MANIPULATION of objects of less than ten micrometers

Transcription

1 IEEE/ASME TRANSACTIONS ON MECHATRONICS 1 Haptc Teleoperaton for 3-D Mcroassembly of Sphercal Objects Aude Bolopon, Hu Xe, Member, IEEE, Dogan Snan Halyo, and Stéphane Régner Abstract In ths paper, teleoperated 3-D mcroassembly of sphercal objects wth haptc feedback s presented. A dual-tp grpper controlled through a haptc nterface s used to pck-and-place mcrospheres (dameter: 4 6 μm). The proposed approach to algn the grpper wth the spheres s based on a user-drven exploraton of the object to be manpulated. The haptc feedback s based on ampltude measurements from cantlevers n dynamc mode. That s, the operator perceves the contact whle freely explorng the manpulaton area. The data recorded durng ths exploraton are processed onlne and generate a vrtual gude to pull the user to the optmum contact pont, allowng correct postonng of the dual tps. A prelmnary scan s not necessary to compute the haptc feedback, whch ncreases the ntutveness of our system. For the pck-and-place operaton, two haptc feedback schemes are proposed to ether provde users wth nformaton about mcroscale nteractons occurrng durng the operaton, or to assst them whle performng the task. As expermental valdaton, a two-layer pyramd composed of four mcrospheres s bult n ambent condtons. Index Terms Blateral couplng, dual-tp grpper, haptc nterface, mcroassembly, vrtual gude. Haptc F op, F h F hx, F hy, F fe hz, F hz as H(s) k x, k y, k z NOMENCLATURE User and haptc force, respectvely. Haptc force on x, y, z (ether for feelng or assstance) axes. Haptc devce transfer functon. Haptc stffness for the x, y, and z force feedback. n, d Number of ponts and dstance between two ponts for the vrtual gude generaton. z d Equlbrum poston for transport when the haptc feedback s provdng assstance. α a, α f, α d Ampltude, force, and dsplacement scalng factor, respectvely. Photodode A, F Ampltude and force measurement from the photododes, respectvely. A CP Resdual ampltude at full contact. A I, A II Ampltude output of Photododes I and II. A t Ampltude of the th recorded contact pont durng the tappng-mode exploraton. A 0, F 0 Ampltude when the tp s oscllatng at ts natural frequency, force when the tp s placed n contact wth the sphere before the graspng operaton. F ao Adheson forces between the sphere and the substrate. F I, F II F I0, F II0 k n S n V I, V II β Coordnates h 0 p h =[x h,y h,z h ] p n =[x n,y n,z n ] p p =[x p,y p,z p ] x CP x t, y t, z t x 0, y 0, z 0 z PO Force output of Photododes I and II. Force output of Photododes I and II when the tp s placed n contact wth the sphere before the graspng operaton. Stffness of the cantlever. Senstvty of the photodode. Voltage output of Photododes I and II. Calbrated converson factor of the photodode from voltage to ampltude. z-poston of the tps whle algnng the grpper. Coordnates of the haptc handle. Coordnates of the nanostage. Coordnates of the pezotube. x-poston of the tp at full contact. Coordnates of the th recorded contact pont durng the tappng-mode exploraton. Coordnates of the sphere center. z-poston of the pull off. Manuscrpt receved March 19, 2010; revsed September 18, 2010; accepted July 15, Recommended by Techncal Edtor S. Fatkow. Ths work was supported n part by the French Agency of Research (ANR) through the PAC- MAN Project and the NANOROL Project under Grant PSIROB The authors are wth the Insttut des Systèmes Intellgents et de Robotque, Centre Natonal de la Recherche Scentfque UMR 7222, Unversté Perre et Mare Cure Pars 6, Pars, France (e-mal: bolopon@sr.upmc.fr; xe@sr.upmc.fr; halyo@sr.upmc.fr; regner@sr.upmc.fr). Color versons of one or more of the fgures n ths paper are avalable onlne at Dgtal Object Identfer /TMECH I. INTRODUCTION MANIPULATION of objects of less than ten mcrometers s a challenge, as t shares many dffcultes wth nanomanpulaton. Among them, the vsual feedback from the optcal mcroscope s lmted and does not enable the accurate postonng of tools and objects. Due to the scale reducton, adheson forces become predomnant over gravtatonal ones [1]. Tools to manpulate these objects must be carefully desgned [2]. Fully automated mcromanpulaton s dffcult to acheve, gven /$ IEEE

2 2 IEEE/ASME TRANSACTIONS ON MECHATRONICS the hgh nfluence of envronmental parameters and the lack of repeatablty. Moreover, t would result n a lack of flexblty n the overall system. Haptc feedback appears as a promsng soluton to provde assstance to operators [3], n partcular for atomc force mcroscopy (AFM)-based manpulaton [4], [5]. Solutons to assst mcromanpulatons through haptc feedback are the frst steps toward ntutve nanomanpulatons. To provde hgh qualty force feedback, haptc couplng schemes have been developed to handle mcro and nanoscale specfcty [6], [7]. In partcular, passvty controllers, wdely used n macroscale teleoperaton [8], [9], have been adapted to mcroscale n order to preserve long-range attracton forces such as van der Waals nteractons [10], [11]. Based on these haptc couplng schemes, several teleoperated mcro- and nanoscale tasks are reported n the lterature. Early examples deal only wth feelng the shape of substrates or objects [12]. Marlere et al. [13] report a haptc mplementaton of a approach/retract task of an AFM probe.the frst remote tasks nducng a modfcaton of the sample are ndentatons (e.g., drect patternng on a substrate [14]). Tasks nvolvng pushng/pullng or cuttng objects are also of prmary nterest [15], [16]. As only two measurements bendng and torson are drectly avalable from an AFM cantlever, 3-D haptc feedback of nanoscale nteractons between the tool and the object can only be acheved by the use of contact mechancs models [16]. Such models are used for 3-D haptc feedback n surface ndentaton and touchng mcroobjects [17] [19]. However, so far, no manpulaton tasks have been reported. In addton to transmttng mcro- and nanoscale nteractons, haptc feedback s used as an enhancement for user assstance, wth vrtual gudes for pushng and pck-and-place by adheson tasks [20]. In ths case, haptc feedback s used to keep the user s moton on a specfed path. All the aforementoned works use a sngle AFM cantlever and statc measurement. Consequently, only one controllable contact pont s avalable to the user. Usng the AFM n contact mode mples also some lmtatons compared to usng t n dynamc mode. The dynamc mode, where the force measurement