follow. Section 2 preent a general four-channel bilateral controller baed on acceleration control. The propoed three-channel architecture and the conventional four-channel architecture are decribed and analyzed by uing hybrid atrice in Section 3. In Section 4, experiental reult are hown. Finally, concluion are decribed in Section 5. 2. General Bilateral Control baed on Acceleration Control In thi ection, the diturbance oberver (DOB) and the reaction torque oberver (RTOB) ipleented for robut acceleration control are preented. In addition, a general fourchannel architecture for bilateral teleoperation baed on acceleration control i decribed. 2.1 Diturbance Oberver (DOB) In actual robot Fig. 1. Diturbance oberver Fig. 2. Equivalent yte of Fig. 1 control, diturbance torque τ di a well a generated torque i exerted on a robot. The diturbance oberver (DOB) etiate diturbance torque ˆτ di including external torque ˆτ ext and copute copenation current I cp for robut otion control (4). The block diagra of a 1-DOF (degree-offreedo) robot yte including DOB i hown in Fig. 1. In Fig. 1, g di, J, K t, θ re and are cut-off frequency of a lowpa filter (LPF) in DOB, robot oent of inertia, torque coefficient, a poition repone value and an acceleration reference value, repectively. The ubcript n denote a noinal value, and denote a Laplace operator. Fig. 2 how the equivalent yte of Fig. 1. By uing DOB, diturbance torque i input to the yte through a high-pa filter (HPF) equivalently a (1). θ re = 1 { } J J n 2 n τ di (1) + g di Ideal robut acceleration control i achieved, if /(+g di ) 0, i.e. g di. 2.2 Reaction Torque Oberver (RTOB) DOB i alo utilized a the reaction torque oberver (RTOB) for etiation of external torque (17). While RTOB etiate wider bandwidth force inforation than force enor, it require identification of frictional torque in advance. By ean of RTOB, etiated external torque ˆτ ext i given by ˆτ ext = g reac τ ext, (2) + g reac where τ ext i actual external torque and g reac i cut-off frequency of LPF in RTOB. Perfect torque etiation i achieved, if g reac /( + g reac ) 1, i.e. g reac.intheexperient of thi reearch, external torque exerted on a robot i etiated by uing not a force enor but RTOB. D 127 12 2007 1225
Fig. 4. Network repreentation of teleoperation yte Fig. 3. General four-channel architecture 2.3 General Four-Channel Architecture Two robot, i.e. a ater robot which i anipulated by a huan operator and a lave robot which contact reote environent are conidered here. Fig. 3 how a general four-channel architecture for bilateral teleoperation. In Fig. 3, θ re, θ re,ˆτ ext and ˆτ ext are the poition of the ater robot, the poition of the lave robot, the etiated external torque exerted on the ater robot and the etiated external torque exerted on the lave robot, repectively. The external torque exerted on the robot ˆτ ext and ˆτ ext are etiated by uing not force enor but RTOB. T 1 denote delay tie fro the ater ide to the lave ide and T 2 denote delay tie fro the lave ide to the ater ide. C 1, C 4, C and C are poition control paraeter, and C 2, C 3, C 5 and C 6 are force control paraeter. 3. Analyi of Four-Channel and Three-Channel Architecture In thi ection, the four-channel architecture hown in Section 2, the four-channel architecture with CDOB and the propoed three-channel architecture are analyzed by uing hybrid atrice. 3.1 Hybrid Paraeter Fig. 4 how network repreentation of teleoperation yte. Hybrid paraeter H 11, H 12, H 21 and H 22 are defined a (3). [ τ ext θ re ] = [ ] H11 H 12 H 21 H 22 } {{ } H [ θ re τ ext ] (3) Perfect tranparency i achieved if H 11 = H 22 = 0and H 12 = H 21 = 1. In thi cae, (4) and (5) are atified. τ ext θ re = τ ext (4) = θ re (5) 3.2 Conventional Four-Channel Architecture The four-channel architecture hown in Fig. 3 i analyzed here. In thi analyi, it i aued that cut-off frequency of LPF in DOB and RTOB approache infinity (g di and g reac ) and ideal robut acceleration control i achieved. In Fig. 3, the bilateral yte i expreed a (6) and (7), if DOB work ideally and diturbance torque including paraeter fluctuation i uppreed copletely, i.e. θ re and