Home

HAPTIC SURGICAL AID SYSTEM WITH MAGNETORHEOLOGICAL

1. 3 Passive Haptic Interaction 5 Experiment setup for rotary MR brake ooonooocccnnnnoccccnonococnonononancnnnnonanccnnnanancnnnnnnass 6 Spherical MIR 2 uu bane beput ote be POR M 7 Braking torque in rotary MR brakes aaa aa aaa eee enen eee eee vene 15 Serpentme figenete TUX PAL EE 16 View of serpentine flux path using FEM analysis 18 Perro Hue sesini EEN 19 MR spherical Drake Tlux pall sionist 21 Calculating cross sectional area for the forward and return paths 23 Magnetic flux density at different azimuth angles 24 Calculating torque about the x and Z axes 25 Force icedback JOY SUCK ene n 27 Optical triangulation system using IR sensors for position measurement 28 Position and orientation of the optical sensors on the joystick handle 29 A A e noes E EEE iep ROE 32 Passive APLICO IMA dE R A A AN 34 Mass Bala nCIN On 35 Work volume of the haptic EEN 38 Ge EEN 39 Haptic arm with the PC Interface and Haptic Rendering 40 ix 6 1 6 2 6 3 6 4 6 5 6 6 6 7 6 8 6 9 6 10 6 11 6 12 6 13 6 14 6 15 6 16 6 17 6 18 Experimental setup for rotary MR 44 Braking torque of rotary MR brake versus current
2. 45 Simulation of collision with a virtual without using the torque sensor 47 Simulation of collision with a virtual with the torque sensor 48 MR brake as damper in haptics tlc e asa Eege 49 MR brake representing Coulomb friction a s 50 Transient response and time constant of the MR brake 51 Hysteretic braking torque of spherical MR brake versus current 55 Simulation of collision with a virtual wall 58 Transient response of spherical MR brake 59 Viscosity simulation with the MR spherical brake as a damper 60 Spherical MR brake simulating Coulomb friction along x axis 61 Virtual environment simulation for a manual gear shifter in an automobile 62 Collision of the tip with a virtual wall 65 STOOL HW reet 68 Unsmoothwalldisplayu egen Een 68 Usability experiments t usa u a E ERR geg 70 Virtual environment for haptic hole placement 71 CHAPTER 1 INTRODUCTION Dental implants have become a routine procedure in prosthetic dentistry to replace missing teeth Figure 1 1 With the use of dental implants a patient with mis
3. xD 2 sqrt pov commandTorqueX 2 torqueX 2 2 sqrt torqueX 2 torqueX 2 i 115 else xD 2 0 Output outCommand 0 xD 0 outCommand 1 xD 1 outCommand 2 xD 2 CalculateAlpha Block Discrete update if Fh 0 Fd 0 Fh 1 Fd 1 Fh 2 Fd 2 lt 0 f xD 0 Fh 0 Fd 0 Fh 1 Fd 1 Fh 2 Fd 2 Fc 0 Fd O Fc 1 Fd 1 F c 2 Fd 21 else xD 0 1 Output alphalol xD 0 116 B 3 HAPI WINCON HOOKUP CODE B 3 1 WinconHapticsDevice cpp include lt HAPI WinconHapticsDevice h gt tif defined HAVE_WINCONAPT using namespace HAPI namespace WinconHapticsDeviceInternal string libs_array 1 list lt string gt wincon_device_libs libs_array libs_array 1 HAPIHapticsDevice HapticsDeviceRegistration WinconHapticsDevice device_registration Wincon amp newlnstance WinconHapticsDevice gt WinconHapticsDeviceInternal wincon device libs bool WinconHapticsDevice initHapticsDevice int thread frequency initialize device set success to true if bool success false cout lt lt Using Wincon Haptics Device n Open shared memory Haptics API open shmem to read no handshaking shmem API read SHMEM Create T Wincon Haptic API To H3D DNORD shmem API read size SHMEM NO HANDSHAKING open shmem to vrite no handshaking shmem API write SHMEM Create T Wincon Haptic
4. Enable or disable the haptic device power amplifiers void enable_amplifiers bool enable Enable or disable the haptic device position watchdog void enable_position_watchdog bool enable Set the Damping Gains void set_damping_gains Vec3 gain Set the Stiffness Gains void set_stiffness_gains Vec3 gain Set the Stiffness Position Setpoints void set_stiffness_position_setpoints Vec3 position Return true is one of the fatal errors happened false otherwise A fatal error flag requires the user to restart his her application bool is_fatal_error Return true if the shmem checksum does enable the haptic device power 120 amplifiers i e if the shmem communication is still valid Return false otherwise Used for monitoring bool does checksum enable 7 Return true if the shmem read timeout watchdog does enable the haptic device power amplifiers i e if the shmem communication is still valid Return false otherwise Used for monitoring bool does vrite timeout enable Return true if the shmem vrite timeout vatchdog does enable the haptic device power amplifiers i e if the shmem communication is still valid Return false otherwise Used for monitoring bool does_read_timeout_enable protected Get the device values position orientation etc virtual void updateDeviceValues DeviceValues gdv
5. xuonisod 4 1 UT9 8AS qnS Id VGEH LTT 97 indinopue uuluogpezijeubevuap 2 guld V ld Jej onpuoguonezneuBeueg indujajbue ejBuyindur Ca uolyeunp H lonuoouonezn uSeul q uoneinqjaubewap SG yndujpuewwoo ED 1193SAS qNS T yfo yuoyuonezn useu p ST TA 98 INCINON9 OMUOD LSO9S A EL 1 0 40 040 0 100 1 Cija PpueWWODINO s lonuo quoi noo plous lu 144 En plousaly quiojnoo AAHOOT A A HOOT A lt lt X 90 A XAMOOT A O 1193SAS qAS 001 11140000110 QIT TA 2014 ULEDALISOOSIA 19 0O JUOD JISO9SIA JAH apoienno 19110 U0D SOSIA XAHOOT A AAHOOT A RG XAMH OT A uro9js s qng 19 01JUO JD JISOISIA 6 TT 99 indino jeubispuew woo indino Bojeuy Uleoduwyowes jeudispueuiluogmel ova vo 1J8SuenO UH9 SAS qn jeusigpueumuoo Z TT dg sojjosjuoguonjeznaubetuaq jueunoindu jueunogindui4 3 uomneinp ndinol llonuo uonez neuBeuiep C 24 oud JejjoquoguonezneuDeuieq uoneigieuBeuep 4 4 ss pieoq sy smopulM pessald Aen indu 44 1 au sl 1193SAS qNS H yloyuoyuonezn us
6. HAPTIC SURGICAL AID SYSTEM WITH MAGNETORHEOLOGICAL BRAKES FOR DENTAL IMPLANTS Abstract by Doruk Senkal MLS Washington State University December 2009 Chair Hakan Gurocak This research explored a passive haptic interface as a surgical aid tool for dental implant surgery The placement of a dental implant is critical since mistakes can lead to permanent damage in the nerves controlling the lips long lasting numbness and failure of the implant and the crown on it Haptic feedback to the surgeon in real time can decrease the dependence on the surgeon s skills and experience for accurate implant positioning and increase the overall safety of the procedure The device developed in this research is a lightweight mechanism with weight compensation Rotary magnetorheological MR brakes were custom designed for this application using the serpentine flux path concept The resulting MR Brakes were 33 smaller in diameter than the only commercially available brake yet produces 2 7 times more torque at 10 9 Nm Another contribution of the research was a ferro fluidic sealing technique which decreased the off state torque A spherical brake as a multi DOF actuator was also developed as a possible candidate for actuation of the wrist joint of the haptic interface To the best of our knowledge our design is the first ever multi DOF spherical brake using MR fluid The control system implemented the passive force manipulability ellipsoid as an analytica
7. Output outCommand 0 xD 0 demagnetizationController Block Discrete update xD 2 if Input 0 gt 0 xD 0 1 xD 1 0 else 103 if xD 4 1 amp amp xD O 1 xD 3 xD 2 xD 4 0 xD 1 1 xD 0 0 if xD 2 xD 3 gt demagnetDuration 0 xD 0 1 xD 1 0 xD 4 1 Output if xD 0 gt 0 1 pin0 0 inputCurrent 0 else pinolol 0 if xD 1 1 pin0 0 1 inputCurrent 0 104 jndjngpuewwoy ujreublgpuewwoD 10199 jaguorjejN LIS 4 uoneulxolddy olo i 8910 JPUBL OI indinopueuiuo yguoyeuwixolddyas104 indujanbio pue oo ynduir 4 13 011uoyBulyaje7 lt ndjngpuewwo 21ejo11uo oBuruo1e1 1 4 HnoyoL il65uvz 5elloAsloq morf lt 9910 j19sn indu lt T ZAX IdV qe H2910 jpueLULIOI G dy asHenbioj puewwoo lt T indindp indurpoL 1nor01 lt I ejoul lt 191 1esyoet 1esyoct 1esyo Lt jnduige jnduize SHOA Jesyoet 0 jesyoct 82 esou 1 0 ee 44 os2 s8v 0 ze oro Buiuonipuoojejeuionuejod indinouajewonuejog
8. brake after the joystick was pulled away from the wall 56 The algorithm for the wall collision experiment is as follows while simulation is running do if dotProduct commandForce userForce lt 0 Condition for engaging the brake activate brake in forward direction else if readyToDemagnetize is true and brake is active Condition for demagnetization reset counter set readyToDemagnetize to false activate brake in reverse direction else Condition for deactivating the brake deactivate brake end if if counter gt demagnetizationDutation Setting the demagnetization duration set readyToDemagnetize to true deactivate brake end if end while loop Where commandForce is the force vector from the virtual environment and userForce is the input force obtained through the force sensor readyToDemagnetize and demagnetizationDutation are variables used for controlling the demagnetization 57 To conduct the wall collision experiment the brake was driven at 1 5A current using the Quanser board 2 T T T T T 1 51 E 2 T o E e 0 5 1 Or 4 0 5 1 l l 1 l 0 1 0 0 1 0 2 0 3 0 4 0 5 Position rad Figure 6 9 Simulation of collision with a virtual wall at position 6 0 The joystick handle is first pulled away from position zero through approximately 0 45 radians Then it is pushed back towards the virtual wall for the collision simulation 6 2 3 Transie
9. was used and static friction torque Tstatic Was set at 5 Nm dynamic friction torque Taysamic Was set at 3 Nm Torque N m o 10 5 0 5 10 Velocity rad s Torque N m o 10 5 0 5 10 Velocity rad s Torque N m o Time s Velocity rad s o 2 3 4 Time s o Figure 6 6 MR brake representing Coulomb friction 50 6 1 5 Transient Response The transient response was obtained using the wall collision results The purpose of the experiment was to find the time constant of the brake and to explore the effect of input current on the response The experiment was repeated for current levels of 0 25 0 5 0 75 and 1A Transient response was recorded after the initial contact with the wall The torque output resembles a typical first order system response in reaction to step changes in current Figure 6 7 The time constant was measured to be 60 milliseconds by overlaying a simulated first order system response on the experimental data Torque N m S 0 0 2 0 4 0 6 0 8 Time s a Transient response at 0 25A 0 5A 0 75A and 1A Torque N m 0 L 0 0 2 0 4 0 6 0 8 Time s b Simulated first order response overlaid on the experimental data at 0 75A Figure 6 7 Transient response and time constant of the MR brake 51 6 1 6 Discussion Hysteresis behavior can be observed in the torque
10. xD n TO4 j i n for i 0 i lt 2 1 for j 0 j lt 2 jt t xD n jacobian i j n Output inb i L for i 0 i lt 2 1 theta i xD i for m 3 m lt 18 m TO4out m 3 xD m for m 19 m lt 27 m J m 19 xD m LatchingController2 Block Discrete update xD 2 f Input 0 gt 0 1 xD 0 T 114 xD 1 0 else if xD 4 1 amp amp xD 0 1 xD 3 xD 2 xD 4 0 xD 1 1 xD 0 0 if xD 2 xD 3 gt demagnetDuration 0 xD 0 1 xD 1 0 xD 4 1 Output if xD 0 gt 0 output 0 inputCurrent 0 output 1 inputCurrent 0 output 2 inputCurrent 0 output101 0 output 1 0 output 2 0 if xD 1 1 output 0 1 inputCurrent 0 output 1 1 inputCurrent 0 output 2 1 inputCurrent 0 PashaSimpleHapticControllerwSelectiveBrakingwContForceA pprox Block Discrete update if commandTorqueX 0 torquex 0 lt threshold 0 amp amp torqueX 0 0 xD 0 sqrt pow commandTorqueX 0 torqueX 0 2 sqrt torqueX 0 torquex 0 else xD 0 0 if commandTorqueX 1 torqueX 1 threshold 0 amp amp torqueX 1 0 xD 1 sqrt pow commandTorqueX 1 torqueX 1 2 sqrt torqueX 1 torqueX 1 else xD 1 0 if commandTorqueX 2 torqueX 2 lt threshold 0 amp amp torqueX 2 0
11. 112 B 2 2 Haptic Surgical Aid System Simulink Code potsVoltage2Angle Block Discrete update double thetal double T04 4 4 double dl inb i di a2 a3 d3 tid t2d t3d xD 0 xD 1 xD 2 thetal theta2 a2 j Ni 0 a2input a3input 0 0 0 0 1 2 t3d tid 3 t2d 3 theta2 theta3 tld t2d t3d jacobian 3 3 d3 a3 1 1 4 412 357 143 16 90 tloffset 0 1 5 805 340 9 t2offset 0 4 6 357 143 61 t30ffset 0 141592653589793 180 141592653589793 180 theta3 t3d 3 141592653589793 180 TO4 0 0 cos thetal cos theta2 cos theta3 cos thetal sin theta2 sin theta3 TO4 0 1 cos thetal cos theta2 sin theta3 cos thetal sin theta2 cos theta3 TO4101121 sin thetal TO4 0 3 cos thetal cos theta2 cos theta3 cos thetal sin theta2 sin theta3 a3 cos thetal cos theta2 a2 sin thetal d3 TO4 1 0 sin thetal cos theta2 cos theta3 sin thetal sin theta2 sin theta3 TO4 1 1 2 sin thetal cos theta2 sin theta3 sin thetal sin theta2 cos theta3 T04 1 2 2cos theta1 TO4 1 3 sin thetal cos theta2 cos theta3 sin thetal sin theta2 sin theta3 a3 sin thetal cos theta2 a2 cos thetal d3 TO4 2 O sin theta2 cos theta3 cos theta2 sin theta3 TO4 2 1 2sin theta2 sin theta3 cos theta2 cos theta3 TO4 2 2 0 TO4121131 sin
12. Aq UMDIJ OOS 06268 AUSIOAIUN lDIS UOJBUIUSO M Ki 05695 wiy 0000 87 SZ EBEN 5031198 6 Rd wor d r SEO TLL GIS b EI ysn az te Km geen 5005 ron oos dus 0 TERA zoo t x x TT Caen TT urma sg li swa m ns su dj GJE HOLE pegpeds uu 6 IAE ISNVI AVN Hd 30 39V duU3LNI O332X3 LON OO DOVWVO_TVLLNSLOd OL 300 S31V1d 39V3N31NI AOW 3i YO NISOOT LON OG Ss LI E Ve UD A or Ze x L WA ULL TYE jo epeus age pe joo pur Dugunojv 88 APPENDIX B SOFTWARE 89 yeuBigpue wwos jeubispuewwo mel uo pajas uone nuis L D J9 oyuoguonezneuDeuiep 4 jndingaejo1uo uonezn uSeul p 1 lo uoouonouqquionoo Ahuooja A jndingje o1uoA uro no g 49 01 UODA IS OSIA A Ayooja A AA A A A H lonuoouonezn uSeul p jndujpueuiuoo jndjnQpuewwogpezijeubewap jndujajbue indingaejo1u0 2 YSOISIA x Auoojo A lt jndujuonsog J l omuooS A 1uBiH inoajelsjjem urAuonisod uruonisod nopueuiuoo dv AEH UX9910 J G mosjbue UIA9910J XAHOOL A lt 500 amp xuon sod A YIOJ
13. azimuth angle degrees Figure 3 7 Magnetic flux density at different azimuth angles MR spherical brake can exert moments along all 3 axes hence maximum torque along each axis should be individually calculated This is accomplished by integrating the tangential component of yield stress along the ball surface for each axis It should be noted that because of the symmetry along z axis maximum torque that can be exerted ee E 66 5 along x and y axes will be equal 24 Figure 3 8 Calculating torque about the x and z axes o is the azimuth angle is in the horizontal x y plane Torque about the z axis can be calculated by integrating shear stress on the sphere from o 0 to o a Figure 3 8 By taking advantage of the symmetry along the z axis the moment arm can be written as r sin o and the shear force on an infinitesimally thick ring around the sphere can be written as T 21 1 sina do Then the integral becomes T fo r sina r 21r sin o do 3 8 r 2x y3 Sinzeyl T eren 2 2 Jl 3 9 56 98 66 9 Using a similar approach torque along the x and y axes can be calculated Hovvever the absence of symmetry along the x axis makes it impossible to use the above equation directly Instead the torque that can be created by the surface 25 area of the opening on the socket is subtracted from the torque that can be created by a comp
14. motion it is impossible to create frictionless wall surfaces After the initial contact with the wall if the user wanted to slide on the wall surface the brake had to be kept engaged in order to prevent further penetration into the wall This resulted in friction forces that were as high as contact forces on virtual surfaces Another artifact was the problem caused by noise in the optical positioning system When the joystick was kept very close to the wall surface the noise in the positioning system caused the tracker to penetrate the 63 wall surface even though no motion was present in that direction Since demagnetization was performed after each wall collision to increase the backdrivability of the device it resulted in slight magnetization in the opposite direction or friction due to the combined effect of the demagnetization pulses This problem can be avoided in the future by using a different positioning system or using optical sensors with better noise characteristics 64 6 3 Passive Haptic Interface Experiments Two set of experiments were conducted with the developed dental implant aid system The first set involved characterization of the device in displaying rigid virtual objects The second set involved analysis of the overall accuracy of manipulation tasks with the system as a group of users tried to drill holes in polyurethane foam simulating the dental implant surgery 6 3 1 Virtual Wall Collision 20 T T T T
