A Study on Selecting the Method of Constructing the Information to be Exchanged in Unlimited-workspace Bilateral Teleoperation

Transcription

1 TrC-IFToMM Symposum on Theory of Machnes and Mechansms, Izmr, Turkey, June 14-17, 2015 A Study on Selectng the Method of Constructng the Informaton to be Exchanged n Unlmted-workspace Blateral Teleoperaton M. G. Karabulut * M. İ. C. Dede B. Taner İzmr Insttute of Technology İzmr Insttute of Technology İzmr Insttute of Technology İzmr, Turkey İzmr, Turkey İzmr, Turkey Abstract In ths paper, a study on selectng the mappng method for nformaton exchange n unlmtedworkspace teleoperaton s presented. In spte the fact that n most of the blateral teleoperaton systems, the master system sends moton demands and receves nteracton force from the slave system, the nformaton to be exchanged through the communcaton lne between master and slave s selected to be force n both drectons n ths study. Ths approach s expected to ease the navgaton of the user when a lmted-workspace master system s used to control an unlmted-workspace slave system. As the unlmted-workspace slave system, a vrtual flyng scalpel s used and human skn s modeled to represent the envronment around the slave system. Two methods of constructng force nformaton, or n other words, nformaton mappng between the two systems, are developed and evaluated va user studes. However, one of them comes out to provde acceptable results n the selected unlmted-workspace teleoperaton task. Expermental results for the method that provdes acceptable results are presented. Keywords: unlmted-workspace teleoperaton, workspace mappng, blateral teleoperaton. I. Introducton 1 Teleoperaton systems allow the human operators to execute tasks n remote or hazardous envronments, such as space, underwater, nuclear plants, medcne and mltary applcatons. Havng a crucal role n robotcs, many scentfc studes dealng wth the teleoperaton have been done n the last decades [1, 2]. In blateral teleoperaton, the master and the slave systems are coupled wth a communcaton lne n whch nformaton flows n both drectons (from master to slave and slave to master). Blateral teleoperaton systems can be grouped nto two regardng ther lmtatons of the workspace, as unlmted- and lmted-workspace teleoperaton. In general, the operator manpulates the slave robot from a dstant locaton. Therefore, the success of operaton depends on the nformaton suppled by n between the master and slave systems usng the communcaton lne [3, 4]. * mehmetkarabulut@yte.edu.tr candede@yte.edu.tr barstaner@yte.edu.tr 1 In unlmted-workspace teleoperaton, the man problem n nformaton exchange s on mappng a lmted-workspace master system to an unlmtedworkspace slave system moton. Mappng methods are generally chosen wth respect to the workspace lmts. The selecton crtera for the mappng methods are studed by Lee et al. [5], Elhaj et al. [6], Farkhatdnov and Ryu [7]. The mappng methods for blateral teleoperaton systems havng lmted-workspace slave robots have already been well-studed n the lterature [8, 9]. However, the workspace of slave and master devce s lmted by the smaller of the two. A soluton can be derved by scalng up or down the workspace that causes drop n the accuracy or lmts teleoperaton system workspace [10, 11]. Therefore, compared to the lmtedworkspace teleoperaton systems, unlmted-workspace teleoperaton systems have workspace mappng problem [5] due to knematc dversty of two sub-system. A proposed method to overcome for the stated problem above for unlmted-workspace teleoperaton systems s the poston-velocty mappng [5-7, 9, 10]. Postonvelocty mappng s only lmted by the slave devce's workspace, whch s theoretcally nfnte. On the other hand, ths type of control s nadequate n terms of postonal accuracy [9]. In order to provde a soluton to ths problem, a control method for the blateral teleoperaton of a moble robot that combnes postonposton and poston-speed was presented by Farkhatdnov and Ryu [7]. A haptc teleoperaton control algorthm based on knetc scrollng-based poston mappng for unmanned aeral vehcles was presented n [12]. Knetc scrollng, whch s adapted to ths control algorthm, s used n all