s obtaned through varatons on the ampltude or frequency of a vbratng probe, s generally consdered of fner qualty than the statc contact mode, where the measured force s drectly proportonal to the deflecton of the probe [21]. In addton, a frequent tme-consumng factor s the prelmnary scan, whch s requred, especally n the case, where vson qualty s poor (for objects smaller than few mcrometers), contact mechancs models are used for 3-D feedback, or path plannng s requred to mplement vrtual gudes. Ths paper valdates the feasblty of usng haptc feedback for 3-D manpulaton. A two-layer pyramdal structure based on four 5 ± 1 μm mcrospheres s selected to llustrate the study snce these objects are commonly consdered at ths scale [22]. Ths paper s based on the analyss of haptc couplng schemes as we dd prevously [23]. The frst results of 2-D teleoperaton of 50 μm mcrospheres wth haptc feedback usng rollng were obtaned n [24]. Ths report deals wth 3-D mcroassembly n ambent condtons of mcrospheres ten tmes smaller. The system uses two ndependent AFM probes to collaboratvely grasp and poston each object, as reported n [25]. Snce ths setup enables 3-D manpulaton of dfferent objects wth force measurement [26], the proposed method could be adapted to dfferent objects provded that the equaton of ther shape s known. Teleoperaton through haptc feedback s extensvely used n every step of the operaton. Snce we have chosen to concentrate on the feasblty of a fully teleoperated 3-D manpulaton, automated manpulaton s not consdered even though good results could be obtaned. The applcatons of these two modes are complementary: automated manpulaton s desgned to perform repeated tasks effcently. Teleoperaton provdes hgh flexblty and enables specfc manpulatons of ndvdual objects, such as placng them on a TEM grd for physcal propertes analyss. As the proposed approach reles on the operator to ncrease the flexblty of the system, the prescan step s avoded. It s replaced by a user-guded ntal exploraton. Ths exploraton allows for onlne calculaton of vrtual gudes, helpng the operator to correctly algn the dual-tp grpper wth respect to the manpulated object, even n the case of poor vsual accuracy. Addtonally, dfferent feedback schemes are presented for pck-and-place of mcrospheres. Compared to exstng teleoperaton systems at the mcro/nanoscale, ths paper presents two man contrbutons. Haptc feedback based on dynamc-mode AFM data s presented for the frst tme. Ths mode s most often used for mcro/nanomanpulaton as the measurement of the ampltude of the oscllatons provdes accurate nformaton about the poston of the object. From these data, the graspng pont s determned based on a least-mean-square algorthm. Next, two dfferent haptc feedback methods are proposed for two-probes-based pck-and-place operatons. The frst one provdes nformaton on the measured nteracton forces drectly, whle the second one asssts the user n mprovng dexterty and avodng collson. These feedbacks are proposed for the same task, but they correspond to dfferent applcatons (comprehenson of physcal phenomena or safe and effcent manpulaton). Ths paper s organzed as follows. The expermental setup and the manpulaton protocol are descrbed n Sectons II and III, respectvely. Haptc feedback based on dynamc-mode measurements and onlne constructon of vrtual gudes to accurately algn the grppers to mcrospheres s dscussed n Secton IV. Secton V explans and analyses n detal the pck-and-place experments, and Secton VI, the constructon of a two-layer pyramd. A table summarzng the notatons used n ths paper can be found n Secton VIII. II. EXPERIMENTAL SETUP Detaled specfcatons of the manpulaton setup are dscussed n [27]. A bref summary s gven here. The mcromanpulaton platform s depcted n Fg. 1. The AFM grpper s equpped wth an optcal mcroscope, and two sets of nanopostonng devces and optcal levers to coordnate two AFM cantlevers wth protrudng tps (namely, Tps I and II, Nanosensors ATEC FM) facng each other, formng a dual-tp grpper. Tp I s fxed on an X Y Z motorzed mcropostonng stage, whle Tp II s actuated by an open-loop X Y Z pezotube mounted

3 BOLOPION et al.: HAPTIC TELEOPERATION FOR 3-D MICROASSEMBLY OF SPHERICAL OBJECTS 3 manpulates the handle and the resultng poston p h s scaled down to be used to control the actuators (nanostage and pezotube). The haptc force F h sent to the user through the haptc nterface s based on measurements from the two photododes (V I and V II ). As represented by the swtches S 1 and S 2 n Fg. 1, dfferent translators and feedbacks are used at each step of the mcroassembly. The next sectons detal the use of the haptc nterface to nteractvely perform a mcroassembly task. Fg. 1. AFM grpper-based telemcromanpulaton system. (Left) Haptc devce provdng a user nterface to control the 3-D mcroassembly wth real-tme haptc feedback. (Rght) Dual-probe grpper comprses two AFM cantlevers wth protrudng tps for pck-and-place mcromanpulaton. TABLE I SPECIFICATIONS OF MOTION STAGES on a X Y Z manual stage. A closed-loop X Y Z nanostage s used to actuate the sample holder durng mcroassembly. Coarse algnment of Tps I and II s acheved under the optcal mcroscope wth large dsplacements of motorzed and manual stages. Detaled specfcatons of each moton stage are summarzed n Table I. Each cantlever dsposes of ts own optcal lever, comprsng a laser source and a four-quadrant photodode. Data acquston occurs at Hz for statc force samplng and at 600 khz for ampltude through a NI 6289 DAQ card. Cantlevers can be used n two dfferent modes: tappng and statc. For the tappng mode, a pezoceramc exctes each probe at ts natural frequency. The ampltude of the resultng oscllatons s measured through the varatons of the voltage output on the photodode A = β ΔV (1) where A s the ampltude measurement, β =10 6 m V 1 s a calbrated converson factor, and ΔV s the dfferental voltage response of the photodode. In statc mode, the normal force appled on the cantlever F s measured drectly from the output voltage of the photodode F = k n S n ΔV (2) where k n =2.8N m 1 s the normal stffness of the cantlever, and S n = m V 1 s the senstvty of the optcal levers. An Omega haptc nterface, manufactured by force dmenson 1 s provded for ntutve user control of the manpulator. Ths master arm s a 3 degrees of freedom devce. The user 1 III. 3-D MICROASSEMBLY PROTOCOL At mcroscales, complex strateges must be used to manpulate objects [28], [29]. Dual-tp manpulaton has certan advantages over other 3-D manpulaton technques. These alternatves are adheson-based pck-and-place or monolthc twofnger grppers. The former requres mechancally complcated strateges for an accurate placng on the substrate [30]. The latter suffers from the lack of force sensng of the grpper [31] and from the ll-controlled grpper/object nteractons, resultng n serous drawbacks n a scenaro requrng precson, such as the one proposed here. On the other hand, n dual-tp manpulaton, the contact area s extremely small, and hence, grpper/object adheson s a mnor ssue. As both probes provdes AFM-grade force sensng, t s possble to montor all nteractons, ncludng contacts, adheson forces, and prmarly, the graspng force. However, these advantages come at the cost of augmented complexty of the overall manpulaton process. Dfferent delcate steps are requred to poston each tp correctly and to coordnate the pck-and-place of the object. An overvew of the complete manpulaton scenaro s depcted n Fg. 2. In order to grasp the object between two tps, each contact pont has to be algned wth the center of the manpulated sphere. Gven the relatve szes of manpulated objects, AFM cantlevers and ther protrudent tps, a vson-based control scheme under an optcal mcroscope does not provde suffcent resoluton and precson. An ntal AFM scan would gve addtonal nformaton for correct postonng, but that s a tme-consumng step and comes at the rsk of dsturbng the manpulaton scene. The approach proposed here s based on a user-drven exploraton of the manpulated object. Haptc feedback allows the operator to feel when he/she touches the object whle freely explorng the manpulaton area. Note that durng ths operaton, the vertcal poston of the probes are constraned to a few mcrometers above the substrate and the operator controls only the horzontal moton. The data recorded durng ths exploraton s processed onlne, whch generates a vrtual gude to pull the user to the optmum contact pont. The user feels and sequentally adjusts the contact force for both tps, ensurng an adequate grp on the object. In the thrd phase of the manpulaton, both grppers are mmoblzed on both sdes of the object and the operator controls the moton of the sample holder, whle stll recevng haptc feedback calculated from the output of the two probes. The choce of the partcle to be manpulated s made usng a top vew optcal mcroscope (from an Olympus BX50WI mcroscope). The coarse postonng of the tps s also performed usng ths vsual feedback. They are moved usng the mcropostonng modules (manual and motorzed stages) descrbed n

4 4 IEEE/ASME TRANSACTIONS ON MECHATRONICS Fg. 2. Steps of dual-tp pck-and-place manpulaton. The manpulaton area and a coarse postonng of the tps are determned usng the optcal mcroscope. The operator then sequentally places each tp on both sdes of the object usng the Omega, controllng, respectvely, the sample holder through the nanostage and Tp II through the pezotube. In both steps, haptc feedback s provded based on the ampltude varatons of each tp n tappng mode. Once the object s held between the two tps, the lftoff and placng on the substrate are acheved by haptc control of the sample holder. Contact-mode measurements are used to provde the force feedback. Fg. 3. Schematc dagram of haptc exploraton by local scan of the lower semmcrosphere usng a oscllatng cantlever: (a) Top vew of desred grasp confguraton. (b) Front vew shows the tp tappng the mcrosphere whle approachng on the x-axs. (c) Sde vew shows the tp tappng the mcrosphere when scannng on the y-axs wth a fxed x-poston. (d) n contact ponts recorded wth random x t -andy t -postons wth matchng ampltudes A t. Table I. The postonng at the correct heght s acheved by automated detecton of the substrate. The user then selects the task to be realzed usng a user nterface. Ths protocol could be smplfed by automated transtons between these steps. In ths case, the completon of the pck-and-place operaton would be determned by the user s decson, to gve hm or her the possblty of pckng up agan the sphere to brng t elsewhere. Tp-algnment phases, ncludng the haptc feedback and vrtual gude generaton and pck-and-place phases wth two dfferent haptc schemes are detaled n the followng. IV. ASSISTED GRIPPER ALIGNMENT The algnment of each tp s a user-drven process. The operator moves the tp whle recevng haptc feedback derved from ampltude measurements of the AFM probe. Durng the ntal exploraton and pror to the generaton of vrtual gudes, the haptc feedback s only on the x-axs [see Fg. 3(a)]. As the operator manually scans the surface of the to-be-manpulated object, the data are recorded to reconstruct ts shape and create the vrtual gude. Ths vrtual gude generates the haptc feedback along the y-axs, pullng the tp to the calculated grasp lne y 0, parallel to the x-axs, and crossng the sphere s center. A. Tappng-Mode Measurements In tappng mode, each probe s excted at ts natural frequency. At constant, z-poston above the substrate and away from objects, ths results n oscllatons at constant ampltude, noted as A 0. Whle approachng an object, startng from a few hundreds of nanometers, the tp contacts the object ntermttently and the ampltude A t decreases untl a mnmum value A CP s reached at full contact between the tp and the object. Fg. 3 llustrates the prncple of object detecton from ampltude varatons. The tp s frst set to a gven z-poston above the substrate h 0. Ths step s acheved n an ntal phase and the user controls the moton only n the (x, y) plane parallel to the substrate. Whle the tp moves on the grasp drecton of the grpper, the x-axs, ampltude decreases untl contact [see Fg. 3(b)]. On the y-axs, perpendcular to the grasp drecton, both tps must be algned wth the center of the sphere. Ths matches the mnmum of ampltude along the y-axs, at a fxed x-poston [see Fg. 3(c)]. B. Haptc Feedback for Tp Algnment To algn the grpper wth the sphere and brng t to contact the vsual feedback from the optcal mcroscope does not provde suffcent resoluton. Haptc force ams to compensate for ths lack of vsual feedback. The haptc couplng used s depcted n Fg Each tp s sequentally algned on the grasp lne and brought to contact. 1) x-axs: The haptc feedback along the x-axs should provde the followng nformaton. 2 More nformaton about ths control scheme can be found n [23]. Ths remark also apples for the couplng depcted n Fg. 10.