15. three un actuated joints at the wrist were arranged such that the combined motion creates a spherical joint at the tip The wrist joints use 2 ball bearings each to reduce friction and play The brakes provide actuation and structural support for the remaining joints They were built with a pair of tapered roller bearings under pretension to minimize play Three 10 turn precision potentiometers were used at the actuated joints for position sensing Model 3509s 8 202 from Bourns Inc 50 They were connected to analog input ports of a Q4 hardware in the loop interface card by Quanser Inc 51 The manufacturing specifications for these potentiometers are 0 021 resolution with 0 5 linearity The range of motion is 360 around Ji 70 around and 170 around J with a maximum reach of approximately 0 86m Table 4 1 shows the Denavit Hartenberg parameters for the arm Table 4 1 Denavit Hartenberg parameters for passive haptic arm i ai 1 di 0 1 0 0 0 0 2 7 2 0 0 02 3 0 0 411m 0 3814m 03 4 0 0 0 0 37 The haptic arm has a range of motion of 360 around Ji 70 around and 170 around J3 With these ranges of motion given link lengths and the interference from the base plane the work volume of the arm can be constructed for the center of the wrist joint Figure 4 3 This gives a maximum reach of 0 86m for the arm 0 8 0 6 0 5F 0 4 0 3 0 2 Z position m 0 1 0 1 0 2 0 3
16. API From H3D DWORD shmem API write size SHMEM NO HANDSHAKING for int i i shmem API write num doubles i write_API_inputs i 0 0 if SHMEM Write shmem API write 0 DWORD shmem API write size amp write API inputs DWORD api integer timeout device_id 0 success true return success else stringstream s s lt lt Error writing to Wincon Haptic API shared memory lt lt Warning Failed to initialize Wincon haptic device setErrorMsg s str setErrorMsg s str success false 117 return success bool WinconHapticsDevice releaseHapticsDevice release all resources allocated in initHapticsDevice disconnecting the device if device id 1 1 clean up the shared memory reset checksum_api_w in API shmem to 0 for int i 0 i lt shmem API write num doubles i write_API_inputs i 0 0 if SHMEM Write shmem API write 0 DWORD shmem API write size amp write_API_inputs DWORD api_integer_timeout stringstream s s lt lt Warning Failed to reset the Wincon haptic device setErrorMsg s str reset the device id device id 1 do not free the shared memory return true void WinconHapticsDevice updateDeviceValues DeviceValues amp dv HAPITime dt HAPIHapticsDevice updateDeviceValues dv dt if device id 1 1 re
17. Cartesian coordinates Tg and Tp are the command torques in joint coordinates and user input force in joint coordinates respectively The force approximation algorithm compares the two joint torques and computes command signals depending on their relative direction T Tq if Ta 14 lt 20 if Ta Th gt 0 5 2 Where T is the command torque The command torque is applied by the brake if its direction is opposing to the users input force If the user s input and the command torques are in the same direction then there is no need to activate the brakes z lt 0 Implementation of the low level controller in simulink can be found in Appendix B 2 43 CHAPTER 6 EXPERIMENTS AND RESULTS 6 1 Rotary MR Brake Experiments A test setup was constructed to identify the parameters of the prototype MR brake Figure 6 1 The setup consisted of a torque sensor from Transducer Techniques Inc 52 attached to the brake chassis a brush assembly to allow multiple rotations of the shaft and a high precision potentiometer to measure position Real time control was implemented using a Quanser Q4 Series hardware in the loop board connected to SIMULINK via WinCon software 51 Force sensor MR Brake Potentiometer Brush assembly Figure 6 1 Experimental setup for rotary MR brake 44 6 1 1 Braking Torque In this experiment the goal was to determine the braking torque as a function of coil current The current on
18. HAPITime dt Send forces and torques to render virtual void sendOutput DeviceOutput amp dv HAPITime dt Initialize the haptics device Use the HapticThread class in Threads h as the thread for haptic rendering Nparam thread frequency is the desired haptic frequency 1000 is the maximum allowed frequency that can be specified Setting this parameter to 1 means run as fast as possible It is recommended to use the default value for most users virtual bool initHapticsDevice int thread frequency 1000 Releases all resources allocated in initHapticsDevice virtual bool releaseHapticsDevice The Wincon Haptic API device IDentification number for this device int device id shared memory handles shmem t shmem API write shmem t shmem API read characteristics on Wincon Haptic API outputs number of doubles to read const static int shmem API read num doubles 10 size in bytes of shared memory to read const static DWORD shmem API read size shmem_API_read_num_doubles sizeof double characteristics on Wincon Haptic API inputs 121 number of doubles to write const static int shmem_API_write_num_doubles 6 size in bytes of shared memory to write const static DWORD shmem_API_write_size shmem_API_write_num_doubles sizeof double loop timing ms const static int api_update_dt_ms 1 shmem timeout value ms con
19. Nm braking torque Another contribution of the research is an optical position measurement system that eliminates the gimbal mechanisms that are typically used in spherical joints for position measurement In the following sections a review of the relevant studies in the literature is provided for MR Brakes 1 1 1 MR Brakes MR brakes create braking torque by changing the viscosity of the MR fluid inside the brake In the inactive state the fluid has a viscosity similar to low viscosity oil Upon activation with a magnetic flux it changes to a thick consistency similar to peanut butter MR brakes are used in many applications including prosthetics automotive vibration stabilization and haptics The MR brakes provide quick response with simple control When used alone or in combination with motors MR brakes have been shown to provide realistic rigid virtual object simulations in haptics applications 26 29 However to obtain significant braking forces torques the brakes are required to be rather large and use high input current There is a commercially available MR brake by Lord Corporation 30 This brake model RD 2087 01 has 96 6 mm diameter 43 7 mm width and can provide 4 Nm torque with 1 5 A current input An MR brake was designed as a clutch for automotive applications 31 The clutch had 152 mm diameter and required 4 A to generate a braking torque of 6 9 Nm A single disk MR brake was designed for haptic rendering 32 The brak
20. O b aouejsig JOSUIS euro urew giueiuoo b 7 lt osr 0 I 1 uogouny Jeysued au L pazyersig Kuonsod sets L 1xuonisod Lindu Boyeuy Sc Jesueno UT 3SAS qns jnaugpuonisoq srra 95 IdydqeH 1noeu am uong lt 90104 paiseq o yuone queno lt moauu peas uo91iduollelu uo lt 4 Zanbio puew oo me Auonejueuo 4 A nb o j puew woo z faioojo nk A All ol A lt Z8910 jPUBLULUOI xiooje c zuonisod le A990 4puew woo X90104pue wwo9 1e onuooondenuossi oo EM Sqy99104 Sqy201041no ejBuyuonisuen 1noejejS ew la Ce 91e1S Ie uno A9910 PUB UU ULIOI mey eubigpueuwuoo q llonuooondeHuossir 0011544 inoajbue X9910 jpue wwo I3 a ajbuyijno ux 9910 000 z 4 UIA 9910 1 UrojsAS qn Zou PA TUSH OVA Aejaqdniamod uku ngd IO AULA lt ulxuonsod ox ux lt 96 Zenbio puewwoo puew woo yx nbio jpueuulio2 790104 PUB WOO A9910 jPUBLUILLIOI X8910 jPUELLUUIOI Luong OL NOHUONEJUSHO J KMotuajy paeus alum Vol duorejualo 1noeui L L meAuollejuallo Bumeuspuey ON IN3INHM Jesueno ZAHOO A A JIDO 9A Aiowsy parys pesH gt gt o S o gt 1noeui peas 2 A zuonsod Bumeuspuey ON IN3WQu 1esueno Auollisod z
21. T Force N o I 4 a4 1 L 1 1 l 1952 0 9 0 88 0 86 0 84 0 82 Position m 0 8 0 78 pth gr Et a a 20 0 78 0 8 P d 0 82h 10 _ 2 0 84 I 8 5 EP NS 8 E 2 age ol e 0 88 Bp e ve 0 9 Sos 1 15 2 25 3 35 4 45 99 os 1 15 Time s b 2 25 3 35 4 45 Time 3 c Figure 6 14 Collision of the tip with a virtual wall 65 Wall collision experiments were conducted for a very stiff wall k 10 N mm A wall surface was created at 0 8 m using H3DAPI Figure 6 14 Starting at 0 9 m the hand piece of the haptic arm was moved towards the wall until the tip made contact with the wall As the tip was pressed into the wall the force felt in the user s hand increased to about 15 N Then the contact with the wall was kept for about second and the hand piece was moved away from the wall back to the 0 9 m position The results show very rigid wall collision with clean release However as it can be seen on Figure 6 14 there is significant off state friction This off state friction is mainly due to magnetization of steel elements inside the MR brake leaving a residual magnetic field over the MR fluid hence higher off state torques even when the current to the brake is turned off Previously we were able to collapse the residual magnetic field by giving a reverse current pulse to the brake Figure 6 4a This reduced the of
22. brake could apply bottom vertical part of the curve in Figure 6 3b This caused the sticky wall feeling and also increased the observed velocities after the release since the user had to pull the handle very hard to break away from the virtual wall It was found that the MR brake could simulate coulomb friction with good accuracy Although the mathematical model for such a simulation is rather simple active actuators actuators that can add energy into the system such as servo motors have great difficulty in simulating discontinuous forces because of stability issues 56 The transient response of the brake closely resembled first order system behavior with a time constant of 60 ms The experiment was repeated for different input currents and it was found that input current had no effect on the time constant 53 The 10 9 Nm maximum torque output obtained from the experiments was slightly higher than the 10 83 Nm obtained from the calculations This small deviation can be attributed to mathematical simplifications such as Bingham Plastic model inherent errors that were caused by finite element analysis of the magnetic field and neglecting some physical effects such as Coulomb friction 54 6 2 Spherical MR Brake Experiments A series of experiments were conducted with two objectives in mind First objective was to identify the parameters of prototype spherical MR brake and the second objective was to test the applicability of t
23. current curves in figure 6 2 This behavior is due to the magnetization of steel elements in the MR brake 57 Magnetization in ferromagnetic elements does not relax back to zero even if the imposing magnetic field is removed Unfortunately this behavior not only adversely affects the controllability of the MR brake but also increases the off state torque greatly In our experiments the residual magnetization kept the off state torque at 1 05 Nm after applying and removing 1 5A coil current It caused unwanted off state friction and reduced the backdrivability of the brake To overcome this problem the controller was modified At the instant the brake was turned off it was reactivated in the reverse direction with a very short current impulse An impulse with amplitude of 1A and 55 ms duration Current plot in Figure 6 3a and 6 4a was found to be enough to collapse the residual magnetic field reducing the off state torque to 0 08 Nm from 1 05 Nm The effect of hysteresis is even more apparent in the damper experiment Figure 6 5 This was mainly caused by the continuous nature of this experiment leaving no time to apply current in the reverse direction in order to demagnetize the brake As a result the system output two different torque outputs for a given velocity and also limited the minimum amount of viscous torque that could be simulated with the system Another difficulty in this experiment was the time lag of the system having an adverse effect o
24. display the controller renders the wall almost frictionless by locking joint 2 at the same time releasing joints 1 amp 3 resulting in the arc seen in Figure 6 15a For the unsmooth wall display the controller does almost the same thing initially by locking joint 2 and releasing joints 1 amp 3 However the resulting arc is not parallel to the wall surface Figure 6 16a When the user drags the hand piece along the wall surface he she is constrained on this arc creating a much desired although artificial pullback action at the same time giving the wall its jagged feeling 6 3 3 Drilling with Haptic Feedback 6 3 3 1 Objective The goal of this experiment was to analyze the overall accuracy of manipulation tasks with the system The experiment involved drilling holes in polyurethane foam simulating the dental implant surgery The hypothesis was that the passive haptic device would result in positioning accuracies similar to or better than the accuracies of other dental surgery systems 6 3 3 2 Procedure Twenty one users 18 male and 3 female performed the usability experiment Each user first went through a training session where the experiment setup was introduced The users then performed a trial run where they placed holes at 3 different 69 locations No data was collected during the trial run and the hole locations were different than the actual experiment Figure 6 17 Usability experiments Following the
25. haptics applications Figure 3 9 The joystick handle equipped with force sensing was attached to the spherical MR brake An optical position measurement system was also attached to the handle Handle Optical Positioning System Force Sensor Spherical MR Brake Figure 3 9 Force feedback joystick 27 3 2 1 1 Optical Position Measurement System Conventional spherical joints use encoders attached to three different axes of the joint through a gimbal mechanism to measure the orientation 47 Although this is a viable approach we explored another approach to meet the design goal of building a compact system Three IR sensors by Sharp Inc were used to build an optical triangulation system to measure the position of the joystick handle Figure 3 10 Figure 3 10 Optical triangulation system using IR sensors for position measurement The sensors measure distance by sending an infrared signal and receiving the signal that bounces back from a surface They have a range of 4 to 30 cm and generate an analog signal corresponding to the measured distance The sensors were placed in a triangular arrangement angled slightly outward and facing down 28 Figure 3 11 Position and orientation of the optical sensors on the joystick handle The data received from the sensors are the distance measurements to three points on the base plane The coordinates of these three points with respect to the joystick handle can be foun
26. oNuome nuis ureJgerq yunus ur 3s4AS piy epis ing ndeH p z g IN31SAS QIV 1VII94NS DILdVH 5 C 8 105 Buidurequonisog indjnolajeuuorjualod CC 9191114104 G S AL 91911 104 G S G 4 4 t1941 440d vz G D lt E 9H 14104 Mey SIOd lt lt 00 35 4 002 bi z 01911 104 4 0025 d oav roO 003 Jasueno 1191114104 003 s 002 J0128 es Buidueg L ur sAKs qn SUIMONIpuoDIO OMIONUEIOg TZ g 106 indinopue u uuogiajjosuogbulyojenj ce jueunoindui zisjjoruogbulyojej uonemgjsubewap inopue uiuloo SCHEI HIN ova vo 1esueno 1 8164000 1 juaing indu ES uomneang 1euDeuieq 49 pueoghey smopulM passald Aen 1 uls WoISAS qng Tal 9774 oneinjeg obeo ENCA ujyeuBispuewwo0g WoIsAS qng nd n pusunu62 TT Z q 107 uonesuaduogeuydje eydjyarjejnojeg 991041e5n a Cs 1099 8533 is Cs xo1ddy ao1o juoombuneigenoajashuajjonuogondeHajdulseused n q rdiinui eydje pudiy lei