modern smart phones, where the user can nteract wth the knetc scrollng n a natural and ntutve manner. In order to confgure an unlmted-workspace teleoperaton system, usually a vrtual nteracton force computed on the bass of obstacles surroundng the moble vehcle s used n to ncrease the user s percepton of the moble robot s workspace [13]. Only n ths way, ths type of a system can be called unlmted-workspace blateral teleoperaton system. In the case where a mappng scheme s selected based on moton capabltes, such as poston-velocty, the conventonal mappng strateges fal to reflect the forces n a sutable way, whch are acqured from the slave, to the human operator

2 TrC-IFToMM Symposum on Theory of Machnes and Mechansms, Izmr, Turkey, June 14-17, 2015 through the master system. The novelty of ths study s that we are attemptng to formulate a new way of mappng for unlmted-workspace blateral teleoperaton systems. In ths paper, after gvng a revew of the man concepts and defntons about unlmted-workspace blateral teleoperaton system, a comparatve study on two technques formulated for mappng the nformaton exchange between the master and the slave systems n unlmted-workspace blateral teleoperaton are presented. The developed methods are examned by desgnng a benchmark test scenaro. Novnt Falcon s used as the master devce and the vrtually constructed flyng scalpel s selected as the slave object. An nteracton wth a an object that has relatvely less mechancal mpedance (e.g. soft tssue) s modeled for the test scenaro. In order to evaluate the methods, the performance of operator n path trackng and task completon tme are measured n the tests. Fnally, the test results n terms of statstcal evaluatons are presented. As a result of ths study, t s shown that one of the proposed methods overcome the master workspace lmtatons and ncrease the performance of the user n teleoperaton task by selectng mappng parameters carefully. II. Unlmted-Workspace Teleoperaton If a robot's workspace s relatvely very large wth respect to the workspace of the master system's workspace or f t s practcally wthout any bounds, ths system can be defned as the unlmted-workspace teleoperaton. Unlmted-workspace teleoperaton s wdely used for controllng moble vehcles, underwater remotely operated vehcles (ROVs) and unmanned aeral vehcles (UAVs). However, n our study we selected the flyng scalpel as a slave system whch also has an unlmted-workspace when t s not attached to a robot arm. Unlmted teleoperaton system can be dvded nto two man categores as unlateral and blateral. A. Unlateral teleoperaton In unlateral teleoperaton, the slave system s drven by the operator s demands such as poston, velocty, or force whch s acqured by the master system s shown n Fg. 1. However, nformaton flows only n one drecton through the communcaton lne. In ths type of teleoperaton, there s not a feedback from the slave to nform operator about the slave envronment[3]. Fg. 1. Unlateral teleoperaton (no force feedback s avalable from the slave unt) 2 B. Blateral teleoperaton In blateral teleoperaton, the nformaton flows n two drectons, not only from the master to the slave system but also from the slave to the master system. In ths knd of a settng, the slave robot s drven by the operator s commands and the necessary feedback such as vsual, force, audtory feedback(s) are sent from the slave to the master system n order to nform operator about the slave envronment [8]. Ths helps the operator to ncrease hs feelng of beng present n the slave envronment. A common case n blateral teleoperaton s that moton demand s sent to the slave and force feedback s receved from the slave. Ths type s presented n Fg. 2. Fg. 2. Blateral teleoperaton (force-feedback sgnal s avalable from the slave to the master control unt) III. Selecton of the Informaton Exchange Method In ths secton, two methods are presented n generaton of moton commands that are sent to slave sub-system and force feedback that s appled to the operator. Both methods are devsed to ease the use of the teleoperaton system by the human operator and enhance the teleoperaton systems overall precson. In both methods, the moton of captured