5 BOLOPION et al.: HAPTIC TELEOPERATION FOR 3-D MICROASSEMBLY OF SPHERICAL OBJECTS 5 Fg. 4. Haptc couplng for dual-tp grpper algnment. The user manpulates the actuators by settng the poston of the haptc devce. Haptc feedback s derved from ampltude measurements. Dependng on the consdered tp, the swtches S 1 and S 2 enable the user to manpulate the nanostage or the pezotube, and accordngly receve the ampltude measurement from Photodode I or II. 1) R 1 : Force null when the tp s far from the object. 2) R 2 : Increasng force as the tp approaches the object. 3) R 3 : Increasng force as the tp apples a force on the object. Accordng to the varaton of ampltude descrbed prevously, the followng haptc feedback F hx s proposed and satsfes requrements R 1 R 3 : { αa (A A 0 ), f A>A CP F hx = (3) α a (A A 0 )+k x (x x CP ), else where α a s a scalng factor. The ampltude A 0 s measured at the begnnng of the experment, whle the tp oscllates at ts natural frequency. Two cases are dstngushed. 1) Before contact (frst equaton): As the ampltude s decreasng whle the tp approaches the object, an ncreasng repulsve force s sent to the user so that he or she s aware of the presence of the object. 2) Whle n contact (second equaton): Asprngk x between the poston of the contact pont x CP and the current poston of the tp x s added to the feedback of the frst equaton. It smulates the force appled by the tp to the sphere. The contact pont locaton x CP s set when the ampltude measurement reaches A CP at full contact. A CP s an arbtrary threshold, set accordng to the conclusons gven n [25]. 2) y-axs: The force perceved along the y-axs must enable the user to algn the tps wth respect to the sphere on the grasp lne. The haptc feedback along the y-axs s not avalable before all the ponts have been recorded and the computaton of the vrtual gude by (7) s acheved. Durng ths exploraton n search of the y 0 -poston, the x-axs haptc feedback s provded to the user so that he or she perceves the sphere s locaton. When y 0 s computed, a haptc feedback F hy smulatng a sprng k y between y 0 and the current poston y of the tp s sent to the user F hy = k y (y y 0 ). (4) C. Vrtual Gude Generaton Durng the ntal exploraton n tappng mode, n contact ponts (x t,yt), wth ther matchng ampltudes A t ( = 1,...,n) are collected. The contact poston data are acqured only f the actual ampltude A t s n the [15%A 0, 70%A 0 ] nterval to avod false postves. In order to defne the z t-coordnate for each contact pont (x t,y t), the approxmaton z t = A t/2 s proposed. Ths s a relatve poston snce the cantlever s oscllatng around the z-poston h 0, set manually. The calculaton of the z 0 -coordnate of the sphere s thus relatve to h 0, and s not accurately known. However, as the only parameter useful for haptc feedback s the y 0 -coordnate, ths approxmaton s acceptable. Fg. 3(d) represents the n ponts (t 1,...,t n ) recorded durng the exploraton process. These ponts are used to reconstruct the shape of the manpulated sphere, calculate the grasp lne, and provde the haptc feedback along the y-axs. Wth pror knowledge of the shape of the object, and the n recorded ponts, the sphere can be reconstructed from the surface equaton (x x 0 ) 2 +(y y 0 ) 2 +(z z 0 ) 2 = R 2 (5) where R s the radus of the sphere, and x 0, y 0, and z 0 are the coordnates of ts center. Ths can be wrtten as follows: x 2 + y 2 + z 2 C a x C b y C c z + C d =0 (6) where C a =2x 0, C b =2y 0, C c =2z 0, and C d = x y0 2 + z0 2 R 2. Fndng the coeffcents C a, C b, C c, and C d allows us to defne the coordnates at the center of the mcrosphere and ts radus. To do so, the equaton whch best fts the n recorded ponts s determned usng a least-mean-square algorthm. The soluton of the followng system of equatons gves the coordnates of the sphere s center: x 1 t yt 1 z 1 t 1 C a (x 1 t ) 2 +(yt 1 ) 2 +(zt 1 ) 2 x 2 t yt 2 zt 2 1 C b C c = (x 2 t ) 2 +(yt 2 ) 2 +(zt 2 ) x n t yt n zt n 1 C d (x n t ) 2 +(yt n ) 2 +(zt n ) 2 (7) The coeffcents C a, C b, C c, and C d can be deduced from: [ C a C b C c C d ] T = M 1 Y, where T s for the matrx transposton and (x t ) 2 x t yt x t zt x t x t yt (y t ) 2 y t zt y t M = x t zt y t zt (z t ) 2 zt x t yt zt n x t [(x t) 2 +(yt) 2 +(zt) 2 ] y t [(x t) 2 +(yt) 2 +(zt) 2 ] Y =. z t [(x t) 2 +(yt) 2 +(zt) 2 ] [(x t) 2 +(yt) 2 +(zt) 2 ] The poston of the graspng pont along the y-axs s then computed as y 0 = C b /2. Note that as all the ponts are on the same sde of the sphere along the x-axs (x <0 for Tp I and nversely for Tp II), the calculated x 0 -coordnate may be naccurate.

6 6 IEEE/ASME TRANSACTIONS ON MECHATRONICS Fg. 5. (a) Optcal mage of a mcrosphere under 100 optcal magnfcaton (the scale bar represents 5 μm). (b) FM mage scan on the mcrosphere. However, as stated earler, the only parameter used for vrtual gude s y 0. The generaton of ths vrtual gude depends on the number of ponts n and ther respectve postons. An emprcal analyss to defne a mnmum value for n and the dstrbuton of the ponts recorded s presented n the followng along wth expermental analyss. D. Expermental Valdaton of Tp Algnment Manpulated objects are mcrospheres, wth a dameter of 4 6 μm. An mage of a sphere from an optcal mcroscope, and one from an AFM scan are presented n Fg. 5(a) and (b). Before algnng the tps wth the spheres, they are frst postoned manually at the correct vertcal poston (around nm) above the substrate. Bolopon et al. [23] propose a haptc feedback soluton for ths step. Each tp s then sequentally postoned by the operator at each sde of the object. For Tp I, the user actually controls the nanostage transportng the sample holder. As the nanostage ncludes a closed-loop poston controller, the Omega supples drectly set-pont values for ts moton. For the algnment of Tp II, the pezotube actuator s used. As ths actuator s n open loop, t does not provde accurate postonng. However, couplng the pezotube wth a haptc nterface s equvalent to a closed-loop force-feedback scheme wth the operator as the controller. All expermental data presented n the followng are acqured usng the nanostage and provdes accurate poston nformaton. The pezotube, although lackng precse poston measurements, gves qualtatvely smlar results. 1) x-axs Haptc Feedback: Expermental results acqured whle movng the tp along the x-axs and contactng the mcrosphere are depcted n Fg. 6. The poston of the tp s represented n Fg. 6(a), the ampltude measurement s depcted n Fg. 6(b), and the haptc feedback n Fg. 6(c). In area 1, the tp s away from the sphere and the feedback s null. In area 2, the user dstnctly perceves the haptc feedback as the tp approaches the sphere and ntermttent contact starts. An addtonal feedback s transmtted when an effort s appled by the cantlever on the sphere n area 3, ncreasng the sensaton of stffness. Compared to usng drect force measurement from a cantlever n statc mode, the tappng-mode ampltude measurement provdes a better senstvty on the x-axs. In statc mode, as the measurement drecton s almost algned wth the probes length, the equvalent stffness s extremely hgh compared to k n on the z-axs. Hence, a statc detecton on the x-axs would only occur Fg. 6. Haptc feedback along the x-axs whle explorng the to-bemanpulated sphere. (a) Tp poston. (b) Measured tp oscllaton ampltude. (c) Force sent to the user. For the haptc feedback, the coeffcents are set to: α a = N m 1, k x = N m 1,anda CP =0.1μm. TABLE II PARAMETERS USED FOR VIRTUAL GUIDE GENERATION when a qute mportant force s already appled on the object. In contrast, the use of the tappng mode allows earler detecton of the object, as only ntermttent contact wth the object produces a detectable sgnal. Ths allows users to be aware of the object s presence and prevents ther nvoluntarly pushng. 2) Generaton of the Vrtual Gude: As stated earler, the prelmnary exploraton provdes the data ponts to generate the vrtual gude and the assocated y-axs feedback. The nfluence of two parameters the number of ponts n used n (7), and the mnmum dstance between two ponts (noted d) s expermentally explored n order to optmze the vrtual gude. Table II summarzes the dfferent trals. Each experment (for a gven (n, d) couple) s repeated fve tmes. Fg. 7 compares a reference AFM scan and y 0 obtaned by shape reconstructon, for the consdered (n, d) couples. The standard devaton s dsplayed n Fg. 7(c). As all these results are user dependent snce users are free to choose any trajectory for the ntal exploraton they should be treated only qualtatvely. The frst observaton s that ncreasng the number of ponts n has lttle mpact on the accuracy of the vrtual gude f these