27. of the haptic feedback In this reaserch we explored an alternative sealing method called ferro fluidic sealing Ferro fluidic seals are normally used for sealing the lubricants inside rotating assemblies like gearboxes bearings etc By placing permanent ring magnets at the two ends of the rotor shaft we aimed to solidify the MR fluid inside the end caps and hence prevent rest of the fluid from leaking out 2 2 Development of a Spherical MR Brake as a Potential Wrist Mechanism In this first prototype the commonly used pen based haptic devices 21 were taken as the basis Hence three joints were actuated with MR brakes for creating the haptic feedback and the remaining three joints at the wrist were left un actuated to provide motion in 6 DOF In essence when the MR brakes are activated the position of the base of the hand piece is constrained whereas the orientation of the hand piece is not In the future prototypes the last 3 DOF need to be actuated if orientation also needs to be 11 constrained This can be accomplished by using smaller MR brakes at the wrist However since rotary MR brakes are 1 DOF devices a gimbal mechanism need to be employed with 3 rotary MR brakes to create the spherical joint at the wrist We explored another alternative by designing a spherical MR brake The spherical brake allows motion about any arbitrary axes When it is activated it restricts motion around all three DOFs simultaneously To the best of ou
28. po commandForce input 0 ta offset userForce luserForce PotsVoltage2Angle HighLevelControllerHookUp ForceApproxim 3i FixedStepDiscrete Figure D 2 Simulink diagram 8 Execute real time code by clicking on Start button in Wincon Server Figure D 3 Untitled WinCon Server File Client Model Plot Window View Help seni di Figure D 3 VVincon Server 9 Remove the force sensor bias by first holding the handpiece of the haptic arm lightly and then pressing L key on the keyboard while simulink window is active 128 10 Start H3D Viewer by clicking on Start gt All Programs gt H3DAPI gt H3Dviewer 11 Open model file DrillingReal x3d by clicking on File gt Open File in H3D Viewer Figure D 4 L H3DViewer File Rendering Device Control Yiewpoints Navigation Advanced Help DrillingReal x3d C X3Dsoftware x3DScenes PashaExperiments Figure DA H3D Viewer with the model file loaded 12 The simulation should now be running adjust the calibration constants a2 a3 tloffset offset and t3offset if necessary 129
29. ratios For that reason either very large motors need to be used or a very high reduction ratio needs to be employed by using transmission mechanisms such as gears or pulleys The transmission elements add to the overall robot size as well as unwanted effects such as backlash deflection or slippage As MR brakes usually have several orders of magnitude higher torque to size ratios than DC motors a design that uses MR brakes with direct coupling to the joint axis was chosen for the haptic interface To the best of our knowledge there is only one commercially available rotary MR brake in the market 31 Although this brake provides comparably higher torque to same 10 sized DC motors it is a disc type MR brake along with many other MR brakes in the literature In this research we explored the design of new type of brake based on serpentine flux path concept Using this method we aimed to build compact yet high torque drum type MR brakes Another contribution of this research was in sealing of MR fluid inside the MR brakes Traditionally rubber seals are used to keep the MR fluid from leaking out Although this is a perfectly viable method for MR brakes that are used in applications such as exercise equipment or automobile clutches the friction created by such a sealing method is highly undesirable in haptics applications Any unwanted off state friction would reduce the back drivability of the haptic device effectively decreasing the realism
30. the tip of the arm renders the virtual world graphics and computes collisions between the virtual objects and the tip to generate command forces to be applied to the user s hand 5 1 2 Low Level Controller The low level controller receives joint positions and force input data from the user s hand It computes the forward kinematics and the Jacobian matrix for the haptic arm Figure 5 1 Forward kinematics is used to find the Cartesian position of the tip Then the tip position is sent to the high level controller which generates the command force necessary to create the haptic sensation The command force is returned to the low level controller which uses the Jacobian matrix to calculate the necessary command 41 torque at the joints User s input force is also converted into joint torques Both the input and the command torques need to be used to calculate the command signal going into the MR brake servo amplifiers Appendix C This is accomplished using a force approximation algorithm explained next As passive devices can only apply forces opposing the user s motion traditional control algorithms for haptics cannot be used for the control of MR brakes The problem occurs mainly in tasks that require following a rigid surface wall following In following an ideal frictionless wall the force vector applied to the user s hand must be normal to the surface regardless of the configuration of the mechanism or the direction of th
31. 02 peznersiq 0248 4 lt puonouny 1ejsuel 02 peznalosia lt oz s 13519 peri adi guonouny Jejsuel L4 1nduj Dojeuy 0c peznolosiq indu E S ino E Woe ai guonouny Jejsuel 9QV JIN SERIES oz slu uunulsul jeuoneN 21 9 0610 erejs ox 4 025 juomnound 1ejuel L p z nesia lt 11 025 al pieoqfoy SMOPuIM x p ss ld Aen o euis urojs s qng 1osuoS LAH ILV T T 8 93 ueiqooepiosues y yndjnox CO ueiqooer Imdu LV MEHJOSUES d UIIJSASQAS uUEIQOSE Y FITA 94 guley eAejaq 9ureo LAejag Z os lt Z 14 09 zuonouny Jajsuel ues ule guonouny peznessiq poenqns zKejeq 196o1uj peznexsig urea AAIDOIaA 236 penqns Aey q 1obaluj 5 4 2 X PIOJ A ere x se Se Cs fea L lt ZAAMOOJOA LXANIOJSA gules Zules d d 1ejj ouoo SANA SOW SPHOM zuonsod Luonoun J 4 Lzuonsod p zn osiq LAuonsod P S Sc lt lt Sc rAuonsod uuojsugi uone nDueu p J0joeAuonsod uonoun4
32. 6 3 1 Virtual Wall Collin ut AS 65 6 3 2 Virtual Wall Following 66 6 3 2 1 Smooth Wall Display 67 6 3 2 2 Unsmooth Wall Display 68 6 3 3 Drilling with Haptic Feedback 69 O V 69 0 7 ToP aa RE PE 69 6 3 Resul ssn po lub 71 07 sas a Qu yaa tact eer D ee 72 7 CONCLUSIONS AND FUTURE RECOMMENDATIONS 74 BIBLIOGERONDPITY EE 77 APPENDIX A ASSEMBLY DRA WINGS 000 dos 82 Bic SOPEW ARES L AL uu 89 C SERVO AMPLIFIER CIRCUITS voii di ERR s 126 LP USERMANUA EE 127 vii 3 1 4 1 6 1 7 1 LIST OF TABLES Theoretical braking torques for different Spherical MR Brake sizes 27 Denavit Hartenberg parameters for passive haptic arm 37 Results of haptic drilling experiment a 71 Design Specifications of the Prototype MR 74 viii 1 1 1 2 1 3 1 4 1 5 1 6 3 1 3 2 3 3 3 4 3 5 3 6 3 7 3 8 3 9 3 10 3 11 3 12 4 1 4 2 4 3 4 4 5 1 LIST OF FIGURES X Ray image of two dental implants eene 1 Mandibular canal with Mandibular Nerve A 2 Computer milled surgical template and opto electronic sensors
33. 8 abrasive on the O rings In this research we developed a novel sealing approach using ferro fluidic sealing assemblies on both sides of the brake shaft Figure 3 4 The ferro fluidic sealing assemblies have three primary roles in the brake assembly 1 Prevent the leakage of the MR fluid 2 hold the roller bearings necessary to keep the shaft in place and 3 connect the chassis to a stationary point via aluminum flanges Bearing Magnet ring Magnet ring Flux path Ferrofluidic Seal Brake shaft Ferrofluidic seal Steel casing Air gap Figure 3 4 Ferro fluidic sealing using a ring magnet left Ball bearing supports the rotor Cross sectional view of the ferro fluidic seal and the flux path right These ferro fluidic seals work by using the MR fluid itself as a sealing element Circular magnets placed at both ends produce magnetic flux paths that cross the gap between the shaft and the chassis Figure 3 4 The magnetic field increases the yield stress of the MR fluid inside the gap which builds a pressure differential that keeps the MR fluid from leaking out As there is no active contact between solid bodies this approach helps decrease the off state friction tremendously which is critical in haptic displays 19 3 2 Spherical MR Brake Devices that use MR fluids have several advantages over devices with ER fluids Yield shear stress of MR fluids is much higher than ER fluids this leads to higher torque outp
34. Device device i create func if device gt initHapticsDevice thread frequency hd reset device hd device state HAPIHapticsDevice INITIALIZED setup_haptic_rendering_callback hd gt setup_haptic_rendering_callback if setup_haptic_rendering_callback hd shaptic rendering callback data this if hd thread thread hd thread max stiffness hd gt getMaxStiffness break else delete device ifdef WIN32 tendif if hd get stringstream s s lt lt Could not init any haptics device Make sure one is lt lt connected properly x lt lt ends setErrorMsg s str return false else return true B 3 4 AnyHapticsDevice h ifndef __ANYHAPTICSDEVICE_H define __ANYHAPTICSDEVICE_H include HAPI HAPIHapticsDevice h namespace HAPI class HAPI API AnyHapticsDevice public HAPIHapticsDevice public Constructor AnyHapticsDevice Destructor virtual AnyHapticsDevice Returns a pointer to the device that is actually used 123 PE that cont inline HAPIHapticsDevice getActualHapticsDevice return hd get Enable the device Positions can be read and force can be sent inline virtual ErrorCode enableDevice ErrorCode e HAPIHapticsDevice enableDevice if hd get hd gt enableDevice return e Disable the device in
35. HAPTIC SURGICAL AID SYSTEM WITH MAGNETORHEOLOGICAL BRAKES FOR DENTAL IMPLANTS By DORUK SENKAL A dissertation thesis submitted in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE IN MECHANICAL ENGINEERING WASHINGTON STATE UNIVERSITY School of Engineering and Computer Science DECEMBER 2009 To the Faculty of Washington State University The members of the Committee appointed to examine the dissertation thesis of DORUK SENKAL find it satisfactory and recommend that it be accepted Hakan Gurocak Ph D Chair Xiaolin Linda Chen Ph D Wei Xue Ph D 11 ACKNOWLEDGMENT I am deeply indepted to Dr Hakan Gurocak my thesis advisor for his valuable support through my Master s studies Not only he helped me through every step of the research with his supervision and advice but he also helped me with matters outside the lab I believe his guidance will be influential throughout the rest of my career I gratefully thank Dr Ilhan Konukseven my professor at the Middle East Technical University for pointing me in the right direction while I was working for my Bachelor s Degree I would also like to express my sincere gratitude to Troy Dunmire and Chad Swanson for machining numerous intricate parts and their helpful input during the manufacturing phase of this project Finally my sincere appreciation and thanks are due to my family for their encouragement and continued support iii
36. L to render graphics and has its own haptics renderer called HAPI We chose the proxy based Ruspini algorithm for the haptic rendering 42 Low level controller computes coordinate transformations for the force sensor and the triangulation for the optical sensors The data are then sent to the high level 32 controller which generates the command force necessary to create the haptic sensation The command force is returned to the low level controller which processed the command signal and the force input from the user to compute the necessary braking torque signals This signal then goes to the spherical MR brake through a current controlled servo amplifier Details and implementation of the controller for the spherical MR brake can be found in Appendix B 1 along with the Simulink H3DAPI interface code in Appendix B 3 33 CHAPTER 4 PASSIVE HAPTIC INTERFACE FOR DENTAL IMPLANT SURGERY 4 1 Passive Haptic Interface Design The prototype consists of four components 1 MR brakes 2 Position sensors 3 A 6 DOF force sensor and 4 Hand piece for drilling Figure 4 1 Pen based haptic devices 22 were taken as an example in the design of this first prototype Hence three joints were actuated with MR brakes for creating the haptic feedback and the remaining three joints were left un actuated at its wrist to provide motion in 6 DOF CAD drawings of the passive haptic interface can be found in Appendix A Handpiece MR brake Force sen
37. O A zuonisod Josuegedio4 A99104 X99104 oguonejnuls ureJderq xurnurs eILIeyds p Tg 3NV49 YW 1TVOlH4HdS 3HVMLJOS VG 90 A9910 j ED X82103 4019 AA X92910J C Uo1IMSJosuegeo10J ueIgooer 1OSU9S 1 4 ILV indino A ndui Ly ndino 1 4 indinox JosUASSI1O jMEPUOISS X92910 1 ON10Su S 82103 WIIJSAS qng JOSUIS TT TU 91 Lpaeoqfoy SMODUIM LUONEIQI ED pessald Ae indui o eui S X92103 110 gz oneiqileo 21 s induj bojeuy 10 98 99104 OQV SIN C sjueuiniisu jeuoljeN uomneiqi eo indui A90104 pIeoq ay SMODUIM IO uonemieo pessaid Ae a1e1Shey 9 au S urojsAs qng 10SUISIIAO J NBPUOIIS TT TY 92 Z uonouny 2 9 06 0 4 025 4 02 LLuoroun Jejsuel pezneisiq S k Ke e oz O uUonouUNny saysues L P9Z1 919S 1 Jojoe4Bureog xuyeyyuoneuqileo 4 oz s indjno 1 4 L e 4 lt 02 6001104 JejsueJ L pezijemsiq A 1 zi oz guonouny pezieisiq lt 0c s Zuojjouny de p z nesia E Lal 0c guomnoung 1ejsue pezieios iq GJ ozs e guonoun
38. R brakes can be added to the wrist joint to control the orientation effectively building a 6 DOF haptic display Potential solutions already exists where MR brakes as small as a U S quarter 40 were previously designed Another solution which was explored in this research is a spherical MR brake 25 that is capable of locking all 3 DOF at once To the best of our knowledge this design is the first ever multi DOF spherical brake using MR fluid The prototype spherical MR brake has a diameter of 76 2 mm and can apply up to 3 7 Nm braking torque Since the torque is proportional to the cube of the radius of the steel ball the brake can scale up very well Using this design it is possible to make much more powerful brakes without increasing the overall dimension of the brake significantly An optical position measurement system that eliminated the gimbal mechanisms that are typically used in spherical joints for position measurement was also developed as part of this research However the optical sensors turned out to be susceptible to noise leading to inaccuracies in the position measurements Another method is necessary to measure the position of the steel ball in the brake Design options for the wrist joint using small rotary brakes or the spherical brake can be explored in a future study 75 Experiments to identify the characteristics of the haptic interface for dental surgery have been conducted Initial experiments show that passive hap
39. Vol duorejualo 1noeui L L MeAuonejueuo Bumeuspuey ON IN3INHM Jesueno ZAHOO A A JIDO 9A Aiowsy parys pesH gt gt o S o gt 1noeui peas 2 A zuonsod Bumeuspuey ON IN3WQu 1esueno Auollisod xuonisod woysks qns LIVOCH O T C H 110 MOQUBISIO0 ISJUOLIOJNME1 4 100 151 WO ME lt gBuiyjoowsgea04 ureozw 0935 ZHOQU SSIN 05 lt yBuigjoouiseo103 sil uuowionpoid 0935 MOQU S L wreoiw ND lt lt uiroousssro Aidninywl 09 dosuas 1 V L 14 xuooubsiy vue lt ogts 4 Pommes 0S lt lt q ndino 8 m 1 3 1zuoQuDis4 fun S991041onpoud os 1001562104 lt be LAuoguBs4 4 4 ei kid 0S Surnnoouisssioj dini Ki lt Lxuogubis4 wi lt xujew lt 095 L 4 4 os lt I AHAUI Ay u jet oO 100 jsaq104med i Zu u zu yionpoid Au lt EI dual 40109 q suoneloH G L oum voy ur9js s qng 401001 LT d 111 uoliisod Y SHOA 0979 0 7820 1 1ESHOX nox C e indujuonsogl y Woysks qng uonre ndiuejquonisod 8 T CH
40. Z position m 0 6 0 4 0 2 0 2 0 4 X position m 05 o d Y position m 0 6 0 2 0 0 2 X position m 0 4 0 8 0 8 0 6 04 0 2 0 0 2 0 4 0 6 0 8 X position m Figure 4 3 Work volume of the haptic arm A 6 DOF load cell Mini 45 from ATI Industrial 52 was used for force sensing It is capable of sensing torques and moments in 3D The sensor was placed at the base of the wrist joint to measure the user input forces 38 4 3 1 End Effector The end effector of the haptic interface uses a spherical joint at the wrist and a small drill bit at the end of the hand piece Figure 4 4 The x y z coordinate of the drill bit can be constrained using the haptic feedback created by the MR brakes In this first prototype orientation was not constrained An alignment plate with a sleeve was attached onto the drill bit to provide angular guidance during user experiments explained later Force sensor Spherical joint Drill bit Alignment plate Figure 4 4 Handpiece 39 CHAPTER 5 SOFTWARE DEVELOPMENT 5 1 Control System The system was interfaced to a computer using a Q4 hardware in the loop card by Quanser Inc 51 and a PCI MIO 16E 4 data acquisition card by National Instruments 53 Figure 5 1 Control signal fe Force data Fi Joint positions EE q M PC interface Hapti