θ re = θ re f / 2 are atified. = / 2 ( 2 + C )θ re = C 6 ˆτ ext ( 2 + C )θ re = C 3 e T1 ˆτ ext C 4 e T 2 θ re C 2 e T2 ˆτ ext (6) C 1 e T 1 θ re C 5 ˆτ ext (7) Here, the four-channel bilateral control yte baed on acceleration control i preented (6). In (6) and (7), control paraeter are et a (8) and (9) C 1 = C 4 = C = C = C p (), (8) C 2 = C 3 = C 5 = C 6 = C f, (9) where C p () = K p +K v and C f = K f are a poition controller and a force controller, repectively. K p, K v and K f denote poition feedback gain, velocity feedback gain and force feedback gain, repectively. While PD control i ipleented for poition control, P control i ipleented for force control. Thi i becaue differentiation of force repone i too noiy for practical purpoe. Fro (6) (9), ater and lave acceleration reference value in the conventional four-channel architecture are calculated a (10) and (11). = C p ()(θ re e T2 θ re ) C f (ˆτ ext + ˆτ ext e T2 ) (10) = C p ()(θ re e T1 θ re ) C f (ˆτ ext e T1 + ˆτ ext ) (11) Then, hybrid paraeter can be calculated a (12) (15), if the external torque are etiated preciely by ean of RTOB, i.e. τ ext = ˆτ ext and τ ext = ˆτ ext are atified. T denote round-trip delay tie (T = T 1 + T 2 ). H 11 = 2 ( 2 + 2C p ()) + C p () 2 (e T 1) (12) D() H 12 = C f ( 2 + 2C p ())e T 2 (13) D() H 21 = C f ( 2 + 2C p ())e T 1 (14) D() H 22 = C f 2 (e T 1) (15) D() { D() = C f 2 + C p ()(1 + e T ) } (16) If the yte doe not include counication tie delay (T 1 = T 2 = 0), hybrid paraeter are calculated a (17) (20). H 11 = 2 (17) C f H 12 = 1 (18) H 21 = 1 (19) H 22 = 0 (20) 1226 IEEJ Tran. IA, Vol.127, No.12, 2007
Three-Channel Architecture for Bilateral Control with Tie Delay Perfect tranparency i achieved in the conventional fourchannel architecture without tie delay, if the force control paraeter C f i large enough. 3.3 Propoed Three-Channel Architecture The propoed bilateral control yte ha three counication channel between ater and lave robot. In concrete ter, thi yte ha two traniion channel of poition and force inforation fro the ater ide to the lave ide and one traniion channel of force inforation fro the lave ide to the ater ide. The ater controller of the propoed three-channel teleoperation yte doe not include a poition controller, i.e. only force control i ipleented in the ater ide, in order to iprove operationality in the ater ide. In the propoed controller, odel of a lave robot and counication tie delay are not required differently fro conventional ethod uch a the Sith predictor, CDOB and o on. Here, control paraeter in (6) and (7) are et a (21) (23) differently fro the four-channel architecture. C 1 = C = C p () (21) C 4 = C = 0 (22) C 2 = C 3 = C 5 = C 6 = C f (23) In the propoed three-channel architecture, ater and lave acceleration reference value are calculated a (24) and (25). θ re f θ re f = C f (ˆτ ext + ˆτ ext e T2 ) (24) = C p ()(θ re e T1 θ re ) C f (ˆτ ext e T1 + ˆτ ext ) (25) Then, hybrid paraeter can be calculated a (26) (29), if the external torque are etiated preciely by ean of RTOB, i.e. τ ext = ˆτ ext and τ ext = ˆτ ext are atified. H 11 = 2 (26) C f H 12 = e T 2 (27) H 21 = e T 1 (28) H 22 = C f (e T 1) (29) 2 + C p () If the yte doe not include counication tie delay (T 1 = T 2 = 0), hybrid paraeter are calculated a (30) (33). H 11 = 2 (30) C f H 12 = 1 (31) H 21 = 1 (32) H 22 = 0 (33) Perfect tranparency i alo achieved in the propoed threechannel architecture without tie delay, if the force control paraeter C f i large enough in the ae way a the cae of the conventional four-channel architecture. 3.4 Dicuion of Architecture with Tie Delay In cae that the yte include counication tie delay, all hybrid paraeter are function of a Laplace operator. Therefore, hybrid paraeter of the propoed three-channel Fig. 5. Gain characteritic of hybrid paraeter H 11 Fig. 6. Gain characteritic of hybrid paraeter H 12 Fig. 7. Gain characteritic of hybrid paraeter H 21 Fig. 8. Gain characteritic of hybrid paraeter H 22 architecture with tie delay are copared with thoe of the conventional four-channel architecture by uing Bode diagra. Fig. 5 Fig. 8 how gain characteritic of hybrid paraeter H 11, H 12, H 21 and H 22, repectively. Control paraeter are et to C p () = 400 + 40 and C f = 3. Delay tie i et to T 1 = T 2 = 100 [] (contant value). In thee figure, hybrid paraeter of the conventional four-channel architecture with/without CDOB (11) and the propoed three-channel architecture are plotted. Here, the coputation of hybrid paraeter in the conventional four-channel architecture with CDOB i decribed a a guide. In the conventional four-channel architecture with CDOB, ater and lave acceleration reference value are D 127 12 2007 1227