7 BOLOPION et al.: HAPTIC TELEOPERATION FOR 3-D MICROASSEMBLY OF SPHERICAL OBJECTS 7 Fg. 7. Results of the vrtual gude generaton. (a) Reference scan at several x-postons from the sphere (the dash lne represents the grasp lne y 0 ). (b) Estmated locaton of the grasp lne. (c) Standard devaton. (d) Tme needed to complete the localzaton of y 0. Fg. 9. Haptc feedback along the y-axs whle algnng the tp wth the sphere: poston of the tp and force computed wth k y = N m 1. Fg. 10. Haptc couplng for pck-and-place. The user manpulates the nanostage wth respect to the two tps by settng the poston of the haptc devce. Haptc feedback s derved from force measurement from the sum of the two photododes outputs. Fg. 8. Sphere reconstructed based on n =12and d =0.4μm. The red dots represent the ponts acqured durng the exploraton step. ponts are close to each other. Moreover, f n s set too hgh, more ponts on the edges of the semsphere are requred. As the poston data at these locatons are less accurate, the standard devaton s hgher. On the other hand, settng a mnmum value for d forces data ponts to be more evenly dstrbuted on the surface and leads to a better estmaton (see Fg. 8) as supported by the decrease of the standard devaton. Fg. 7 also shows that except for small d values, ncreasng the number of ponts hghly ncreases the tme cost of the vrtual gude generaton. Snce only the hemsphere facng the tp s accessble and the tp s constraned n the vertcal drecton, settng d to a mnmum value lmts the maxmum number of ponts that can be acqured. Choosng d =0.3 μm and n =12s a good tradeoff between precson of the results and tme cost of the gude generaton. These values are selected for the followng manpulatons. 3) y-axs Haptc Feedback: The y-axs feedback s effectve as soon as the vrtual gude s generated. Its value s calculated usng (4). Fg. 9 represents the force perceved by the user. The poston of the tp as well as y 0 are also gven. The haptc feedback on the y-axs helps the user to algn the tp wth respect to the sphere, as y 0 s at the equlbrum pont of the vrtual sprng. Precse postonng s acheved snce contact nformaton s transmtted to the user through the x-axs of the haptc devce F hx, and algnment s ensured due to the haptc force F hy. V. PICK-AND-PLACE WITH HAPTIC FEEDBACK After the object s grasped between the two tps, t wll be lfted from the substrate, transported to the target locaton, and placed on the substrate. Two dfferent haptc feedbacks are proposed for ths task. The frst one renders to the user drectly the forces measured by both probes, wth proper scalng. The second one calculates a vrtual gude usng these measurements to assst the user to lftoff the object to a vertcal poston set suffcently hgh to avod any contact wth other objects or the substrate. It also ensures that the placng s voluntary. In both cases, as depcted n Fg. 10, the Omega haptc devce s used for poston control of the sample holder through the nanostage, whle the two tps holdng the mcrosphere are mmoblzed. The force data are obtaned n statc mode, from the deflecton of each probe measured drectly on photododes usng (2).

8 8 IEEE/ASME TRANSACTIONS ON MECHATRONICS Haptc feedback s rendered along the vertcal ascendng z- axs. Hence, a postve value results n a force pushng the haptc handle upward, away from the substrate; t s hence called repulsve, whle a negatve value pulls the handle downward toward the substrate ( attractve ). The manpulaton s carred out n ambent condtons, at 20 C, and relatve humdty of 48%. A. Haptc Feedback of Nanoscale Interactons Ths frst haptc feedback returns to the user the nanoscale nteractons of the pck-and-place operaton as fathfully as possble. It s syntheszed from force responses of Tps I and II. As detaled n [25], adhesve forces F ao between the sphere and the substrate can be estmated as follows: F ao = F I + F II (8) where F I (respectvely, F II ) s the force appled to Cantlever I (respectvely, II). Hence, the haptc force rendered to the user s computed as follows: F fe hz = α f (F ao F 0 ) = α f [(F I F I0 )+(F II F II0 )] (9) where F 0 = F I0 + F II0 s the force measured when the tps are holdng the sphere before lftoff and t s naturally proportonal to the graspng force appled by the tps to the object. Removng ths offset allows the user to dscard the graspng force, whch s not useful for pck-and-place. Moreover, n the case, where the grasped object s lost hazardously durng the lftoff, the measured forces F I and F II wll fall back to zero, and (9) wll gve a negatve value, pullng back the probes to the substrate. A force amplfcaton factor α f s used to scale the measured forces and the haptc force sent to the user. The nomnal value used here s α f = Ths coeffcent s set consderng the magntude of nanoscale nteractons that should be felt by the user (n partcular the pull-off force). Detaled dscusson on the defnton of ths parameter can be found n [23]. Fg. 11 represents the haptc feedback durng a pck-andplace operaton of a 5 μm sphere from a glass substrate and the nsert depcts forces measured from probes. The curve s startng pont s the contact state between the mcrosphere and the substrate. As the nanostage moves down (hence the object held by the tps s lfted), probes are bent down measurng negatve forces (nset ). Durng the pckup, when the nanostage poston reaches around 900 nm, the mcrosphere pulls off the substrate wth a mnmum force of 1125 nm overcomng the adheson. Note that after the pulloff, the measured force falls to 550 nm, and not to the prepck-up null value (nset ). Actually, as the tp/object contact ponts are n the lower hemsphere, durng the lftoff, the object sldes slghtly down, ncreasng the graspng force. 3 Durng the transport phase, a 3 Both pck-up and unntentonal loss of a sphere yeld a negatve force feedback untl the pull off (ether the pull off of the sphere from the substrate, or the tps from the sphere). After the pull off, n case of pckng-up the sphere, the force feedback remans negatve due to the graspng force. In case of an unntentonal loss of the sphere, the cantlevers wll go back to ther neutral poston,.e., wth no bendng. In ths case, the haptc feedback sent to the user Fg. 11. Measured normal force responses from both mcrocantlevers durng the pck-and-place manpulaton of a mcrosphere. () Pck-up occurs. () Mcrosphere s detached from substrate after the pull off. () Mcrosphere snaps nto the substrate. (v) Grpper/mcrosphere pulls off. (v) Grpper snaps nto the substrate. (v) Manpulaton ends wth slght bendng of the mcrocantlevers. change on the force can be noted. Ths s agan due to the sldng of the sphere n the grpper. Ths hypothess s backed up by the approxmate 0.2 μm, dfference seen between pckup and touch down postons along the z-axs. Durng the placng operaton, the mcrosphere snaps-n the substrate (nset ). As the object s pushed to the substrate, between () and (v), the contact force compensates the graspng force, untl the tps pull off the sphere (nset v) and slde down from the object to the substrate (nset v). At ths pont, t s suffcent to move apart both tps along the y-drecton to release the sphere from the grpper and acheve the operaton. Note that snce the contact area at object/tp nterface s much smaller than at the object/substrate nterface due to the sharp tps used, the problem of the object adherng to a probe s lmted. B. Haptc Feedback Provdng Assstance The haptc scheme presented earler allows the user to feel the nanoscale nteractons, especally the adheson and the wellknown pull-off phenomena. However, t s arguable f ths feedback has a postve effect on manpulaton dexterty. It may be more nterestng to conceal these effects of whch the operator s unfamlar wth and to replace the haptc feedback wth a vrtual gude. An assstve haptc feedback, wth a postve effect on the dexterty, should fulfll followng requrements. 1) The user does not have to use great effort to lft the sphere off. 2) The sphere should be kept at a gven z-poston durng transport to avod contact. 3) The sphere should be placed voluntarly only, not because the operator touches the substrate accdentally. 4) Whle placng the sphere, the effort appled should be as lttle as possble to enable an easy release of the tps. wll become postve, snce the zero haptc feedback corresponds to the ntal graspng force appled on the sphere. The user wll thus be able to dstngush the pck-up and the unntentonal loss of the sphere.