41. a Computer milled surgical template 14 b Opto electronic sensors with light emitting diodes on the hand piece with Graphical User Interface of the system 16 Another method for implant placement uses Image Guidance Implantology systems IGI 17 18 These systems use optical sensors with light emitting diodes Figure 1 3b attached to the hand piece of the drill and to the patient to track their relative positions Combining this information with the volume data of the patient s jaw structure makes it possible to view the preoperative planning together with the drill position in real time Using these systems intra operative safety can be increased as critical anatomic structures such as nerves can be avoided with the aid of the graphical user interface from the system 16 However even with the help of the graphical user interface it is difficult to achieve proper position and angulation as random factors such as trembling cannot be eliminated without the guidance of a mechanical system Errors as high as 1 23 0 28 mm on average and a maximum of 1 87 0 47 mm between the planned position of the fiducial point marker have been reported 19 For this reason the quality of the intervention is still largely dependent on the surgeon s skills and experience 1 A surgical robot system for maxillofacial surgery 20 has been developed With this system the surgeon interactively programs the robot during the surgery after which the robot
42. ad the shared memory of the Wincon haptic API all at once if ISHMEM Read shmem API read DWORD shmem API read size SAPI outputs read DWORD api integer timeout stringstream s s lt lt Warning Failed to update values from the Wincon haptic device setErrorMsg s str current position in world coordinate frame Cartesian position and velocity in metres dv position Vec3 API outputs read 0 API_outputs_read 1 API outputs read 2 current linear velocity in world coordinate frame m s dv velocity Vec3 API outputs read 3 API outputs read 4 API outputs read 5 118 Device Euler angles yaw pitch roll radians Vec3 r API_outputs_read 6 API_outputs_read 7 API_outputs_read 8 dv orientation Rotation r Ji bitmask for buttons bit is button 0 bit 1 button 1 and so on value of 1 indicates button pressed dv button status API outputs read 9 void WinconHapticsDevice sendOutput DeviceOutput amp dv HAPITime dt if device id 1 1 latest data from H3D loop to send to the Wincon Haptics API write_API_inputs 0 dv force x JU Fx write_API_inputs 1 dv force y Fy write_API_inputs 2 dv force z Fz write_API_inputs 3 dv torque x Mx write_API_inputs 4 dv torque y My write_API_inputs 5 dv torque z Mz if SHMEM Write shmem API write 0 DWORD shmem API write size amp w
43. and 3 need to be balanced The weight of the linkages is transferred equally to the joints they are attached Therefore mass balancing can be done by considering the joints as point masses For J3 d m d4 m4 mA cos zt f d cos a 4 1 cos z terms cancel out giving d my ds ms m m6 4 2 Since no position variables remain in equation 4 2 Js can be statically balanced Similarly for J2 d m da m cos 0 d cos 0 d cos z B m d cos 0 d B m ds cos 0 ms 4 3 ds cos 0 ds cos z B m d cos 0 dg cos z BI mg Rearranging and substituting equation 4 1 into equation 4 3 cancels out the cos z term leaving dz m da m td m td mp d Mm d my dg 4 4 Since no position variables remain in equation 4 4 J gt can also be statically balanced It should be noted that gravitational constant g was omitted in all of the above equations as it would cancel out in the end 36 4 3 Prototype Implementation Following the design principles of pen based haptic devices 22 three joints were actuated with MR brakes for creating the haptic feedback leaving the remaining three joints un actuated at the wrist to provide motion in 6 DOF The links were made out of carbon fiber tubing and aluminum caps to keep the mass of the system low The
44. ave been developed using the serpentine flux path technique The design activates more surface area of the MR fluid in a given volume by weaving the magnetic flux through the MR fluid multiple times Table 7 1 compares the design specifications of the rotary brake to a commercially available MR brake Table 7 1 Design Specifications of the Rotary MR Brake Designed RD 2087 01 MR Brake by LORD Corp Diameter mm 63 5 96 6 Length mm BOT 43 7 Max torque Nm 10 9 4 0 Max off state Nm 0 1 0 4 Max current A 1 5 1 0 Time constant ms 60 Includes ferro fluidic seal end caps 27 0 mm The prototype MR brake is about 2 7 times more powerful than the commercially available RD 2087 01 rotary brake by Lord Corporation and about 33 smaller in diameter Furthermore with 10 9 Nm torque output it is more powerful than the closest brake in the literature with 6 9 Nm torque 30 but only approximately half the diameter and requires 62 5 less current Another contribution of the research was a ferro fluidic seal which reduced the 74 friction that would normally occur due to using conventional O rings The off state friction was further reduced down to a negligible 0 08 Nm by applying a current pulse in the reverse direction This improved the backdrivability of the brake which is much desired in haptics applications Current prototype is capable of rendering only 3 DOF x y z position In future studies additional M
45. c arm Quanser Q4 HIL card gt MR brakes Servo amplifiers i i i i i i 1 D 1 i 1 i 1 H Position sensors PCI MIO 16E 4 DAQ card 6 DOF Force sensor Matrix multiplication f 1 i i 1 1 1 i 1 Figure 5 1 Haptic arm with the PC Interface and Haptic Rendering The control algorithm was implemented using Simulink by Mathworks Inc 49 along with the WinCon software which enables real time code generation from Matlab Simulink diagrams The Q4 handles signals coming from the position sensors 40 and the command signals going out to the MR brakes The PCI MIO 16E 4 handles the analog signals coming from the force sensor For haptic rendering we implemented a two layer control architecture consisting of a low level and high level controller Figure 5 1 5 1 1 High Level Controller The high level controller is for the virtual environment We used H3DAPI which is an open source haptics package by SenseGraphics AB 41 The H3DAPI uses OpenGL to render graphics and has its own haptics renderer called HAPI The proxy based Ruspini algorithm was chosen for the haptic rendering 42 H3DAPI was interfaced to WinCon API using memory sharing to access I O signals on the hardware in the loop card Code for interfacing between H3DAPI and simulink can be found in Appendix B 3 The controller receives the position of
46. ct on the amount of magnetic flux that can pass through it This phenomenon is also known as core saturation because the magnetic flux density in a magnetic circuit is ultimately limited by the saturation point of the magnetic material being used Another parameter that needed to be optimized was the homogeneity of the magnetic field across the MR fluid gap For our design we aimed at obtaining 1 Tesla throughout the MR fluid which according to manufacturer s specifications is very close to the saturation point of the MR fluid MRF 132LD from Lord Corp 46 This helped make maximum use of the MR fluid in the gap but it also required design of a well balanced magnetic circuit An unbalanced magnetic circuit would cause the fluid to saturate in one part of the brake while leading to insufficient magnetic flux densities at other points For this reason the position of the aluminum ring along the MR fluid gap circumference is critical as it divides the surface area of the MR fluid gap to forward and return paths Figure 3 6 which determine the ratio of magnetic flux density along these two surfaces The magnetic flux is the same at any point along the magnetic circuit Prorward Preturn 3 3 In terms of magnetic flux density B and the surface area A through which the flux flows the same equation can be written as Brorwara Aforvara Breturn Areturn 3 4 22 The ring must be placed at a location where the result
47. d Corp Online Available http www lord com Accessed October 2009 Tsumaki Y Naruse H Nenchev D N and Uchiyama M 1998 Design of a compact 6 DOF haptic interface Proceedings of the 1998 IEEE International Conference on Robotics amp Automation Leuven Belgium vol 3 2580 2585 Rosenberg L B and Adelstein B D 1993 Perceptual decomposition of virtual haptic surfaces Proceedings IEEE Symposium on Research Frontiers in Virtual Reality IEEE Computer Society Press Los Alamitos CA 46 53 MathWorks Inc Online Available http www mathworks com Bourns Inc Available Online http www bourns com Accessed October 2009 Quanser Available Online http www quanser com Accessed October 2009 ATI Industrial Automation Inc Available Online http www ati ia com Accessed October 2009 National Instruments Available Online http www ni com Accessed October 2009 Sakaguchi M Furusho J Takesue N Passive force display using ER brakes and its control experiments Virtual Reality 2001 Proceedings IEEE vol no pp 7 12 17 17 March 2001 80 55 56 57 58 Advanced Motion Controls Online Available http www a m c com Accessed October 2009 An J and Kwon D S Stability and Performance of Haptic Interfaces with Active Passive Actuators Theory and Experiments In Int Journal of Robotics Research vol 25 no 11 pa
48. d as x1 y1 21 w siny d 0 cosy d 3 14 X2 V2 Z2 cos 60 w cos 60 sin y d sin 60 w sin 60 siny 45 cosy d 3 15 2 cos 60 w cos 602 sin y d sin 60 w sin 60 sin y 45 cosy d3 3 16 where w is the distance of each sensor from the centre of the handle y y 30 is the angle between the handle and the direction of the optical sensors and 4 is the distance measurement for sensor n Figure 3 11 29 Once the three points on the base plate are known position of a virtual plane overlapping the base plate can be calculated A x B y C z D 0 3 17 where the coefficients A B C and D can be found by using the following determinants 1 y 2 1 zi X y 1 X1 41 A 1 V2 Z2 B X2 1 Z2 C X2 y 1 D X2 yz Z2 3 18 1 ya Z3 X3 1 Z3 X3 ya 1 X3 ya Z3 Then the relative angles between the base plate and the handle Figure 3 12 can be found through a unit vector that is collinear with the joystick handle NET B 0 sin a 3 19 Mr A 6 sin ere ESCH The optical system can measure the joystick handle position in 3D as the user moves it in any direction As the system works by measuring the relative orientation of the base plate rotating the base plate would have absolutely no effect on the position sensor readings d d2 d3 Therefore the system cannot measure rota
49. d to the hand piece They do not require sophisticated controllers 2 4 Integration with the virtual reality environment and controller As in other pen based haptic devices the user interacts with the virtual environment by grasping the hand piece and moving it The base of the hand piece center of the gimbal mechanism at the wrist joint is represented by a point in the virtual environment The device creates the haptic feedback by selectively locking the MR brakes The resulting resistance forces are used to render virtual surfaces As MR brakes can only apply forces opposing the user s motion traditional control algorithms for haptics cannot be used for the control of MR brakes For this reason a two tiered control algorithm has been implemented in this research First required output forces based on the position of the haptic arm are calculated using a penalty based haptic renderer 41 42 Later these command forces are passed through a force approximation algorithm to create the closest force output that can be obtained by using MR brakes 43 44 This way haptic feedback of tasks such as wall following could be approximated 13 CHAPTER 3 MR BRAKES 3 1 Rotary MR Brake Rotary MR brake designs have two varieties 1 Disk type and 2 Drum type In these designs the braking torque is due to the shear stress of the MR fluid placed in a gap between two rotating surfaces Often the behavior of controllable fluids is represen
50. e 6 18 Virtual environment for haptic hole placement 6 3 3 3 Results After the experiment the hole locations were carefully measured using a caliper In each measurement deviation from actual dimensions x y 50mm amp z 4mm was calculated as the positioning error The average error for each dimension and the standard deviation are given in Table 6 1 Table 6 1 Results of haptic drilling experiment Error Standard deviation x axis 2 88mm 2 51mm y axis 1 90mm 1 46mm Z axis 1 16mm 0 94mm 71 6 3 3 4 Discussion An image guided implantology system had an average error of 1 23 0 28 mm and maximum error of 1 87 0 47 mm 19 Robot assistant systems report deviations of 66 99 6699 1 2 mm 23 24 Our passive haptic system had errors below 2 mm for y and z axis and a little below 3 mm for the x axis These results are on the same order of magnitude as the optical tracking system and other dental robots hence passive haptic devices have potential to be a viable alternative to active servo controlled haptic devices Deflection was the primary cause for the errors Although the haptic arm was built with deflection in mind by using materials such as steel aluminum and carbon fiber as opposed to polymer based materials common in today s haptic devices main cause of the errors was the deflection in the system This became especially problematic in the x axis direction as forces in this direction resulted in
51. e crown cannot support the loads put on them Systems for guiding the implantologist can provide additional safety One approach is to use templates However they have significant shortcomings Conventionally fabricated templates which are based on wax model of the patient s teeth structure do not take the thickness of the mucosa topography of the underlying bone or vital anatomical structures into consideration In addition limitations of current fabrication techniques do not allow fabrication of a template that remains stable during surgery 10 Hence they can only be used to optimize the position of the implants for later prosthodontics treatment For this reason methods that use templates based on volume image data Figure 1 3a have recently been developed 11 12 These methods usually use volume image data of the patient s underlying jaw structure obtained through computer tomography CT or digital volume tomography DVT Advanced manufacturing techniques such as stereo lithography together with specialized implant planning software 13 are used to manufacture drilling templates 14 Although these templates take the hidden anatomy into consideration and significant improvement in placement accuracy is reported they suffer from high costs of CAD CAM processing 15 and added lead time In addition they lack flexibility as any change in the planning requires the manufacturing of new templates 8 a b Figure 1 3
52. e had a diameter of about 80 mm It was smaller than some of the other examples in literature but could only provide 1 4 Nm torque in spite of the 4 A input current Another single disk MR brake with about 130 mm diameter was designed and experimentally tested 33 This brake provided 1 4 Nm maximum braking torque at 0 75 A input current To improve the performance of MR brakes a design optimization method using Finite Element Analysis was proposed 34 The method resulted in 25 height reduction leading to a brake with 120 mm diameter and 38 mm height At5 A input current the brake provided 4 25 Nm torque output 1 1 2 Spherical MR Brake A two DOF joystick was developed for haptics applications 35 The design integrated two one DOF MR disc brakes into a joystick using a gimbal mechanism The MR disc brakes had 78 mm diameter The overall prototype joystick had a base of about 160 mm x 160 mm and provided up to 10 Nm braking force to the joystick handle Another multi DOF device was designed using two groups of MR actuators to simulate virtual forces in 2D 36 The system used four MR rotary brakes each with 170 mm diameter and 10 mm height The overall system size was 630 mm x 540 mm x 970 mm The maximum output torque on the handle was 10 Nm Two other multi DOF devices were reported that used electrorheological ER fluids The first device used both clutch and brake mechanisms to achieve active and passive force feedback 37 The device integ