calculated a (34) and (35) =C p ()(θ re e T2 +θ cp cdob θre ) C f (ˆτ ext + ˆτ ext e T2 ) =C p ()(θ re e T1 θ re ) C f (ˆτ ext + ˆτ ext e T2 ), (34) =C p ()(θ re e T1 θ re ) C f (ˆτ ext e T1 + ˆτ ext ), (35) where θ cp cdob = θre e T1 θ re e T2 denote a copenation value output fro CDOB (12). Then, hybrid paraeter of the conventional four-channel architecture with CDOB can be calculated a (36) (39), if the external torque are etiated preciely by ean of RTOB, i.e. τ ext = ˆτ ext and τ ext = ˆτ ext are atified. H 11 = 2 (36) C f H 12 = C { f ( 2 + C p ())e T 2 + C p ()e 1} T (37) D() H 21 = e T1 (38) H 22 = C f 2 (e T 1) (39) D() { D() = C f 2 + 2C p () } (40) In Fig. 5, it i hown that the gain characteritic of the propoed three-channel architecture and the conventional fourchannel architecture with CDOB are not affected by tie delay. However, they have large gain in high-frequency area. Thu, the force control paraeter C f ha to be et to a larger value to uppre the gain of H 11. Both in Fig.6 and in Fig. 7, it i hown that the gain characteritic of the propoed three-channel architecture are zero in all-frequency area. Therefore, the propoed three-channel architecture atifie the condition H 12 = H 21 = 1 (0 db), while the other architecture do not atify the condition. In Fig. 8, all three kind of architecture have quite all gain in all-frequency area. The condition H 22 = 0ialotatifiedbyuingevery architecture dicued in thi ection. The poition control paraeter C p () ha to be et to a larger value to uppre the gain of H 22. 4. Experient In thi ection, experiental reult are hown to confir the validity of the propoed three-channel architecture in bilateral teleoperation with tie delay. 4.1 Experiental Setup The experient are perfored by uing 1-DOF ater and lave anipulator hown in Fig. 9. In the experient, the lave anipulator had contact with aluinu rod in contact otion. Paraeter utilized in the experient are lited in Table 1. 4.2 Experiental Reult Fig. 10 Fig. 12 how experiental reult with tie delay in the cae uing the four-channel architecture without CDOB, the four-channel architecture with CDOB and the propoed three-channel architecture, repectively. Virtual and randoly-fluctuating counication tie delay (50 [] T 1, T 2 150 []) wa iulated on a coputer. In Fig. 10(b), Fig. 11(b) and Fig. 12(b), the ater force repone ultiplied by ( 1) are Fig. 9. 1-DOF ater and lave anipulator Table 1. Paraeter in experient Noinal oent of inertia J n 0.00535 [kg 2 ] Noinal torque contant K tn 0.156 [N/A] Poition feedback gain K p 400 Velocity feedback gain K v 40 Force feedback gain K f 3 Cut-off frequency of DOB g di 200 [rad/] Cut-off frequency of RTOB g reac 200 [rad/] Control period t 0.1 [] (a) Poition repone (b) Force repone Fig. 10. Experiental reult (conventional four-channel architecture without CDOB) plotted to copare the ater repone with the lave repone. In Fig. 10(a), it i hown that the poition repone of the ater and lave robot alot perfectly tracked each other. However, the operator felt large anipulating force in free otion a hown in Fig. 10(b). Thi i becaue the ater and lave robot tried to track each other, although the counication line between the ater and lave ide had tie delay eleent. Epecially in free otion, it wa true that the poition repone of the ater and lave robot alot perfectly tracked each other, but the operator could not anipulate the ater robot o quickly becaue of the large anipulating force. Therefore, the conventional four-channel architecture without CDOB generated the large anipulating force in free otion. In Fig. 11(a), it i hown that the poition repone of the ater and lave robot alot perfectly tracked each other. In addition, the operator did not feel large anipulating force 1228 IEEJ Tran. IA, Vol.127, No.12, 2007