9 BOLOPION et al.: HAPTIC TELEOPERATION FOR 3-D MICROASSEMBLY OF SPHERICAL OBJECTS 9 TABLE III ASSISTANCE BASED ON HAPTIC FEEDBACK FOR PICK-AND-PLACE OPERATION Fg. 12. Haptc feedback assstance for pck-and-place operaton. (A) Pckand-place operaton begns, the sphere s on the substrate. (B) and (C) Sphere pulls off. (D) Sphere s mantaned above the substrate at a gven vertcal poston. (E) Sphere s moved toward the substrate. (F) Sphere s placed down on the substrate. Suffcent haptc nformaton should be provded so the user can effectvely feel that the placng has been acheved. The proposed approach s based on the use of the opposte of the measured nteracton force as haptc feedback. As such, for example, the pull-off force wll result n a postve force on the haptc handle, actually pushng the held object away from the substrate. The expected percepton s comparable to releasng a pressed keyboard button. As prevously, the haptc feedback s computed from force measurement from both photododes. To fulfll the requrements stated earler, the output of the photodode s converted nto the haptc force F as hz : { Fhz as αf (F = ao F 0 ), f z<z PO α f (F ao F 0 )+k z (z z PO ), else (10) where α f s a force amplfcaton factor, F s the force measured from the photodode, F 0 s the graspng force as earler, z s the poston on the vertcal axs, and z PO s the vertcal poston of nanostage correspondng to the end of pull-off phenomena. Its value s detected onlne durng the manpulaton from the sudden drop n force measurement; k z s the stffness of the vrtual sprng of the haptc gude, whch wll effectvely pull the object above z PO and keep t at a constant vertcal poston. The result of ths feedback scheme s depcted n Fg. 12 for a pck-and-place operaton on the substrate. A repulsve force, proportonal to the opposte of the measured adhesve forces, asssts the user to lft the object durng the pull off. Immedately after the pull-off phenomena, the measured force does not fall back to ts ntal value, due to the sldng of the sphere between the tps as explaned earler. Thus, a resdual porton of the repulsve force remans at z z PO. As the sprng k z s actvated at z PO, ths force s counterbalanced by the vrtual gude at z D, above z PO. Then, users can freely move the sphere above the substrate n the horzontal plane, whle the sprng analogy of the vrtual gude ensures that they keep a relatvely constant Fg. 13. Task plannng of the mcropyramd assembled by four mcrospheres. The rght-sde dagram shows the assembly protocol, n whch mcrospheres 1, 2, and 4 are assembled by pck-and-place manpulaton, whle mcrosphere 3 s pushed to ts target poston. poston on the z-axs. To place the sphere on the substrate, the user has to counteract the sprng k z between z D and z PO, and the resdual repulsve force below z PO. Ths condton ensures that the object s not be placed by mstake. When the force measurement falls below F 0, the vertcal moton s automatcally stopped, so that no addtonal force s appled as the sphere reaches the substrate. Ths facltates the releasng of the tps and protects fragle objects and the grpper. As ths results n the haptc force becomng null, the user clearly dscerns the placng. For haptc feedback, the parameters are set to: α f = , k z = N m 1. Forces hgher than 3 N are truncated.