53. e user s motion However with passive devices this is rarely the case In situations where the user is trying to push into the wall and slide on the surface at the same time the braking torques that are preventing the user from penetrating into the wall also prevent him from sliding on the surface This creates a sticky feeling on the wall surface which greatly degrades the haptic feedback Algorithms to overcome this problem have been proposed A method that uses a very narrow band along the virtual wall was proposed By selectively locking the brakes to keep the tip position inside this narrow band the haptic device was forced to move along the wall surface in a zig zag fashion 54 This algorithm was implemented in a 5 bar planar mechanism using electro rheological brakes A force approximation method called Passive Force Manipulability Ellipsoid Analysis FME was also proposed 43 44 Because of its general applicability to the whole work volume of the haptic device and to the existing proxy based rendering techniques we chose FME for the haptic rendering 42 The controller converts the Cartesian command force coming from the high level controller and the user input force coming from the force sensor into joint torques Ta J Faan tp J F 5 1 Where J is the Jacobian matrix for the haptic arm Fg is the command force generated by the high level controller in Cartesian coordinates Fp is the input force from the user in
54. er mandibular implant placement Journal of Oral and Maxillofacial Surgery 57 1408 1412 1999 Ellies L G Altered sensation following mandibular implant surgery a retrospective study Journal of Prosthetic Dentistry 68 664 671 1992 Serhal C B Steenberghe D Quirynen M and Jacobs R Localisation of the mandibular canal using conventional spiral tomography a human cadaver study In Clinical Oral Implants Research 12 3 Pages 230 236 2001 Berberi A Le Breton G Mani J Woimant H amp Nasseh I Lingual paresthesia following surgical placement of implants report of a case International Journal of Oral and Maxillofacial Implants 8 580 582 1993 Berglundh T Persson L and Klinge B A systematic review of the incidence of biological and technical complications in implant dentistry reported in prospective longitudinal studies of at least 5 years Journal Of Clinical Periodontology 29 3 Pages 197 212 2002 Gray s Anatomy 20 US Edition 1918 Brief J Edinger D Hassfeld S Eggers G Accuracy of image guided implantology Clinical Oral Implants Research Volume 16 Number 4 August 2005 pp 495 501 7 Buser D Martin W Belser U C Optimizing esthetics for implant restorations in the anterior maxilla Anatomic and surgical considerations Int J Oral Maxillofac Implants 19 7 Pages 43 61 2004 Lal K White G S Moreaand D N Wright R F Use of stereo
55. eu p TT TT d B 1 2 Spherical Brake Simulink Code Calibration Block Discrete update if keyState 0 1 xD 0 input 0 Output out 0 xD 0 BiasFT Block Discrete update int 1 if keyState 0 1 for i 0 ic6 i xD i inputlil Output int i for i 0 ic6 outlil i xDlil TriangulationTransform Block Discrete update double xi yl zl x2 y2 22 x3 y3 23 xl 1 0625 0 5 d 0 yl 0 zi 0 866 d 0 x2 1 0625 0 5 0 5 0 5 d 1 y2 1 0625 0 866 0 866 0 5 d 1 z2 0 8660 d 1 x3 1 0625 0 5 0 5 0 5 d 2 1 0625 0 866 0 866 0 5 d 2 z3 0 8660 d 2 A y2 z3 yl z2 zl y3 x1 y2 z2 y3 B xl z3 x2 zl x3 zl z1 x3 z2 x1 C xl y2 x2 y3 yl x3 yl x2 y2 x3 D xl y2 z3 x2 y3 z1 yl z2 x3 zl y xD 0 asin B sqrt A A B B C C xD 1 asin A sqrt A A B B C C xD 2 C sqrt A A B B C C Output positionX 0 xD 0 101 A B C D betaPrime z syi Z3 x2 y3 xl 2 43 z2 y3 x1 z3 yl x2 positiony 0 xD 1 positionz 0 xD 2 wallCollisionHapticController Block Discrete update double forceAngle int wallState forceAngle acos commandForceX 0 forcexX 0 commandForceyY 0 forceY 0 pow force X 0 forceX 0 forceY 0 forceY 0 0 5 pow commandForceX 0 commandF orceX 0 c
56. f state torque from about 10 of the full torque range down to about 1 This result was encouraging but had limitations It is nearly impossible to implement this approach in applications where the MR brake needs to apply variable torque as an actuator In the duration of the reverse pulse the brakes cannot be engaged Tasks such as wall following become very difficult since they require continuous brake control 6 3 2 Virtual Wall Following Wall following experiments were conducted to analyze the haptic arm s ability to simulate moving over a virtual wall surface A virtual environment containing flat walls were created using H3DAPI The experiment consisted of dragging the tracker virtual arm tip along the virtual wall surface 66 Because of the force approximation in the control system the haptic arm showed varying wall following performance at different locations in its work volume If the mechanism was at a configuration such that the force vector created by the haptic device could be rendered normal to the virtual wall surface then frictionless walls could be rendered However if the mechanism was in a configuration such that the force vector contained tangential and normal components then the reality of the wall following task suffered This was because of the passive MR brakes on the haptic arm which limited the direction of the force vector to certain angles according to the wall and the user force vector directions In order
57. first rotated away from position zero through approximately 0 9 radians Then it is rotated back towards the virtual wall for collision simulation 48 6 1 3 Damping Experiment A servo amplifier DRIOOEE20A8BDC QDI from Advanced Motion Controls 55 was added to the setup This enabled us to supply variable current to the brake Closed loop PI current control was performed by the servo amplifier The purpose of the experiment was to see how well the MR brake could represent a virtual damper Relationship between velocity and torque was denoted as T b 0 6 1 where damping ratio b was chosen as 1 Nm s rad 6 4 2 o 2 m o 3 a 4 6 1 L 1 1 1 6 4 2 0 2 4 6 Velocity rad s ew E o 0 c o 5 0 1 2 3 4 Time s 5 Ke gt 0 o 9 2s 0 1 2 3 4 Time s Figure 6 5 MR brake as a damper in haptics 49 6 1 4 Coulomb Friction Experiment We also explored the Coulomb friction characteristics of the MR brake Karnopp model was used to represent Coulomb friction due to its simplicity and ease of implementation 56 This model uses a velocity dead band to define static friction range instead of using absolute zero velocity which is very difficult to obtain with digital systems due to discretization The model can be represented as _ nana 1 sgn Iob A friction max T ric T pisa o k AQ 6 2 where a 0 5 rad s velocity deadband Aw
58. gent Material Systems and Structures 10 9 714 717 79 39 40 41 42 43 44 45 46 47 48 Senkal D and Gurocak H B 2009 Compact MR brake with serpentine flux path for haptics applications World Haptics 2009 Third Joint EuroHaptics conference and Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems Salt Lake City Utah 91 96 Blake J and Gurocak H B 2009 Haptic glove with MR brakes for virtual reality IEEE ASME Transactions on Mechatronics In press H3DAPI Sensegraphics AB Available Online http www sensegraphics com Accessed October 2009 Ruspini D C Kolarov K and Khatib O 1997 The haptic display of complex graphical environments Proceedings of the 24th Annual Conference on Computer Graphics and interactive Techniques ACM Press Addison Wesley Publishing Co New York NY 345 352 Cho C Kim M Song J B Direct control of a passive haptic device based on passive force manipulability ellipsoid analysis International Journal of Control Automation and Systems 2 2 Pages 238 246 2004 Cho C Song J B Kim M Hwang C S Energy based control of a passive haptic device Robotics and Automation 2004 Proceedings ICRA 04 2004 IEEE International Conference on vol 1 no pp 292 297 Vol 1 26 April 1 May 2004 Infolytica Corp Online Available http www infolytica com Accessed October 2009 Lor
59. ges 1121 1136 November 2006 Liu B Li WH Kosasih P B and Zhang X Z Development of an MR brake based haptic device In Smart Materials and Structures vol 15 no 6 pages 1960 1966 2006 Web3D Consortium Online Available http www web3d org Accessed October 2009 81 APPENDIX A ASSEMBLY DRAWINGS 82 6002 9 01 lbq UOIJISS SSOID 0000 oyuas xnuoq Aq UMPA Z 9 DDS Kusisalun 9404S uojBulusb M y 83 m n oom e pig YW Sullu di s 7 R p s 0099 9 1 u sil ni 0SSTY Z 1 31VOS els 000960 V V NOILO3S eee e O IO O O ry A 4 2 2 5 6002 9 01 910a ioguas xnuoq Aq uMDJd 2 1 lO AjISIOAIUN IDJS uo BuIUSD A Y PINSHdS C 1 31VOS V V NOILOAS 84 600Z 9 01 910a MIA 1UOJJ OD USS xnuoq Aq UMDIG ALISIOAIUN IDIS UOJBUIUSO M IE wiy 2uydpH G lD S 85 6002 9 01 910a MOSIA AOD OD USS xnuoq Aq UMDIG ALISIOAIUN IDIS UOJBUIUSO M IE wiy 2uydpH G lD S 0000 0 86 6002 9 01 91oq MIA JUBI 8 497 oyuas ynioQ
60. he brake to a haptics environment 6 2 1 Braking Torque In this experiment the goal was to determine the braking torque as a function of coil current The current on the coil was increased by 0 1 A starting from zero to 1 5 A Then the current was decreased using the same step size The data from each step was taken in one minute intervals to achieve consistency in readings As shown in Figure 6 8 the minimum off state and maximum torque were found as 0 1 Nm and 3 7 Nm respectively This gives a dynamic range of about 31 dB Torque N m E Un b Un Un 0 0 5 l 1 Un Current Amps Figure 6 8 Hysteretic braking torque of spherical MR brake versus current Light gray zero to 1 5 A dark gray 1 5A to zero 99 6 2 2 Wall Collision A virtual wall was placed into the simulation environment The purpose of this experiment was to assess how well the brake could simulate collision with a virtual surface wall A command force vector comes from the virtual environment and is normal to the wall surface An input vector comes from the force applied to the joystick handle by the user indicating the intended motion The control algorithm compares these vectors If the dot product of the vectors is negative this means the user is trying to penetrate the wall further and hence the brake is engaged Furthermore the controller implements a demagnetization algorithm to flush out the residual magnetic field in the
61. ical brake 27 3 2 1 1 Optical Position Measurement System 28 32 52 Force Measurement A 31 3 2 1 3 Haptic Rendering and Control Architecture 31 4 PASSIVE HAPTIC INTERFACE FOR DENTAL IMPLANT SURGERY 34 4 1 Passive Haptic Interface Design 34 NN EE 34 4 3 Prototype Kanten EEN 37 4 91 End Bee 39 9 SOFTWARE DEVELOPMENT Anora 40 ST Control System garen dp 40 JLL Hiesh PeyelControllere EE 41 5 1 2 Low Level ET 41 6 EXPERIMENTS AND RESULTS a 13 6 1 Rotary MR Brake Experiments a 44 11 Brakin Ke EE 45 6 2 WAIPGOIDSIOI 45 6 1 3 Dampine Et 49 6 1 4 Coulomb Friction Rxpermenmt enen eee 50 MA dE ig e s pi dan maa abla 51 Galabe DISCUSSION prapa 25 th 52 6 2 Spherical MR Brake 55 Ga Braking TORQUE nani 55 0 2 2 A ase Teale hh he de at 56 02 5 Transient Response unin NG 58 6 2 4 Damping 5millallOfi TEE 59 6 2 5 Coulomb Friction Simulation eee 60 6 2 6 Virtual Environment Simulation 61 021 DISCUSSION net s x aaa ai l 62 6 3 Passive Haptic Interface 5 65 vi
62. ime s Time s Figure 6 11 Viscosity simulation with the MR spherical brake as a damper top Breakdown of torque left and angular velocity around y axis right components of the viscosity simulation 6 2 5 Coulomb Friction Simulation The same setup as in the viscosity experiment was used to explore the ability of the brake to simulate Coulomb friction Figure 6 12 Karnopp model was used to represent Coulomb friction 56 Velocity deadband Aw was selected as 0 3 rad s and static friction torque Tstatic was set at 2 5 Nm The dynamic friction torque Taynamic was set at 0 5 Nm 60 N N Torque N m o Torque N m e Time s Time s 4922 Figure 6 12 Spherical MR brake simulating Coulomb friction along x axis left desired Coulomb friction right 6 2 6 Virtual Environment Simulation Virtual environment simulations were conducted to verify the applicabilty of the MR spherical brake as a purely passive device The objective of the experiment was to see the applicability of existing penalty based haptic rendering algorithms to a purely passive haptic device A virtual environment containing a manual gear shifter was constructed using X3D 58 Graphical and haptic rendering of the environment was performed by H3DAPI User s goal was to go through each gear and then return to the starting point which is the straight up position of the joystick The trac
63. ing surface areas on both the forward and return flux paths will allow equal flux densities on both sides In other words the forward and return cross sectional areas need to be equal To find the location of the aluminum ring the areas were computed as Figure 3 6 Calculating cross sectional area for the forward and return paths J 2n r sino do D 2n 1 sino do 3 5 2nr cos 8 2nr coso l 3 6 cos SH 3 7 where P is the angle at which the aluminum ring is placed a is the angle where the MR Fluid gap ends and T is the radius of the sphere Increasing a has a positive effect on the torque output because of the increased MR fluid gap area but at the same time it is a limiting factor for the spherical MR Brake s work volume since it reduces the motion 23 range of the handle attached to the ball For this reason first a moderate socket size of 120 was selected Then fj 75 5 was computed as the location for the aluminum ring Accuracy of this method was also verified using FEM Thirteen data points inside the MR Fluid gap were taken Figure 3 7 At 1 5 Amps 1 14 0 02 Tesla was found throughout the MR Fluid gap The flux density in the gap is fairly uniform as a result of the strategic placement of the aluminum ring I 2 T T T l z ce e e lr U r 7 m CG 0 67 7 m 0 47 d 0 21 4 0 L 1 1 H 0 20 40 60 80 100 120 o
64. ker was restricted using virtual walls seen in Figure 6 13 After the experiment was completed x and y angular position readings from the low level controller was compared to the geometry of the virtual gear shifter in order to observe the amount of wall penetration 61 0 3 0 2 0 1 2 B a 3 5 0 1 A 0 2 1 Position 0 rad Figure 6 13 Virtual environment simulation for a manual gear shifter in an automobile left Recorded joystick positions as a user tries to shift gears with haptic feedback from the joystick right 6 2 7 Discussion Similar to the rotary MR brake hysteresis behavior can be observed in the torque current curves Figure 6 8 This behavior is due to the magnetization of steel elements in the MR brake 57 Magnetization in ferromagnetic elements does not relax back to zero even if the magnetic field is removed Unfortunately this behavior not only adversely affects the controllability of the MR brake but also increases the off state torque greatly In our experiments the residual magnetization kept the off state torque at 0 3 Nm after applying and removing 1 5A coil current It caused unwanted off state friction and reduced the backdrivability of the brake To overcome this problem the controller was modified At the instant the brake was turned off it was reactivated in the reverse direction with a very short impulse An impulse with amplitude of 1A and 55 ms duratio
65. l tool for force rendering to follow rigid virtual walls with the passive device Usability experiments were conducted to drill holes with haptic feedback The maximum average positioning error was 2 88 mm along the x axis The errors along the y and z axes were 1 9 mm and 1 16 mm respectively The results are on the same order of magnitude as optical tracking systems and other dental robots hence passive haptic devices can be considered a viable alternative to active servo controlled haptic devices iv TABLE OF CONTENTS Page ACKNOWLEDGEMENTS iii ABSERA E iv STOP TABLES EES viii IST OE PICURBS a usd os Rude dett i ix CHAPTER I INTRODUCTION e 1 1 1 Passive Haptic interiate o loe dad s 5 a o A O EE 8 11 eelere eege ege ga te die al 9 2 PROBLEM STATEMENT AND SCOPE OF RESEARCH 10 2 1 Development of Compact and Powerful Rotary MR Brakes 10 2 2 Development of a Spherical MR Brake as a Potential Wrist Mechanism 11 2 3 Design of the First Prototype Dental robot 12 2 4 Integration with the Virtual Reality Environment and Controller 13 3 MR DRAKE Sido 14 2d A A O A een od N Mas 14 o E EE 15 la S a Wiss uo det a ae ee R 18 9 2 Sph erical MR Drake o n ada aaa aa 20 3 2 1 Force feedback Joystick with MR spher