from the master system s mapped to the force demand to be sent to the slave system n order to control the slave object n ts unlmted-workspace. The nformaton between master and slave s delvered through the communcaton lne. The block dagram n Fg. 3, presents the nformaton flow n ths teleoperaton system settng. Fg. 3. Informaton flow scheme for the desgned blateral teleoperaton system In Fg. 3, F s s the force constructed n the master system to be sent as a demand to the slave system. F e s the nteracton forces measured at the slave system s envronment. F m s the force appled to the user through the haptc master system. The calculaton of the above nformaton dffers n each method presented n the next sub-sectons. A. Poston-force methodology In ths method, a vrtual sprng s ntroduced to the system n order to calculate a force demand to be sent to the slave system. The method has an analogy wth the gas pedals n automobles. For example, gas pedal s poston

3 TrC-IFToMM Symposum on Theory of Machnes and Mechansms, Izmr, Turkey, June 14-17, 2015 s mapped to the torque to be suppled to the drvng shaft through regulatng the fuel and ar supply to the engne. Therefore, the drver controls the acceleraton of the car by changng the poston of the gas pedal. In addton, the gas pedal returns to ts ntal poston when t s not pressed by the drver. Smlarly, the vrtual sprng used n ths method brngs the haptc master system s end-efector to ts ntal poston when the user s not applyng any forces on ts handle. The am of usng the vrtual sprng force whch s appled to the master s to drect the operator to the determned md-pont of Falcon workspace. F s, s the restorng force produced from the dsplacement of the vrtual sprng by the human operator nput. The vrtually constructed force, F s, s sent to the slave system as demand. The calculaton of the F s force along each drecton of moton wth respect to the coordnate frame F{O; u ; = 1,2,3} s provded n Eq. 1. F s = k p m u ; = 1, 2, 3 (1) In Eq. 1, p m s the poston of the master handle wth respect to the ntal poston along the u axs, and k s the vrtual lnear sprng constant along every drecton of moton. F m s the force appled through the master system to the human operator along the u axs. Ths force s calculated by summng the transmtted slave envronmental forces, F e, and restorng force as shown n Eq. 2. F m = F s + F e (2) The restorng force sent to the slave s used as force command n ts equaton of moton as presented n Eq. 3. F s = m s a s + Bv s + F e ; = 1,2,3 (3) We ntroduce the vscous frcton constant, B, for the system to lmt slave object moton n free space. Otherwse, even though the master handle s at the ntal poston, the slave object contnues ts moton under the gven force by user. a s s the acceleraton of slave where the master poston s mapped. v s s the velocty of the slave object. m s s the mass of slave object, whch s defned as a scalpel. From the gven equatons, the slave acceleraton can be derved as follows: a s = F s Bv s F e m s ; = 1, 2, 3 (4) Fnally, the master poston s mapped to the slave system s acceleraton by creatng vrtual sprng n master system and transmttng the nformaton from the master to the slave n terms of forces. As a result of ths, the slave object could be drven n unlmted workspace, though the master system has a lmted workspace. B. Velocty-force methodology In ths method, a vrtual damper s ntroduced n master system, so that the nformaton can be sent as a force demand to the slave system. In ths case, F s, s the force nformaton to be sent to the slave as the demand. Ths force calculaton s defned n Eq. 5. F s = cv m ; = 1, 2, 3 (5) In Eq. 5, v m s the relatve velocty of the master handle, and c s the dampng constant. After sendng force demand to the slave system, ths force nformaton s used n the dynamc equaton of moton of the slave as t was presented n Eq. 3 and the respectve acceleraton of the slave s calculated by usng Eq. 4. As a result, the velocty of the master system s mapped to the acceleraton of slave by means of creatng vrtual dampng n the master system. Ths methodology allows us to delver the nformaton to the slave system as a dampng force through the