10 10 IEEE/ASME TRANSACTIONS ON MECHATRONICS Fg. 14. Teleoperated 3-D mcroassembly demonstraton of a mcropyramd: (a) (d) Four photos ntercepted from the assembly process of the frst layer ofthe mcropyramd. (e) and (f) Assembly process of the second layer (the fourth mcrosphere) of the mcropyramd. The top vew photos (a) (f) are captured under magnfcaton of 20. (g) Mcroassembly result magnfed 100. TABLE IV COMPARISON OF HAPTIC FEEDBACK METHODS FOR PICK-AND-PLACE OPERATIONS the manpulaton operaton. The choce of the method wll thus depend on user needs and the specfctes of the task. The percepton of ths haptc feedback s equvalent to pushng a keyboard button: the object s kept at the vertcal poston z PO durng the transfer and the user has to voluntarly push t back to the substrate to place t. Table III detals each step of the pck-and-place operaton usng ths vrtual gude wth phase transtons and user percepton. Usng ths scheme, the pck-and-place operaton s made easer and safer as the haptc feedback helps the user to perform the gven steps correctly. C. Comparson of Two Haptc Feedback Renderng Two haptc feedback methods are presented n ths paper for pck-and-place operaton. However, the resultng feedback n the users s dfferent (see Table IV). Haptc feedback of nanoscale nteractons ams at transmttng physcal phenomena. The user drectly feels the adhesve forces, the pull-off phenomena, and the contact wth the substrate. It mproves the understandng of these nteractons. The haptc feedback for assstance s not desgned to be related to the physcal forces but to facltate VI. CONSTRUCTION OF A TWO-LAYER PYRAMID In order to valdate 3-D manpulaton capabltes of the haptc system, a mcrostructure s bult. The example of a two-layer pyramd composed of four nylon mcrospheres wth dameter of 4 6 μm s chosen [22]. Mcrospheres were deposted on a freshly cleaned glass substrate. An area of nterest for the experments was selected under an optcal mcroscope wth 20 optcal magnfcaton. Fg. 13 shows the selected area and the nset shows the assembly sequence. After a coarse postonng under the optcal mcroscope, the total manpulaton, whch ncludes the algnment of the grpper and the pck-and-place operaton, takes less than fve mnutes per sphere. Snce ths manpulaton s based on the operator, ths tme manly depends on users sklls. It s gven only as an ndcaton here. The 3-D mcropyramd s bult usng pck-and-place manpulaton. Fg. 14 shows the 3-D mcromanpulaton process. Fg. 14(a) and (b) s captured when the frst sphere s pcked and placed. Fg. 14(c) shows the second sphere manpulated and released at ts fnal poston. The frst layer s fnshed after the thrd mcrosphere has been manpulated by pushng [see Fg. 14(d)]. Fg. 14(e) and (f) descrbes the transport of the last mcrosphere, whch completes the assembly. The ultmate result s shown n Fg. 14(f), and wth a magnfcaton of 100 n Fg. 14(g). Fg. 15 shows the haptc feedback of nanoscale nteractons durng the manpulaton of the fourth sphere. As ths sphere s put back above the others, the placng occurs 4 μm above the ntal poston. In the case, where assstve feedback s used for manpulaton of the fourth sphere, two ponts are consdered. 1) The gven vertcal poston at whch the sphere s mantaned wth zero-force feedback must be hgher than the frst layer of the spheres. Ths s adjusted by changng the

11 BOLOPION et al.: HAPTIC TELEOPERATION FOR 3-D MICROASSEMBLY OF SPHERICAL OBJECTS 11 Fg. 15. Haptc feedback durng the mcroassembly operaton of the fourth mcrosphere. () Pck-up occurs. () Mcrosphere s detached from the substrate. () Mcrosphere snaps nto the frst layer of mcrospheres. (v) Grpper/sphere pulls off. (v) Grppers snap nto the frst layer of mcrospheres. value of the stffness k z. To ncrease the z-poston above the substrate, ths stffness should be decreased. 2) Whle placng the sphere, the z-poston wll not become lower than z PO. The force feedback wll not decrease before placng the sphere. When the measured force wll be such that F F 0 (the sphere s deposted onto the pyramd), the vertcal moton of the sphere wll be stopped and the force feedback wll become null. In these crcumstances, assstve haptc feedback produces results smlar to the ones presented n Fg. 12. Moreover, f the heght at whch the sphere should be placed s known, (10) can be adjusted by tunng k z so that haptc feedback s dentcal to the one transmtted when the sphere s placed down on the substrate. VII. CONCLUSION Haptc feedback s used to perform a 3-D mcroassembly operaton usng AFM. Although a complex tool composed of two cantlevers as a dual-tp grpper has to be used to perform accurate graspng and pck-and-place at ths scale, users unfamlar wth the setup were able to carry out the task, as haptc feedback ensures a hgh ntutveness of the setup. To sequentally place the two tps n the grasp lne and n contact wth the object, both cantlevers are used n dynamc mode. Ths approach enables suffcent feedback to detect the sphere even f the measurement drecton s almost algned wth the probe. Data acqured durng ths exploraton are processed onlne to compute vrtual gudes, whch pull the user to the grasp lne. For the pck-and-place step, haptc feedback allows us to ether fathfully render mcroscale nteractons or provde assstance to the operator for mproved dexterty and collson avodance. These feedbacks are based on force measurements from both cantlevers n statc mode. A two-layer pyramd s bult from four 5 ± 1 μm spheres n ambent condtons for expermental valdaton of the usablty of the overall setup. Ths paper concentrates on the feasblty of a teleoperated system wth haptc feedback for 3-D AFM-based manpulaton. To desgn an effcent setup, t s necessary to clearly defne whch tasks could be automated, and whch ones requre the hgh flexblty of teleoperaton, dependng on the applcaton. User-based tests should then be carred out to measure the performance of the proposed systems. A user evaluaton wll also compare the pck-and-place haptc feedback schemes n terms of effcency and percepton. In ths paper, the example of a pyramd s chosen to llustrate the analyss. The methodology presented could be extended to other type of objects, such as carbon nanotubes or nanowres. In partcular, the vrtual gude along the y-axs could be adapted to nonsphercal objects by takng the object geometry nto account. The system presented s a frst step toward haptc feedback nanomanpulaton. Pror to usng t, though, a real-tme vrtual realty scene should be generated and dsplayed n 3-D, along wth vrtual gudes, to compensate for the complete lack of vsual feedback. REFERENCES [1] R. Fearng, Survey of stckng effects for mcro parts handlng, n Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., 1995, vol. 2, pp [2] A. Mencass, A. Esnberg, I. Izzo, and P. Daro, From macro to mcro manpulaton: Models and experments, IEEE/ASME Trans. Mechatroncs, vol. 9, no. 2, pp , Jun [3] I. Bukusoglu, C. Basdogan, A. Kraz, and A. Kurt, Haptc manpulaton of mcrospheres usng optcal tweezers under the gudance of artfcal force felds, Presence: Teleoperators Vrtual Envron., vol. 17, no. 4, pp , [4] M. Guthold, M. Falvo, W. Matthews, S. Paulson, S. Washburn, D. Ere, R. Superfne, F. Brooks Jr., and R. Taylor II, Controlled manpulaton of molecular samples wth the nanomanpulator, IEEE/ASME Trans. Mechatroncs, vol. 5, no. 2, pp , Jun [5] A. Ferrera and C. Mavrods, Vrtual realty and haptcs for nanorobotcs, IEEE Robot. Autom. Mag., vol. 13, no. 3, pp , Sep [6] M. Boukhnfer and A. Ferrera, H loop shapng blateral controller for a two-fngered tele-mcromanpulaton system, IEEE Trans. Control Syst. Technol., vol. 15, no. 5, pp , Sep [7] M. Boukhnfer and A. Ferrera, Wave-based passve control for transparent mcro-teleoperaton system, Robot. Auton. Syst., vol. 54, no. 7, pp , [8] R. Anderson and M. Spong, Blateral control of teleoperators wth tme delay, IEEE Trans. Autom. Control, vol. 34, no. 5, pp , May [9] B. Hannaford and J.