66. lete sphere T complete Topening 3 10 Tcomplete Can be calculated by using equation 8 with a zr to cover the whole surface area of the sphere T complete STR 3 11 The torque that needs to be subtracted due to the opening can be calculated by a double integration over the opening The moment arm can be written as r coso T sin o sin o and shear stress at any point on the opening can be o 8 written as T r sin o do do Then the double integral for the opening becomes Topening 2 e r sino cos 0 sing sing do do 3 12 Substituting into equation 9 gives Tyy T m r ee sin o y cos o sin o sin o do do 3 13 For r 20 32mm and Tyq 1 Tesla 55kPa for the fluid we used the torque values that can be exerted by the MR Spherical Brake are found as T 3 66Nm and T 3 28N m 26 Because of the r term in equation 12 braking torque scales up very well with radius Table 3 1 shows braking torques for spherical MR brakes at different sizes Table 3 1 Theoretical braking torques for different spherical MR brake sizes Radius mm T TX 5 0 06 0 05 10 0 44 0 39 20 32 3 66 3 28 30 11 79 10 54 50 54 59 48 82 1 Calculations are done at a 120 i Prototype MR Brake 3 2 1 Force feedback joystick vvith MR spherical brake The MR spherical brake was used in the design of a force feedback joystick for
67. line virtual ErrorCode disableDevice ErrorCode e HAPIHapticsDevice disableDevice if hd get hd sdisableDevice return e Register this renderer to the haptics renderer database static HapticsDeviceRegistration device registration otected Implementation of updateDeviceValues using the contained device to get the values param dv Contains values that should be updated param dt Time since last call to this function virtual void updateDeviceValues DeviceValues amp dv HAPITime dt if hd get hd supdateDeviceValues dv dt Needed to correctly calculate device velocity for devices base this on current raw device values for example FalconHapticsDevice hd current raw device values last raw device values Implementation of sendOutput calling sendOutput of the ained device param dv Contains force values to send to the haptics device param dt Time since last call to this function virtual void sendOutput DeviceOutput amp dv HAPITime dt if hd get hd gt output force output force hd gt output torque output torque hd gt sendOutput dv dt Calls initHapticsDevice of the contained device virtual bool initHapticsDevice int _thread_frequency 1000 Releases all resources allocated in initHapticsDevice virtual bool releaseHapticsDevice if hd get 124 bool b hd gt releaseHaptic
68. ling M Watzek G Bergmann H In vitro assessment of a registration protocol for image guided implant dentistry Clin Oral Implants Res 12 1 69 78 2001 Lueth T C Hein A Albrecht J Demirtas M Zachow S Heissler E Klein M Menneking H Hommel G Bier J A surgical robot system for maxillofacial surgery Industrial Electronics Society 1998 IECON 98 Proceedings of the 24th Annual Conference of the IEEE vol 4 no pp 2470 2475 vol 4 31 Aug 4 Sep 1998 Kusumoto N Sohmura T Yamada S Wakabayashi K Nakamura T Yatani H Application of virtual reality force feedback haptic device for oral implant surgery Clinical Oral Implants Research 17 6 Pages 708 713 2006 PHANTOM Sensable Technologies Inc Available Online www sensable com Accessed October 2009 Brief J Hassfeld S Redlich T et al Robot assisted insertion of dental implants a clinical evaluation Lemke HU Vannier MW Inamura K Farman A eds Computer assisted radiology and surgery Proceedings of the 14 International Congress and Exhibition 28 June 1 July 2000 San Francisco CA Amsterdam Elsevier 2000 932 937 Boesecke R Brief J Raczkowsky J Schorr O Daueber S Krempien R Treiber M Wetter T and HaBfeld S 2001 Robot sssistant for dental implantology In Proceedings of the 4th international Conference on Medical Image Computing and Computer Assisted interventi
69. lithographic templates for surgical and prosthodontic implant planning and placement Part I The Concept Journal of Prosthodontics 15 1 Pages 51 58 2006 SurgiGuide Materialize Medical Available Online http www materialise com Accessed October 2009 CompuSurge Template Implant Logic Systems Available Online http www ipr ira uka de compusurge Accessed October 2009 Simplant Materialize Medical Available Online http www materialise com Accessed October 2009 77 14 15 16 17 18 19 20 21 22 23 24 25 Klein M Abrams M Computer guided surgery utilizing a computer milled surgical template Pract Proced Aesthet Dent 2001 Mar 13 2 165 9 Sarment D P Sukovic P Clinthorne N Accuracy of implant placement with a stereolithographic surgical guide The International Journal of Oral amp Maxillofacial Implants 18 4 2003 Ewers R Truppe M Reichwein A Wagner A Computer aided navigation in dental implantology 7 years of clinical experience Journal of Oral and Maxillofacial Surgery Volume 62 Issue 3 Pages 329 334 Robodent Robodent GmbH Available Online http www robodent com Accessed October 2009 IGI Denx Image Navigation Ltd Available Online http www denx com Accessed October 2009 Birkfellner W Solar P Gahleitner A Huber K Kainberger F Kettenbach J Homolka P Diem
70. magneto rheological fluid clutch In Gregory S Agnes editor Smart Structures and Materials 2002 Damping and Isolation Proceedings of SPIE vol 4867 2002 Demersseman R Hafez M Lemaire Semail B and Clenet S Magnetorheological brake for haptic rendering In M Ferre editor Proceedings of the EuroHaptics 2008 Spain 2008 Li W H and Du H Design and experimental evaluation of magnetorheological brake In Int Journal of Manufacturing Technology vol 21 pages 508 515 2003 Nam Y J Moon Y J and Park M K Performance improvement of a rotary MR fluid actuator based on electromagnetic design In Journal of Intelligent Materials Systems and Structures vol 19 pages 695 705 June 2008 Li W H Liu B Kosasih P B and Zhang X Z 2007 A 2 DOF MR actuator joystick for virtual reality applications Sensors and Actuators A Physical 137 2 308 320 Yamaguchi Y Furusho J Kimura S and Koyanagi K 2005 Development of high performance MR actuator and its application to 2 D force display International Journal of Modern Physics B 19 7 9 1485 1491 Han Y M Kang P S Sung K G and Choi S B 2007 Force feedback control of a medical haptic master using an electrorheological fluid Journal of Intelligent Material Systems and Structures 18 12 1149 1154 Bose H and Berkemeier H J 1999 Haptic device working with an electrorheological fluid Journal of Intelli
71. n leo 94 LpIousaru 9910 4pueuiuoo s piosa 4 0 L Aessen I Fe L zionpoid indujanbio pueuiuoo x m puewwog no xoiddysoiojmjuogondeHjiduseusej xanbio puewwoo e pionpoud 94 gionpold anbio peu dur 4 b ajqe jdryoo3enbio L donesiedincs HO IUOOMPUEUILIOI jndjnQopuewwoguonewixoiddyaqo4 Qe lt lt WoISAS qng uonemmxoiddvesioj pT Z qd 108 ZAX Id WO EH 99 09puB uiuo IdyasHanbio puewwos s nbio pueluuloo Ce oio sn g A dninw XEN Ce nbio Leen senbio j1uiof O Ad un Id VGEH9W04pueWWOd jonpold IdydeH uorje WUOJSUBI 99103 d noa wy e14 puoying lt lt ZAXS9210J POH d 1no uuil peal uoidguonejueuo 4 Zanbio pueuiuoo rodeusoy yndindr ED induir me AUOHE1U LO lt 4 Z Ayooja A lt A All of A vonejndiuejyuon sog 1olp j Suonsoq 70910 Jpueuuuioo kk K ox poLledersey induigo1 moz ndujuonisog oum odesa A9910 IpUEULUIO2 X9910 jPUBLULIOI 9 9 CTC 109 Zenbio puewwoo puew woo yx nbio jpueuulio2 X8910 jPUELLUUIOI Luong OL NOHUONEJUSHO J KMotuajy paeus alum
72. n the simulation performance Delays in discrete velocity calculations sometimes caused incorrect torque outputs during acceleration and deceleration When the torque sensor was used our experiments with the virtual wall simulation produced a very crisp reaction force upon initial contact and very high rigidity 92 at the quasi static contact with the wall Figure 6 4 When the lever arm collided with the virtual wall the brake locked the position of the arm at zero radian This created a very rigid wall feeling If the user tried to push the arm into the wall the reaction torque increased sharply top vertical part of the curve in Figure 6 4b When the lever was pulled away from the wall there was about 0 5 Nm torque acting in the reverse direction as the separation occurred Then the brake turned off completely providing nearly zero resistance Most likely the time lag of the brake coupled with the delay in deactivating the brake due to the demagnetization action caused the small initial torque in the reverse direction On the other hand when the torque sensor was not used the release action from the wall had a sticky feeling Figure 6 3b This was due to the fact that the brake was controlled only with position feedback Once the wall penetration occurred the controller had no way of knowing the direction of the user input As the user tried to pull away from the wall the release torque was almost as high as the maximum torque the
73. n the design of the MR spherical brake 20 The MR spherical brake consists of four main components 1 Steel ball 2 Steel socket with an aluminum ring 3 Coil and 4 MR fluid between the ball and the socket Figure 3 5 The aluminum ring sits on the coil It extends into the MR fluid gap to prevent the magnetic circuit from shorting around the coil and to force it to go through the MR fluid gap Starting near the center of the coil the magnetic flux path jumps across the MR fluid gap into the ball Once inside the ball it continues until it passes to the other side of the aluminum ring On the top side of the aluminum ring it jumps across the MR fluid gap once again going back into the socket where it completes its loop back to the center of the coil Due to the difficulty of visualizing and calculating magnetic fields in such complex geometry we used Magnet FEM software by Infolytica Corp 45 Steel socket Steel ball Return path Figure 3 5 MR Spherical brake flux path left and FEM modeling right Our primary design goal was to develop a compact multi DOF actuator with the highest possible torque output In order to obtain this the MR fluid needs to be activated with a strong homogeneous magnetic field 21 The size of the magnetically conductive parts was one of the parameters conflicting with this requirement since reducing the cross sectional area along the path of the magnetic circuit has a negative effe
74. n was found to be enough to collapse the residual magnetic field reducing the off state torque to 0 1 Nm from 0 3 Nm 62 Spherical MR brake was also able to simulate Coulomb friction with good accuracy Figure 6 12 Maximum torque that can be applied on x and y axes was found experimentally to be 3 7 Nm FEM simulations predicted it to be 3 28 Nm The 11 4 difference is probably due to friction forces from the O Ring seal combined with the increased contact force between the ball and socket due to the present magnetic field First order system response was observed during the activation of the brake Time constant was found to be approximately 170ms regardless of the applied current At higher torques there was visible deflection on the joystick handle and it might have adversely affected the transient response of the system Virtual environment simulation using the manual gear shifter revealed that the joystick could create very realistic contact with the virtual surfaces Our experiments with the virtual wall simulation produced a very crisp reaction force upon initial contact and very high rigidity at the quasi static contact with the wall vertical part of the curve in Figure 6 9 However several artifacts on force feedback were observed due to the purely passive nature of the device The most significant was the high friction on the virtual surfaces Since passive devices can only create forces against the user s direction of
75. na oller_wrapper c To get started select MATLAB Help or Demos from t gt gt toj PashaDemagCont mexw32 fg PashaDemagCont tlc ic PashaDemagCont_wrapper c ic PashaDemagnetizationController c eg PashaDemagnetizationController tlc ic PashaDemagnetizationController wrapper c ic PashaSimpleHapticController c Eo PashaSimpleHapticController mexw32 i PashaSimpleHapticController tlc ic PashaSimpleHapticController wrapper c ie PashaSmplHapticCont c tej PashaSmplHapticCont mexw32 z PashaSmplHapticCont tlc lt I Workspace Current Directory Figure D 1 Matlab working directory and loading the model file 127 6 Compile the simulink diagram into real time code by clicking on Wincon gt Build Figure D 2 7 Make sure that Simulation No constant is set to 2 Figure D 2 File Edit View Simulation Format Tools KU Help Ded e z Zei Extemal v Eg m 3 Bes Stop New Plot Open Plot SimulationNo A LatchingControllerCommand Output 6 LatchingController PotentiometerOutput Clean PotentiometerConditioning Clean All 0 48 1 sl a2input Set VVinCon Options az 0 4857750 bsi a3input userTorque b ltorque TO4out a3 PotsVoltage2Angle TO4out Jputput 4j Jinput 0 9 ttoffset toffset commandTorqueH3DAPI iki commandTorque Input Force Approx 28 l t2ottset t2offset commandForceH3DAPI XYZ
76. nt Response The transient response was obtained using the wall collision simulation The purpose of the experiment was to find the time constant of the brake and to explore the effect of input current on the response The experiment was repeated for current levels of 0 5 1 and 1 5A Transient response was recorded after the initial contact with the wall First order system behavior was observed time constant was found as 170ms for all 3 current levels Figure 6 10 58 25 se 5 1 i 2 E 2 15 une 1 E il E ch m 0 5 N 0 5 l y ana D x 0 0 2 0 4 0 6 0 8 1 0 0 2 0 4 0 6 0 8 1 Time s Figure 6 10 Transient response of spherical MR brake left Simulated first order system responses used to estimate the time constant right 6 2 4 Damping Simulation A linear analog servo amplifier was added to the setup This enabled us to supply variable current to the brake Closed loop PI current control was performed by the servo amplifier The purpose of the experiment was to see how well the spherical MR brake could represent a virtual damper Equation 6 1 was used for denoting the relationship between angular velocity and torque Damping ratio b was chosen as 1 N m s rad and 66 5 w was the angular velocity around the y axis 59 Torque N m Torque N m Velocity rad s 3 1 L 3 n A 4 0 2 4 6 8 0 2 4 6 8 T
77. ommandForceY 0 commandForceY 0 0 5 3 141592653589793 180 if forceAngle gt 360 forceAngle 360 if forceAngle lt 360 forceAngle 360 xD 1 forceAngle xD 2 pow forceX 0 forceX 0 forceY 0 forceY 0 0 5 if abs forceAngle gt wallAngle 0 transitionAngle 0 xD 0 1 wallState 1 else if abs forceAngle gt wallAngle 0 amp amp wallAngle 0 transitionAngle 0 gt abs forceAngle xD 0 abs forceAngle wallAngle 0 transitionAngle 0 2 5 wallState 2 else xD 0 0 wallState 0 Output outCommand 0 xD 0 outAngle 0 xD 1 outForceAbs 0 xD 2 demagnetizationController1 Block Discrete update xD 6 xD 2 if Input 0 gt 0 xD 0 1 xD 4 1 xD 5 0 else if Input 0 0 amp amp xD 4 1 amp amp abs inputAngle 0 lt 45 xD 5 1 xD 4 0 xD 3 xD 2 102 if xD 5 1 xD 0 1 if xD 2 xD 3 gt demagnetDuration 0 amp amp Input 0 0 xD 5 0 xD 0 0 Output if xD 0 1 selector 0 1 else if xD 0 1 selector 0 2 else if xD 0 0 selector 0 3 velocityController Discrete update xD 0 1 velocityX 0 Output outVelocity 0 xD 0 CoulombController Block Discrete update if pow velocityX 0 velocityX 0 velocityY 0 velocityY 0 0 5 lt thre shold 0 xDT0 1 5 else xD 0 0 1