communcaton lne. C. The contact model The sprng-damper model s used to smulate the nteracton forces between two vrtual bodes, whch are placed at the slave system s envronment. In Fg. 4, sprng-damper representaton of the slave system and the poston of scalpel body, x d, relatve to the contact surface of human skn are shown. There s no nteracton force appled on master when x d > 0. An nteracton force s generated by sprng-damper equaton dependng on the poston and the velocty of the scalpel when x d <0. Fg. 4. Sprng-damper contact model It should be noted that the modeled skn s assumed to be on the u 2 u 3 plane. Therefore, nteracton force at the slave s envronment along the normal of the surface, u 1 axs, F e, s calculated as presented n Eq. 6. F e 1 = m s x d 1 + c e x d 1 + k e x d 1 (6) In Eq. 6, k e s the sprng constant and c e s the dampng constant. D. Vscous resstance When the object s movng through a flud at relatvely slow speeds, the equaton of vscous resstance s appled 3

4 TrC-IFToMM Symposum on Theory of Machnes and Mechansms, Izmr, Turkey, June 14-17, 2015 n opposte drecton of movement. Smlarly, durng cuttng operaton n surgery, the assumpton made n ths work s that the surgeon feels a low resstance (drag) force stemmng from the nteracton between scalpel and the human skn. Ths force s approxmately proportonal to velocty, but opposte n the drecton of moton. The equaton of the vscous resstance force calculaton along the u 2 and u 3 axs s gven n Eq. 7. Mappng strateges tested by takng force commands from the user through Novnt Falcon. Then the dsplacements and veloctes of the master devce along u 1, u 2 and u 3 axes (Fg. 6) are used n the mappng procedure. F e = bv d ; = 2, 3 (7) In Eq. 7, F e s the vscous resstance force at the slave s envronment along the u 2 and the u 3 axs. b s a vscous frcton constant, whch depends on the human skn characterstcs. v d s the velocty of the slave object along the u axs. IV. Expermental Tests The moton demand from the master devce (Novnt Falcon) s acqured along the u 1, u 2 and u 3 axes as shown n. A. Expermental test setup Teleoperaton systems mentoned n ths work are tested n a sem-expermental test setup, shown n Fg. 6. The test setup s composed of a haptc master devce and a vrtual slave system that s modeled and vsualzed n Matlab Smulnk. The master devce used n the tests are Novnt Falcon, whch s a 3 degrees-of-freedom haptc devce represented wth 1 n Fg. 5. The vrtual slave system conssts of a scalpel, whch has an unlmted-workspace, and a relatvely low mpedance object (e.g. human skn), whch s defned as a vrtual wall. The vrtual scalpel and the vrtual object's surface to be nteracted wth are ndcated n Fg. 5 wth 2 and 3 respectvely. The object to be nteracted wth s modeled based on a lnear mass sprng damper model n Matlab Smulnk. It should be noted that the scalpel s not connected to a robot arm but t s movng freely wthout bounds n ts vrtual envronment. Fg. 6. Novnt Falcon Haptc Devce The test setup s samplng rate s set to 500 Hz. The task s desgned so that the operator tracks three tmes the path ndcated wth the red lne on the skn surface n Fg. 7. The scalpel s ndcated wth (1) n Fg. 7 and the skn to be nteracted wth s ndcated wth (2). x d s the dstance between bodes along the u 1 axs. Fg. 7. Vrtual realty representaton of the slave system The user s exposed to the forces appled through the master system along the 3 axes as mentoned n the methodology secton. The normal force s the summaton of the restorng force and the contact force along the u 1 axs. Whle the operator moves nsde the human skn, the vscous resstance force as well as the restorng force s appled along the u 2, and u 3 axes to smulate cuttng process. The magntudes of generated forces by the master system n N are shown n Fg. 8 n whch the contact moment s ndcated wth the dashed lne. Fg. 5. Expermental setup composed of a haptc master devce and a vrtual slave system 4