-H. Ryu, Tme-doman passvty control of haptc nterfaces, IEEE Trans. Robot. Autom., vol. 18, no. 1, pp. 1 10, Feb [10] S.-G. Km and M. Stt, Task-based and stable telenanomanpulaton n a nanoscale vrtual envronment, IEEE Trans. Autom. Sc. Eng., vol.3, no. 3, pp , Jul [11] C. D. Onal and M. Stt, A scaled blateral control system for expermental one-dmensonal teleoperated nanomanpulaton, Int. J. Robot. Res., vol. 28, no. 4, pp , [12] R. Holls, S. Salcudean, and D. Abraham, Toward a tele-nanorobotc manpulaton system wth atomc scale force feedback and moton resoluton, n Proc. IEEE Conf. Mcro Electro Mechancal Syst., 1990, pp [13] S. Marlere, D. Urma, J. Florens, and F. March, Mult-sensoral nteracton wth a nano-scale phenomenon: The force curve, n Proc. Eurohaptcs, 2004, pp [14] G. L, N. X, H. Chen, P. Crag, and P. Mathew, Vdeolzed atomc force mcroscopy for nteractve nanomanpulaton and nanoassembly, IEEE Trans. Nanotechnol., vol. 4, no. 5, pp , Sep [15] M. Guthold, M. Falvo, W. Matthews, S. Paulson, J. Mulln, S. Lord, D. Ere, S. Washburn, R. Superfne, F. B. Jr., and R. T. II, Investgaton

12 12 IEEE/ASME TRANSACTIONS ON MECHATRONICS and modfcaton of molecular structures wth the nanomanpulator, J. Mol. Graph. Model., vol. 17, pp , [16] M. Stt and H. Hashmoto, Teleoperated touch feedback from the surfaces at the nanoscale: Modelng and experments, IEEE/ASME Trans. Mechatroncs, vol. 8, no. 2, pp , Jun [17] G. L, N. X, M. Yu, and W.-K. Fung, Development of augmented realty system for AFM-based nanomanpulaton, IEEE/ASME Trans. Mechatroncs, vol. 9, no. 2, pp , Jun [18] W. Vogl, B. Ma, and M. Stt, Augmented realty user nterface for an atomc force mcroscope-based nanorobotc system, IEEE Trans. Nanotechnol., vol. 5, no. 4, pp , Jul [19] C. D. Onal and M. Stt, Teleoperated 3-D force feedback from the nanoscale wth an atomc force mcroscope, IEEE Trans. Nanotechnol., vol. 9, no. 1, pp , Jan [20] M. Amm and A. Ferrera, Robotc asssted mcromanpulaton system usng vrtual fxtures and metaphors, n Proc. IEEE Int. Conf. Robot. Automat., 2007, pp [21] M. Stt, Survey of nanomanpulaton systems, n Proc. IEEE Conf. Nanotechnol., 2001, pp [22] S. Sato, H. T. Myazak, and T. Sato, Mcro-object pck and place operaton under SEM based on mcro-physcs, J. Robot. Mechatroncs, vol. 14, no. 3, pp , [23] A. Bolopon,B.Cagneau, S. Halyo, and S. Régner, Analyss of stablty and transparency for nanoscale force feedback n blateral couplng, J. Mcro Nano Mechatroncs, no. 4, pp , [24] A. Bolopon, B. Cagneau, and S. Régner, 2D mcro teleoperaton wth force feedback, n Proc. IEEE Int. Conf. Intell. Robots Syst., 2009, pp [25] H. Xe and S. Régner, Three-dmensonal automated mcromanpulaton usng a nanotp grpper wth mult-feedback, J. Mcromech. Mcroeng., vol. 19, no. 7, p (9pp), [26] H. Xe, S. Halyo, and S. Régner, A versatle atomc force mcroscope for 3D nanomanpulaton and nanoassembly, Nanotechnology, vol. 20, p (9pp), [27] H. Xe and S. Régner, Development of a flexble robotc system for multscale applcatons of mcro/nanoscale manpulaton and assembly, IEEE/ASME Trans. Mechatroncs, 2010, DOI: /TMECH [28] M. Sava and H. Kovo, Contact mcromanpulaton - Survey of strateges, IEEE/ASME Trans. Mechatroncs, vol.14,no.4,pp ,Aug [29] M. Sausse-Lhernould, S. Régner, and P. Lambert, Electrostatc forces n mcromanpulatons: Revew of analytcal models and smulatons ncludng roughness, Appl. Surface Sc., vol. 253, pp , [30] S. Halyo, F. Donnet, and S. Régner, Controlled rollng of mcro objects for autonomous mcro manpulaton, Int. J. Mcromechatron., vol.3, no. 2, pp , [31] D. Herban and G. Gauther, Robotc mcro-assembly of mcroparts usng a pezogrpper, n Proc. IEEE/RSJ Int. Conf. Intell. Robots Syst., 2008, pp Aude Bolopon receved the Ph.D. degree n robotcs from the Unversty of Perre and Mare Cure, Pars, France, n She s currently a temporary Assstant Professor at the Insttute of Intellgent Systems and Robotcs (ISIR), Unversty of Perre and Mare Cure. She has also been a member of the ISIR mcromanpulaton team snce Her research nterests nclude teleoperaton and haptc feedback at the nanoscale for mcro and nanomanpulaton. Hu Xe (S 05 M 07) receved the B.S. degree n mechancal engneerng from Harbn Unversty of Scence and Technology, Harbn, Chna, n 2000, and the M.S. degree n mechancal engneerng and the Ph.D. degree n mechatroncs engneerng from Harbn Insttute of Technology, Harbn, n 2002 and 2006, respectvely. In 2003, he became a Research Assstant n the State Key Laboratory of Robotcs and System, Harbn Insttute of Technology, where he became a Lecturer n Snce December 2008, he has been a Research Assocate at the Insttut des Systèmes Intellgents et Robotque, Unversté Perre et Mare Cure/Centre Natonal de la Recherche Scentfque, Pars, France, where he became a Postdoctoral Researcher n December Hs current research nterests nclude mcrorobotcs/nanorobotcs, mcroscopc vson and automaton on atomc-force-mcroscope-based nanomanpulaton, parallel nanomanpulaton force mcroscopy, and 3-D nanomanpulaton force mcroscopy. Dr. Xe was the recpent of the Best Applcaton Paper Award at the IEEE/RSJ Internatonal Conference on Intellgent Robots and Systems n 2009 and also the Tosho Fukuda Best Paper Award n Mechatroncs Fnalsts at the IEEE Internatonal Conference on Mechatroncs and Automaton n Dogan Snan Halyo was born n Istanbul, Turkey, n He receved the Ph.D. degree n robotcs from the Unversty of Perre and Mare Cure, Pars, France, n Snce 2005, he has been an Assocate Professor at the Insttute of Intellgent Systems and Robotcs, Unversty of Perre and Mare Cure. Hs current research nterests nclude mcro and nanomanpulaton and related topcs such as mcroscale phenomena, remote handlng, and bology-drven self-assembly. Stéphane Régner receved the Ph.D. degree n mechancs and robotcs from the Unversty of Perre and Mare Cure, Pars, France, n He s currently a Professor at the Insttute of Intellgent Systems and Robotcs (ISIR), Unversty of Perre and Mare Cure. He has been Head of the ISIR mcromanpulaton team snce Hs research nterests nclude mcro and nanomanpulaton, teleoperaton and haptc feedback at the nanoscale, mcromechatroncs, and bologcal cell characterzaton.