78. on October 14 17 2001 W J Niessen and M A Viergever Eds Lecture Notes In Computer Science vol 2208 Springer Verlag London 1302 1303 Senkal D and Gurocak H B 2009 Spherical brake with MR fluid as multi degree of freedom actuator for haptics Journal of Intelligent Material Systems and Structures In press 78 26 27 28 29 30 31 32 33 34 35 36 37 38 An J and Kwon D S Five bar linkage haptic device with DC motors and MR brakes In Journal of Intelligent Materials Systems and Structures 00 1 12 2008 An J and Kwon D S Haptic experimentation on a hybrid active passive force feedback device In Proceedings of the IEEE Int Conf on Robotics and Automation Washington DC May 2002 Nam Y J and Park M K A hybrid haptic device for wide ranged force reflection and improved transparency In Proceedings of the Int Conf on Control Automation and Systems Seoul Korea October 17 20 2007 Reed M and Book W Modeling and control of an improved dissipative passive haptic display In Proceedings of the Int Conf on Robotics and Automation New Orleans LA April 2004 Carlson J D Leroy D F Holzheimer J C Prindle D R and Marjoram R H Controllable Brake US Patent No 5 842 547 December 1998 Kavlicoglu B Gordaninejad F Evrensel C Cobanoglu N Liu Y and Fuchs A A high torque
79. performs the pre programmed tasks A haptic system for bone drilling has also been developed 21 This system uses a PHANToM Haptic Device 22 for virtual bone drilling The system was intended for training and not for actual surgery The average misalignment was reported to be less than 0 2 mm indicating that the system is potentially applicable to oral implant surgery A robotic assistant for dental implantology was built which used a robot arm and CT scans to hold a drill guide over a phantom jaw 23 24 Deviations of approximately 1 to 2 mm were obtained using this system 1 1 Passive Haptic Interface Due to the complexity of the anatomic structure and the procedure a surgical aid system for dental implants is highly desirable to ensure the success of the procedure Such a system can decrease the dependence on surgeon s skills and experience for implant position accuracy and increase the overall safety of the procedure The system would track the surgeon s hand piece and the patient to provide graphical user interface and haptic feedback to the surgeon in real time to guide him during the operation In this research we explored a passive haptic interface to be used in such a surgical aid system for dental implants Figure 1 4 The interface uses magnetorheological MR fluid brakes as they are inherently safe and have excellent characteristics in providing rigid interaction forces to the user Figure 1 4 Passive Haptic In
80. puted in the fluid gap The goal was to maximize the flux density in the fluid hence the braking torque while keeping the outer diameter of the brake around 65 mm The torque requirement was set at 10 Nm The magnetic coil was formed by wrapping 800 turns of 26 gauge enameled magnet wire around the spool on the rotor The MR fluid MRF 132LD was purchased from Lord Corp 46 The rotor is a solid steel part that works both as a steel core for the electromagnet and a transmission element for the torque The magnet wire is wound into the groove on the shaft The two ends of the magnet wire pass through tiny holes along the shaft axis to be connected to power supply The shear stress generated by the MR fluid is proportional to the magnetic flux through the fluid If the number of coil turns and the current are increased the flux will increase However this requires thicker wires and results in a larger coil Using more turns of a thinner wire is possible but this time the wire overheats due to the increased current The braking torque is also a function of the fluid gap the drum radius and width The smaller the gap the larger the magnetic flux in the gap since the relative permeability of the MR fluid is much smaller than that of low carbon steel Increasing the drum radius provides larger torque arm as the shear force is applied on the surface of the drum by the fluid Furthermore if the width of the drum can be increased the total surface area whe
81. r knowledge this is the first ever multi DOF spherical brake using MR fluid 2 3 Design of the First Prototype Dental Robot Our primary design goal is to provide haptic feedback to the surgeon through the hand piece Such a system does not have the usual master slave relationship of a teleoperated system In this case they are collocated since the surgeon uses the hand piece as he she normally would and haptic feedback is added to it By eliminating the need for a programmable robotic manipulator we aimed to obtain a high level of transparency which is much desired in surgical procedures For such a system to be useful in a dental implant surgery the system must be lightweight As the surgeon has to work inside the patient s mouth cavity he she has very limited reach A bulky design would seriously deteriorate the surgeon s performance The system must also be safe Due to the nature of the operation any malfunction in the surgical aid system might result in catastrophic results With these design requirements in mind a passive haptic interface with MR brakes has been designed The MR brakes have very high torque to size ratio 25 39 40 hence they are a very good choice for a lightweight design Since MR brakes are 12 passive devices they are inherently safe They cannot add energy into the system but can only dissipate it They have excellent wall collision characteristics enabling near rigid interaction forces to be delivere
82. rated four AC motors with a spherical ER joint at the center The spherical joint assembly had an estimated diameter of 110 mm When the motors were included the system took up about 45 cm x 45 cm area A complex controller was implemented resulting in about 7 N force output on the joystick handle from the spherical joint The second device consisted of a metal sphere which was concentrically mounted in a metal half sphere 38 The gap between them was filled with ER fluid The spherical joint had 102 mm diameter At 2 8 kV mm electric field strength the device produced 1 2 Nm output torque CHAPTER 2 PROBLEM STATEMENT AND SCOPE OF RESEARCH The long term goal is to develop a dental robot to assist in oral implant surgery Development of a first prototype robot as an initial step towards this goal is the basis of this research The objective is to design a lightweight robot with passive actuators to assess the advantages and limitations of such a design The research contains four phases 2 1 Development of Compact and Powerful Rotary MR Brakes A dental robot that would be placed in a dentist s office needs to be lightweight and strong as the surgeon already needs to work in a very limited work volume such as the patient s mouth cavity Actuators are one of the primary components that affect the size of arobot arm Traditionally DC motors are used for controlling a haptic arm However DC motors usually have rather small torque to size
83. re the shear stress is applied will increase Consequently to increase the braking torque the fluid gap must be minimized while the number of coil turns current drum radius and width must be maximized 17 After much iteration the optimal design had a rotor with 52 mm radius 47 mm length and 0 25 mm fluid gap producing an average of 1 03 Tesla flux in the MR fluid Figure 3 3 Based on the manufacturer s specifications for the MRF 132LD fluid the shear stress corresponding to the 1 03 Tesla flux was 55 kPa 46 With these dimensions and the shear stress value the maximum braking torque output for an input of 1 5A current was calculated as 10 83 Nm using equation 2 The Coulomb friction was ignored since the design incorporates ball bearings reducing the mechanical friction between the shaft and its housing down to negligible levels This assumption was later validated in experiments Shaded Plot IBI 1 2 25764 1 80611 1 35458 0 803057 0 451529 Figure 3 3 Quarter sectional view of the MR Brake showing the serpentine magnetic flux using FEM analysis 3 1 2 Ferro fluidic Sealing MR brakes usually employ an O ring between the rotor and stator to seal the fluid in Although this is an effective mechanism to prevent fluid leakage the O Rings increase the off state friction of the MR brake The off state friction is an important parameter for haptic applications Furthermore the MR fluid has been shown to be very 1
84. rite API inputs DWORD api integer timeout stringstream s s lt lt Warning Failed to send outputs to the Wincon haptic device setErrorMsg s str fendif HAVE WINCONAPI I B 3 2 WinconHapticsDevice h if defined WINCONHAPTICSDEVICE define _ WINCONHAPTICSDEVICE H _ include HAPI HAPIHapticsDevice h if defined HAVE WINCONAPI include tchar h include math h include WinConInterface h Wincon shared memory namespace HAPI 119 class HAPI API WinconHapticsDevice public HAPIHapticsDevice public IT Constructor WinconHapticsDevice device id 1 Destructor virtual WinconHapticsDevice Return the Wincon device device_id for this device inline int get device id return device_id Register this device to the haptics device database static HapticsDeviceRegistration device_registration Schedule the haptic device to be calibrated if the flag argument is true Do not schedule the haptic device to be calibrated if the flag argument 3g false void schedule calibration bool calibrate Send the calibration flag to the haptic device void send calibration If scheduled run or skip the haptic device calibration procedure i e reset of the encoders Block wait until completed Optional timeout argument void do_calibration int timeout
85. sDevice hd reset NULL thread NULL return b return true The haptics device actually used auto_ptr lt HAPIHapticsDevice gt hd r tendif 125 SERVO AMPLIFIER CIRCUITS APPENDIX C 3 Solenoid R 1kQ bar OR Ciorokc c 4229901904 QUOO Gere cc lt lt ce Ks c 62 se oo nde I HUN gt A el ELSE ve pe of TE eee la eee T sjelalalalejelejajele Bl Ali E 126 APPENDIX D USER MANUAL 1 Turn on the power supplies make sure that power supply for the potentiometers is set to 5V and the power supply for the MR brake servo amplifiers is set to 18V respectively 2 Switch the output of the power supplies to on by pushing the Output button on each power supply 3 Start Matlab 2006b by clicking on Start gt All Programs gt Matlab gt R2006b 4 Change the Matlab working directory to VPasha which contains the model files Figure D 1 5 Load the simulink model file Pasha01 mdl by double clicking on the file name Figure D 1 MATLAB 7 3 0 R2006b File Edit View Debug Desktop Window Help Dg Ba A SW 2 2 Curent Directory Pasha Shortcuts 2 Howto Add 2 What s New Command Window e 2 All Files This is a Classroom License for instructional us a 7 ie Research and commercial use is prohibited ic LatchingController2_wrapper c c ki
86. sing teeth has a chance to gain the full functionality of his teeth without having to sacrifice aesthetics Figure 1 1 X Ray image of two dental implants A successful procedure results in osseointegration of the implant and an acceptable prosthodentic outcome 1 To achieve these outcomes the implant must not damage critical structures namely mandibular nerve in the lower jaw Figure 1 2 and scheiderian membrane of the maxillary sinus in the upper jaw When implants are placed too close to the mandibular nerve permanent nerve damage in different peripheral nerves may occur 2 3 This is a very serious concern as it can lead to permanent damage in the nerves controlling the lips It can also result in long lasting paraesthesia tingling pricking and numbness feeling 2 6 e Maxillary sinus First and second superior pre molars Superior canines Lateral and medial incisors First and sec ond inferior premolars Mental fora men Mandibular canal i Inferior molars y Figure 1 2 Mandibular canal with Mandibular Nerve 7 The placement of the implant must also allow enough bone structure at the bottom and sides of it for proper support 8 After the implant is in place a crown is mounted on it during prosthetic treatment to achieve the desired aesthetic affect 9 If the implant is not accurately placed then the crown cannot be aligned properly Over time the implant and th
87. sor Position sensor Figure 4 1 Passive haptic interface 4 2 Balancing The weight of the arm should not be felt by the surgeon to ensure transparency of the system Therefore the arm was made out of lightweight components Furthermore 34 its weight was compensated Since there are no active actuators in the system active weight compensation is not possible For this reason static mass balancing was done by using balancing weights and by strategically placing MR brakes on the arm geometry The brakes were positioned close to the center of rotation of joints 2 and 3 in order to minimize the inertia of the system Figure 4 2 Brake 3 at J3 together with balancing weight A ma were used to balance the arm around joint 2 J2 Balancing weight B mj was used to balance the arm around joint 3 J3 Torque applied to is transferred to J3 by mechanical links Hence it can be assumed that joints J4 and J3 are overlapping and the link between m and m is connected to the link between m and mg Using this principle the only weight that tries to rotate the arm in the clockwise direction is ms The rest of the weights try to rotate the arm in the counter clockwise direction Figure 4 2 Mass balancing 35 Since the weight of the hand piece is relatively small mass balancing around the un actuated joints Js and Je was neglected Also as the first joint axis is in the direction of gravity only joints 2
88. st static int api integer timeout int 1 api update dt ms Nincon Haptic API data arrays contain the Wincon Haptic API inputs double write API inputs shmem API write num doubles contain the Wincon Haptic API outputs double API outputs read shmem API read num doubles r tendif HAV E _WINCONAPI tendif WINCONHAPTICSDEVICE H B 3 3 AnyHapticsDevice cpp include HAPI AnyHapticsDevice h include sstream include lt H3DUtil DynamicLibrary h gt using namespace HAPI HAPIHapticsDevice HapticsDeviceRegistration AnyHapticsDevice device registration Any amp newlInstance AnyHapticsDevice gt list lt string gt bool AnyHapticsDevice initHapticsDevice int _thread_frequency hd reset NULL for list lt HapticsDeviceRegistration gt iterator i registered devices begin i 1 registered devices end i if i name Any ifdef WIN32 need to go through list of libs to see if it is even possible to try to initialize the device cout lt lt Devices found i name An bool all libs ok true for list string gt iterator j i libs to support begin j i libs to support end j if H3DUtil DynamicLibrary load j NULL all_libs_ok true cout lt lt Haptic Device library failed Xn 122 break if all_libs_ok tendif HAPIHaptics
89. ted as a Bingham plastic having variable yield strength 33 The flow is governed by r r B eg don where the first term is the dynamic yield stress as a function of the magnetic flux The second term is the shear strain rate with wis the angular velocity r is the radial position 7 1s the viscous friction coefficient and h is the fluid gap In practical applications a small torque due to friction e g due to the seals also exists as a third component in the total braking torque The significant portion of the braking torque is from the dynamic yield stress acting on the outer surface of the rotor Figure 3 1 In haptics applications the brakes rotate slowly Hence the second term in Equation 1 is ignored The total braking torque can then be written as m 2 T 2z r L T ya B T coulomb 3 2 14 where Za B is the MR fluid shear stress as a function of magnetic flux and Tcoulomb is the mechanical friction While the MR brakes are widely used in a variety of applications from exercise equipment to automobiles they are usually too bulky to be effectively used in haptics applications Also the off state friction is an issue in their usage in haptics applications Figure 3 1 Braking torque in rotary MR brakes is a function of the dynamic shear stress on the rotor controlled by the magnetic flux In the existing designs only a limited area of MR fluid in the gap can be kept at the required magnetic field strength