Three-Channel Architecture for Bilateral Control with Tie Delay (a) Poition repone (b) Force repone Fig. 11. Experiental reult (conventional four-channel architecture with CDOB) otion wa achieved at the ae level a conventional ethod. However, perfect tranparency wa not achieved differently fro the previou analyi uing hybrid paraeter. Thi i becaue cut-off frequency of LPF in DOB and RTOB wa et to 200 rad/ in the experient, while it wa aued in the analyi that ideal acceleration control (g di )and ideal force etiation (g reac ) were achieved. The propoed three-channel architecture achieved precie force feedback to the extent poible both in free otion and in contact otion. The validity of the propoed three-channel architecture wa confired by the experiental reult. 5. Concluion In thi paper, a novel architecture for bilateral teleoperation wa propoed. The propoed bilateral control yte ha three counication channel with/without tie delay, and the ater controller doe not include a poition controller. The traniion characteritic of force and poition inforation between ater and lave robot were clarified in the analyi uing hybrid paraeter. In the experient, it wa hown that the three-channel architecture with tie delay gave better perforance than the other conventional ethod. A future work, tability analyi of the propoed threechannel architecture ha to be perfored firtly. Then the control bandwidth hould be brought cloe to the ideal condition (g di, g reac ) to a axiu extent. (Manucript received March 30, 2007, revied Aug. 3, 2007) (a) Poition repone Reference (b) Force repone Fig. 12. Experiental reult (propoed three-channel architecture) in free otion a hown in Fig. 11(b). However, the force repone of the conventional four-channel architecture with CDOB had oe error in the teady tate (20 25 ). Thi i becaue CDOB require lave odel including integral eleent and the perforance i greatly affected by order rearrangeent of tranitted data due to network condition. Therefore, the conventional four-channel architecture with CDOB wa uceptible to network condition becaue of the integral eleent, while it could reduce the anipulating force in free otion. In Fig. 12(a), it i hown that the poition repone of the ater and lave robot alot perfectly tracked each other a in the cae of the conventional four-channel architecture with CDOB. A hown in Fig. 12(b), the operator felt le anipulating force in free otion copared to the other cae, and oe error of the force repone in the teady tate decreaed. In addition, force feedback to the operator in contact ( 1 ) B. Hannaford: A Deign Fraework for Teleoperator with Kinethetic Feedback, IEEE Tranaction on Robotic and Autoation, Vol.5, No.4, pp.426 434 (1989) ( 2 ) Y. Yokokohji and T. Yohikawa: Bilateral Control of Mater-Slave Manipulator for Ideal Kinethetic Coupling Forulation and Experient, IEEE Tranaction on Robotic and Autoation, Vol.10, No.5, pp.605 620 (1994) ( 3 ) D.A. Lawrence: Stability and Tranparency in Bilateral Teleoperation, IEEE Tranaction on Robotic and Autoation, Vol.9, No.5, pp.624 637 (1993) ( 4 ) K. Ohnihi, M. Shibata, and T. Murakai: Motion Control for Advanced Mechatronic, IEEE/ASME Tranaction on Mechatronic, Vol.1, No.1, pp.56 67 (1996) ( 5 ) Y. Matuoto, S. Katura, and K. Ohnihi: An Analyi and Deign of Bilateral Control Baed on Diturbance Oberver, Proceeding of the 10th IEEE International Conference on Indutrial Technology, pp.802 807 (2003) ( 6 ) W. Iida and K. Ohnihi: Reproducibility and Operationality in Bilateral Teleoperation, Proceeding of the 8th IEEE International Workhop on Advanced Motion Control, pp.217 222 (2004) ( 7 ) G. Nieeyer and J.J.E. Slotine: Stable Adaptive Teleoperation, IEEE Journal of Oceanic Engineering, Vol.16, No.1, pp.152 162 (1991) ( 8 ) R.J. Anderon and M.W. Spong: Bilateral Control of Teleoperator with Tie Delay, IEEE Tranaction on Autoatic Control, Vol.34, No.5, pp.494 501 (1989) ( 9 ) K.B. Fite, M. Goldfarb, and A. Rubio: Loop Shaping for Tranparency and Stability Robutne in Tie-Delayed Bilateral Teleanipulation, Tranaction of the ASME, Vol.126, pp.650 656 (2004) ( 10) S. Munir and W.J. Book: Internet-baed Teleoperation Uing Wave Variable with Prediction, IEEE Tranaction on Mechatronic, Vol.7, No.2, pp.124 133 (2002) (11) K. Natori, T. Tuji, K. Ohnihi, A. Hace, and K. Jezernik: Robut Bilateral Control with Internet Counication, Proceeding of the 30th Annual Conference of the IEEE Indutrial Electronic Society, pp.2321 2326 (2004) (12) N. Iiyaa, K. Natori, R. Kubo, K. Ohnihi, H. Furukawa, K. Miura, and M. Takahata: A Bilateral Controller Deign Method Uing Delay Copenator, Proceeding of the IEEE International Conference on Indutrial Technology, pp.836 841 (2006) D 127 12 2007 1229
(13) K. Hahtrudi-Zaad and S.E. Salcudean: Analyi of Control Architecture for Teleoperation Syte with Ipedance/ Adittance Mater and Slave Manipulator, International Journal of Robotic Reearch, Vol.20, No.6, pp.419 445 (2001) ( 14) J.J. Abbott and A.M. Okaura: Stable Forbidden-Region Virtual Fixture for Bilateral Teleanipulation, Journal of Dynaic Syte, Meaureent, and Control, Vol.128, pp.53 64 (2006) (15) K. Hahtrudi-Zaad and S.E. Salcudean: Tranparency in Tie-Delayed Syte and the Effect of Local Force Feedback for Tranparent Teleoperation, IEEE Tranaction on Robotic and Autoation, Vol.18, No.1, pp.108 114 (2002) ( 16) H. Fleer and J. Wikander: Tranparency and Stability Analyi of a Surgical Teleoperator Syte, Proceeding of the 11th Sypoiu on Haptic Interface for Virtual Environent and Teleoperator Syte, pp.382 389 (2003) ( 17) T. Murakai, F. Yu, and K. Ohnihi: Torque Senorle Control in Multidegree-of-Freedo Manipulator, IEEE Tranaction on Indutrial Electronic, Vol.40, No.2, pp.259 265 (1993) Ryogo Kubo (Meber) received the B.E. degree in yte deign engineering and the M.E. degree in integrated deign engineering fro Keio Univerity, Yokohaa, Japan, in 2005 and 2007, repectively. In 2007, he joined Nippon Telegraph and Telephone (NTT) Corporation. He i currently engaged in reearch and developent of optical acce network and yte at NTT Acce Network Service Syte Laboratorie, Chiba, Japan. He i a Meber of the IEEE, the Society of Intruent and Control Engineer (SICE) and the Intitute of Electronic, Inforation and Counication Engineer (IEICE). Noriko Iiyaa (Student Meber) received the B.E. degree in yte deign engineering fro Keio Univerity, Yokohaa, Japan, in 2006. She i currently working toward the M.E. degree in integrated deign engineering at Keio Univerity. Her reearch interet include robotic, otion control and haptic. Kenji Natori (Student Meber) received the B.E. degree in yte deign engineering and the M.E. degree in integrated deign engineering fro Keio Univerity, Yokohaa, Japan, in 2004 and 2006, repectively. He i currently working toward the Ph.D. degree at Keio Univerity. Fro April 2007, he i a Reearch Fellow of the Japan Society for the Prootion of Science (DC2). Hi reearch interet include tie delay yte, bilateral control and network-baed control yte (NBCS). He i a Student Meber of the IEEE. Kouhei Ohnihi (Senior Meber) received the B.E., M.E., and Ph.D. degree in electrical engineering fro the Univerity of Tokyo, Tokyo, Japan, in 1975, 1977 and 1980, repectively. Since 1980, he ha been with Keio Univerity, Yokohaa, Japan. Hi reearch interet include echatronic, otion control, robotic and haptic. Prof. Ohnihi received Bet Paper Award fro the IEEJ and the Japan Society for Preciion Engineering, and Outtanding Paper Award at IECON 85, IECON 92, IECON 93 and IECON 05. He alo received the EPE-PEMC Council Award and the Dr.-Ing. Eugene Mittelann Achieveent Award fro the IEEE Indutrial Electronic Society in 2004. He i a Fellow of the IEEE. Hirotaka Furukawa (Non-eber) received the B.E. and M.E. degree in applied phyic and cheitry fro the Univerity of Electro-Counication, Tokyo, Japan, in 2001 and 2003, repectively. Since 2003, he ha been with NTT DoCoMo, Inc. He i currently a developent engineer at Counication Device Developent Departent, Kanagawa, Japan. 1230 IEEJ Tran. IA, Vol.127, No.12, 2007