5 TrC-IFToMM Symposum on Theory of Machnes and Mechansms, Izmr, Turkey, June 14-17, 2015 Fg. 8. Forces appled to the user along each axs by the haptc devce B. Expermental test results The operator carres out the task usng the veloctyforce nformaton exchange technque. The task s carred out by 10 dfferent operators. One of the operator s test results s shown n Fg. 9. The route to be followed s shown n blue colored sold lne and the recordng from user s moton trackng s shown n red colored sold lne. Durng the task, the user must complete 3 rounds around the assgned zone. The am of ths task s not to go out of assgned red zone shown n Fg. 7 whle the scalpel s nsde of the human skn. In order to make a statstcal assessment, the user s path trackng performance as well as the task completon tme s recorded after the contact between the scalpel and the human skn s acheved. The tests are performed 10 tmes for velocty-force method by assgnng 0 and 2 values for the vscous frcton constant B n free moton. When vscous frcton constant B s selected to be 0, the user s able to control the slave object wthout a workspace lmtaton snce there s no opposte force to the moton. On the other hand, when the vscous frcton constant B s selected to be larger than 0, the resstance force works aganst to the moton of the slave object. Therefore, the slave workspace s partally lmted. However, the control of the slave s moton becomes more convenent for the user. In our case, the optmal value s found to be 2 consderng the ease of use and the slave workspace lmtaton. Fg. 9. The test sample recordng for velocty-force method that shows the assgned route to be followed and the user performance (a) B=0; (b) B=2 C. Evaluaton of the test results For each node along the user s tracked lne, the dstance to the assgned route s named as d and ths s shown n Fg. 9 (a). The calculaton method for d s shown n Fg. 10. From each node, the dstance to the lnes s calculated. Fg. 10. Pont-lne dstance In Fg. 9, the lne s specfed by selected two ponts x 1 =(x 1, y 1 ) and x 2 =(x 2, y 2 ). Then, the vector perpendcular to the lne s gven by: y 2 y 1 v = [ (x 2 x 1 ) ] (8) 5

6 TrC-IFToMM Symposum on Theory of Machnes and Mechansms, Izmr, Turkey, June 14-17, 2015 Let r be a vector from the pont x 0 =(x 0, y 0 ) to the frst pont on the lne: r = [ x 1 x 0 y 1 y 0 ] (9) The dstance d from (x 0, y 0 ) to the lne agan gven by projectng r onto v s calculated n Eq. 10. d = v. r = (x 2 x 1 )(y 1 y 0 ) (x 1 x 0 )(y 2 y 1 ) (x 2 x 1 ) 2 +(y 2 y 1 ) 2 (10) Ths calculaton s repeated for the four lnes that consttute the path. The smallest d s selected to be the offset from the mddle of the red zone lne at that pont. As a restrcton, the operator must reman wthn the red zone on the surface durng the operaton. When the operator goes out of the red zone, then the error s calculated by countng the pkes whch are hgher than 0.03 m. The upper lmt s shown wth red dashed lne and the top of pkes hgher than dashed lne are marked wth the red dots as seen n Fg. 11. Fg. 11. The operator s error calculaton D. Statstcal assessment Test results whch have been recorded durng the operatons are tabulated n TABLE 1 and Table 2. The tests are performed 10 tmes by usng dfferent value for the vscous frcton constant, B. As a remnder, d s the dstance between desred path and the operator s path. Durng the teleoperaton task, scalpel must be controlled n a desred depth nsde the human skn. To do so, the dstance between the tp pont of the scalpel and the desred depth nsde the human skn s defned as x. For each tral, mean value and the standard devaton of dstance d and x are gven as well as error and the task competton tme. Comparatve assessment s done after nspectng the collected data. 6 Trals velocty-force method (B=0) mean d (mm) mean x (mm) std d (mm) std x (mm) tme (sec) error Avg TABLE 1. Experment results for velocty-force method (B=0) Trals velocty-force method (B=2) mean d (mm) mean x (mm) std d (mm) std x (mm) tme (sec) error Avg TABLE 2. Experment results for velocty-force method (B=2) The overall averages of all the recorded test data are tabulated n TABLE 3. The standard devaton on the average of d and x dstances are relatvely hgh. The teleoperaton system s examned by dfferent users wth varous levels of hand-eye coordnaton performance and ths s why there s a relatvely large value for standard devaton. It s observed that the