90. tegrated with haptic devices typically run two processes The first process involves collision detection haptic rendering and updating the graphics in the virtual world with about 15 30 frames per second The second process is the control loop of the haptic device which usually runs at 1000 Hz As shown in Figure 3 12 we implemented a two layer control architecture consisting of a low level and high level controller 31 mus Positioning Spherical MR Force feedback joystick System Brake PCI MIO 16E 4 Quanser Q4 HIL Board u Hardware Interface Transformation Triangulation Hape Low level controller Controller Ji Haptic Renderer HAPI i High level controller Figure 3 12 Control system architecture The low level controller is to control the haptic device It uses a Q4 hardware in the loop card by Quanser Inc and a PCI MIO 16E 4 data acquisition card by National Instruments The control algorithm was implemented using Simulink by Mathworks Inc 49 along with the WinCon software which enables real time code generation from Matlab Simulink diagrams The Q4 handles signals coming from the optical sensors and the command signal going out to the spherical MR brake The PCI MIO 16E 4 handles the analog signals coming from the load cell The high level controller is for the virtual environment We used H3DAPI which is an open source haptics package by SenseGraphics AB 41 The H3DAPI uses OpenG
91. terface We developed a new rotary MR brake as the actuator for the dental haptic interface Figure 1 5 The new brake uses a serpentine magnetic flux path which leads to a more compact brake design Our prototype brake has 63 5 mm diameter and 10 9 Nm torque at 1 5 A current input Another contribution of the research is a ferro fluidic seal In general MR brakes use O rings to prevent leakage of the fluid The O Ring increases the off state friction of the brake In our design we used a ferro fluidic sealing technique which reduced the off state torque and sealed the fluid Figure 1 5 Experiment setup for rotary MR brake All MR actuators are single degree of freedom DOF The wrist joint of the passive haptic interface is a 3 DOF spherical joint with individual yaw pitch and roll axis In order to create a 3 DOF spherical joint three rotational MR brakes are needed in a gimbal arrangement In this configuration the resulting mechanism is usually rather large In this research we explored design of a MR spherical brake as a multi DOF actuator Figure 1 6 Unlike the single DOF brakes the spherical brake allows motion about any arbitrary axes When it is activated it can restrict or lock all three DOFs simultaneously To the best of our knowledge our design is the first ever multi DOF spherical brake using MR fluid 25 Figure 1 6 Spherical MR brake The MR spherical brake has a diameter of 76 2 mm and can apply up to 3 7
92. the coil was increased by 0 1 A starting from zero to 1 5 A Then the current was decreased using the same step size The data from each step was taken in 3 minute intervals to achieve consistency in readings Figure 6 2 shows torque current curve for the full range of the braking torque The minimum off state and maximum torque were found as 0 08 Nm and 10 9 Nm respectively This gives a dynamic range of about 43 dB 12 10 6 Torque N m 0 0 5 1 1 5 Current A Figure 6 2 Braking torque of rotary MR brake versus current 6 1 2 Wall Collision The experimental setup was converted into a 1 DOF haptic device by attaching a lever arm to the MR Brake The purpose of this experiment was to observe how well the brake could simulate a collision with a virtual wall surface The control loop was running at 1000 Hz with 1 A current for the brake during collision The current was 45 turned on fully when collision occurred The experiment was first conducted in two modes 1 without the torque sensor and 2 with the torque sensor to see the effect on the haptic behavior Algorithm for the wall collision experiment with the torque sensor is as follows while simulation is running if position is inside the wall and torque is towards the wall activate brake in forward direction else if readyToDemagnetize is true and brake is active reset counter set readyToDemagnetize to false activate brake in reverse direction else deac
93. theta2 cos theta3 cos theta2 sin theta3 a3 sin theta2 a2 dl 04131101 0 04131111 0 04131121 0 04131131 1 jacobian 0 0 sin thetal sin theta2 sin theta3 a3 sin thetal cos theta2 a2 cos thetal d3 jacobian 0 1 sin thetal cos theta2 cos theta3 cos thetal sin theta2 cos theta3 113 cos theta2 sin theta3 a3 sin theta2 a2 jacobian 0 2 cos thetal sin theta2 cos theta3 cos theta2 sin theta3 a3 jacobian 1 0 cos thetal cos theta2 cos theta3 cos thetal sin theta2 sin theta3 a3 cos thetal cos theta2 a2 sin thetal d3 jacobian 1 1 cos theta2 sin jacobian 1 2 cos theta2 sin jacobian 2 0 T VU sin thetal thetal d3 cos thetal 77 77 57 jacobian 2 2 sin thetal sin thetal cos theta2 cos theta3 sin thetal sin theta2 sin theta3 a3 sin thetal sin theta2 cos theta3 theta3 a3 sin theta2 a2 sin thetal sin theta2 cos theta3 theta3 a3 0 sin thetal sin thetal cos theta2 cos theta3 theta2 sin theta3 a3 sin thetal cos theta2 a2 cos thetal cos thetal cos theta2 cos theta3 ll l Il 5 Il cos thetal cos thetal cos theta2 cos theta3 cos thetal sin theta2 sin theta3 a3 ESTO for i 0 i lt 3 i for 3 0 3 J
94. tic devices can be viable alternatives to active haptic devices During the haptic drilling experiments deflection caused positioning errors despite the rigid construction of the haptic arm This is a challenging issue since the arm needs to be lightweight and the application requires very strict positioning accuracy In the future we plan to explore options for closed loop position control using optical magnetic trackers and or a mathematical model of the deflection in order to compensate for the positioning error Passive haptic devices have advantages over active devices in terms of stability rigid wall collision and inherent safety They suffer from lack of pull back action on the virtual walls unsmooth behavior at the wall surfaces and inability to compensate for positioning errors All of these shortcomings can be overcome by designing devices that use hybrid passive active actuators 26 27 56 76 1 2 3 4 5 6 7 8 9 10 11 12 13 BIBLIOGRAPHY Fortin T Isidori M Blanchet E Perriat M Bouchet H and Coudert J L An image guided system Drilled surgical template and trephine guide pin to make treatment of completely edentulous patients easier A clinical report on immediate loading Clinical Implant Dentistry and Related Research 6 2 Pages 11 119 2004 Bartling R Freeman K amp Kraut R A The incidence of altered sensation of the mental nerve aft
95. tion of the handle about its own axis This would require an additional sensor such as an absolute encoder which was not implemented in this prototype due to the intended joystick application for virtual reality 30 3 2 1 2 Force Measurement When passive actuators like the MR spherical brake are used in haptics applications it is necessary to measure the forces applied by the user in addition to position measurements to control the behavior of the device If only position is measured then the so called sticky wall 48 situation occurs where the joystick will not release the brake as the user tries to pull away from a collision with a virtual object We built a simple load cell with two sets of strain gage full bridges to measure the forces applied by the user on the handle Although the spherical MR brake is able to generate moments in all three DOF in this study we concentrated only on measurement of the user forces in x and y directions For the intended purpose of the prototype as a haptic joystick the moment around the handle was neglected since it was not needed The load cell was made of aluminum and its design was optimized using finite element analysis The strain gages were connected to strain gage amplifiers 5B38 05 by Analog Devices The circuitry was interfaced to a data acquisition card PCI MIO 16E 4 by National Instruments 3 2 1 3 Haptic Rendering and Control Architecture Virtual environments in
96. tivate brake end if if counter gt demagnetizationDutation set readyToDemagnetize to true deactivate brake end if end while 46 The results of virtual wall simulation without the torque sensor are presented in Figure 6 3a and 6 3b 0 2 4 6 10 Time s 5 r r E e I 2 Sh 4 o E 10 L L l l 2 4 6 8 10 Time s 2 r 0 a Leg gt Po Y 1 2 gt 4 0 2 4 6 8 10 Time s a 5 E z 0 d g 5 o 10 L l L 1 L 0 0 2 0 4 0 6 0 8 1 Position rad b Figure 6 3 Simulation of collision with a virtual wall located at position zero without using the torque sensor a Input current torque and velocity b Virtual wall at position 0 The lever arm is first rotated away from position zero through approximately 0 95 radians Then it is rotated back towards the virtual wall for collision simulation 47 The results of virtual wall simulation with the torque sensor are presented in Figure 6 4a and 6 4b Torque N m Current Amps Velocity m s Torque N m 2 4 6 8 10 Time s 4 6 8 10 Time s 4 6 8 10 Time s a una 1 0 2 0 4 0 6 0 8 1 Position rad b Figure 6 4 Simulation of collision with a virtual wall located at position zero with the torque sensor a Input current torque and velocity b Virtual wall at position 0 The lever arm is
97. to activate the fluid Also since magnetic field strength is more or less inversely proportional to flux path cross sectional area the maximum MR fluid gap that can be actuated is limited As a result the only options to obtain high levels of torque are to increase the brake radius coil windings and current All of these options lead to a bulky design due to the size of the coil and the disk or drum 3 1 1 Serpentine Flux Path To design an MR brake with higher torque without increasing the size of the brake more surface area of the MR fluid must be activated by the magnetic flux We 15 achieved this by creating a serpentine flux path By strategically placing magnetically conductive 1018 steel rings and non conductive Aluminum rings it is possible to bend the magnetic field and weave it through the MR fluid gap multiple times Figure 3 2 This led to a more compact brake design and enabled us to increase the braking torque without increasing the size of the brake Bobbin wire MR fluid gap Figure 3 2 Serpentine magnetic flux path weaving through the drum and the outer shell top It enables activation of more of the MR fluid for increased braking torque MR fluid between the outer shell and the rotor with the coil bottom 16 Due to the complex geometry we modeled the brake using the MagNet Finite Element Analysis FEA software by Infolytica Corp 45 The design was optimized based on the magnetic flux density com
98. to demonstrate this phenomenon the experiment was conducted at two different zones inside the work volume In one zone the system could display smooth walls while in the other it could not 6 3 2 1 Smooth Wall Display A vertical virtual wall with k 0 5 N mm was created on y 0 8 m plane The y and z coordinates of the wall following task can be seen in Figure 6 15 67 0 1 T T T O 08 0 06 0 04 M PAS VP APRIL Aen J wav t 002 Or N 0 02 YA VA TL NP A 004 Z position m MARA s 0 06 0 084 01 1 L 1 1 1 1 096 0 94 092 09 088 086 084 0 82 Y position m a b Figure 6 15 Smooth wall display 6 3 2 2 Unsmooth Wall Display A horizontal virtual wall with k 0 5 N mm was created on z 0 plane The y and z coordinates of the wall following task can be seen in Figure 6 16 0 1 T 1 R I 0 056 i 4 kodu a 5 H SEE N by mv Jt 24 ETAT A y 0 05 L 0 6 0 65 0 7 0 75 08 085 F Y position m a b Figure 6 16 Unsmooth wall display 68 Figures 6 15b and 6 16b reveal a significant difference in the rendered wall quality Figure 6 15b shows a smooth wall surface albeit slightly curved whereas Figure 6 16b shows a jagged wall Both of these shapes are the result of the force approximation algorithm For the smooth wall
99. torsion of the mechanism out of the plane of the parallel links in the mechanism This resulted in the highest average error among all three axes 2 88mm Another cause for the deviation although not as significant as the first cause was the lack of pull back action on the virtual walls This is possible with haptic devices with active actuators such as motors The passive actuators can only create resistance forces If the user penetrated into the wall by pressing too hard the system had no way of compensating for the error since it could not pull back the tip A final cause was the resistance force created by the foam due to the penetration of the drill bit Some users mistook the haptic feedback for limiting the z depth as the resistance force of the foam and continued to press down looking for the haptic feedback 72 to stop When the manipulator and the haptic master are the same device haptic feedback forces and reaction forces from the environment are bound to overlap This kind of effect would be more significant when drilling a more rigid object such as bone A viable solution would be to measure the reaction forces coming from the environment and make necessary compensations in the haptic force feedback accordingly 73 CHAPTER 7 CONCLUSIONS AND FUTURE RECOMMENDATIONS A novel haptic interface with MR brakes for dental implant surgery has been built As part of this research rotary MR brakes with high torque output h
100. training the real placement task was performed 3 times by each user The hole placement order was randomized to prevent any task learning There was no time constraint on the experiment In between trials the user was allowed to rest his her arm for one minute Users were instructed to drill three holes with the assistance of the haptic feedback coming from the device Figure 6 17 Hole locations were guided by virtual constraints that resembled angle brackets in the virtual environment Figure 6 18 The corner of each virtual bracket was aligned with the axis of a hole The users moved the hand piece to position the tracker virtual tip near a guide After that they pressed the tracker against both walls of the virtual bracket and felt the haptic feedback in their hand to locate the x and y positions of the hole To successfully guide the drill bit along the axis the users had to keep the tracker in contact with the bracket through haptic feedback in their hand as they moved the bit down towards the foam The z depth of the holes was fixed by a horizontal virtual wall inside the foam As the user continued to 70 penetrate the foam the arm locked up when the required depth was reached by the drill bit A total of 3 holes had to be drilled for each placement task The holes were located 50mm apart in an L shaped arrangement and the depth was limited at 4 mm Vertical constraints Tracker Reference point Hole location Figur
101. ut The ER fluids require potential differences as high as 3 kV in order to be activated whereas devices that use MR fluids have much lower voltage requirements usually in the range of a few volts This becomes even more important if the device is to be used in environments such as haptics where human interaction will be present Failure in a 3 kV circuit can be very dangerous to the user A big challenge in designing devices that use MR fluids is routing the magnetic flux path through the fluid while keeping the overall device size compact and the output torque high Both in ER and MR fluids the fields that are applied to the fluid must be perpendicular to the fluid gap This requirement is satisfied easily with ER devices as applying a potential difference between any two surfaces would automatically create electric fields perpendicular to those two surfaces hence also perpendicular to the ER fluid in the gap With MR brakes it is more difficult since a coil for an electromagnet must be housed in the brake and the resulting flux path must be guided through the fluid by carefully designing the magnetic circuit Previously we designed single DOF rotary compact and powerful MR brakes using a serpentine flux path concept 39 40 In this approach aluminum and steel rings were employed to weave the magnetic flux path through the MR fluid gap This led to activation of much more of the MR fluid in the same compact volume The same concept was adapted i

HAPTIC SURGICAL AID SYSTEM WITH MAGNETORHEOLOGICAL

Contents

Download Pdf Manuals

Related Search

Related Contents