user error s reduced by ncreasng the vscous frcton constant, B. In addton, the magntude of dstances d and x are decreased when the vscous frcton constant B s selected as 2. The task completon tme s also decreased relatvely. In concluson, a more convenent usage condton s provded to the operator by ntroducng an optmal the vscous frcton for the moton of the slave system. TABLE 3. Overall averages of the results for velocty-force method V. Concluson Overall Avg. velocty-foce method B=0 B=2 mean d (mm) mean x (mm) std d (mm) std x (mm) tme (sec) error In ths paper, an unlmted-workspace teleoperaton system s consdered. The nformaton exchange between the master and the slave s selected to be force n both drectons. Ths s done to accommodate the control of an

7 TrC-IFToMM Symposum on Theory of Machnes and Mechansms, Izmr, Turkey, June 14-17, 2015 unlmted-workspace slave wth a lmted-workspace master. In order to construct force nformaton on the master sde to send to the slave system, two methods are developed and examned. Among these, the poston to force method dd not produce acceptable results snce the operators durng the test were not able fnsh the tests. Usng the velocty to force methodology, the tests were able to be carred out by the operators wth success. However, n order to ease the control of the user a vscous frcton effect s added for the slave system s equaton of moton. The addton of ths effect provded better performance results as t can be observed from the statstcal test results. In teleoperaton systems, one of the man areas of study s the stablty of the systems under communcaton tme delays. As a future work of ths study, our am s to nvestgate the performance of ths nformaton exchange methodology n tme delayed teleoperaton systems and develop necessary algorthms to guarantee a stable operaton. Acknowledgement The authors would lke to thank to The Scentfc and Technologcal Research Councl of Turkey for fundng the research presented n ths work. (grant number 113E147) References [1] T. B. Sherdan, Telerobotcs, automaton, and human supervsory control: MIT Press, 1992 [2] T. Fong and C. Thorpe, Vehcle Teleoperaton Interfaces, Autonomous Robots, 11, 9-18, 2001 [3] E. Uzunoğlu and D. M. I. Can, İletşm hatalarına maruz kalan k yönlü teleoperasyon sstemnn gelştrlmş kuvvet takb performanslı model-aracılı denetm, Türkye Otomatk Kontrol Komtes Toplantısı, Malatya, [4] O. N. Şahn, T. Erş, and M. İ. C. Dede, "Unlmted- Workspace Teleoperaton wth Obstacle Avodance Capablty," presented at the Natonal Meetng of Turksh Natonal Automatc Control Commttee, [5] D. Lee, O. Martnez-Palafox, and M. W. Spong, Blateral teleoperaton of a wheeled moble robot over delayed communcaton network, the Internatonal Conference on Robotcs and Automaton, Orlando, [6] I. Elhajj, N. X, and Y.-h. Lu, "Real-Tme Control of Internet Based Teleoperaton wth Force Reflecton," San Franssco, [7] I. Farkhatdnov and J.-H. Ryu, "Hybrd postonposton and poston-speed command strategy for the blateral teleoperaton of a moble robot," presented at the Control, Automaton and Systems, [8] P. F. Hokayem and M. W. Spong, "Blateral teleoperaton: An hstorcal survey," Automatca, vol. 42, pp , December, [9] M. R. Wrock and S. B. Nokleby, "Haptc Teleoperaton of a Manpulator usng Vrtual Fxtures and Hybrd Poston-Velocty Control," presented at the 13th World Congress n Mechansm and Machne Scence, Guanajuato, Mexco, [10]I. Farkhatdnov, J.-H. Ryu, and J. Podurayev, "Control Strateges and Feedback Informaton n Moble Robot Teleoperaton," presented at the Proceedngs of the 17th World Congress, Seoul, Korea, [11]A. Bettn, P. Marayong, S. Lang, A. M. Okamura, and G. D. Hager, "Vson-asssted control for manpulaton usng vrtual fxtures," Robotcs, IEEE Transactons on, vol. 20, pp , [12]A. Ruesch, A. Y. Mersha, S. Stramgol, and R. Carlon, "Knetc scrollng-based poston mappng for haptc teleoperaton of unmanned aeral vehcles," n Robotcs and Automaton (ICRA), 2012 IEEE Internatonal Conference on, 2012, pp [13]N. Dolat and C. Melchorr, "Teleoperaton of a moble robot through haptc feedback," n Haptc Vrtual Envronments and Ther Applcatons, IEEE Internatonal Workshop 2002 HAVE, 2002, pp