ARTICLE International Journal of Advanced Robotic Systems
Customizable Rehabilitation Lower Limb Exoskeleton System Regular Paper
Riaan Stopforth1,*
1 Mechatronics and Robotics Research Group (MR2G) Bio-Engineering Unit, University of KwaZulu-Natal * Corresponding author E-mail:
[email protected]
Received 16 Jul 2012; Accepted 5 Sep 2012 DOI: 10.5772/53087 © 2012 Stopforth; licensee InTech. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Abstract Disabled people require assistance with the motion of their lower limbs to improve rehabilitation. Exoskeletons used for lower limb rehabilitation are highly priced and are not affordable to the lowerincome sector of the population. This paper describes an exoskeleton lower limb system that was designed keeping in mind that the cost must be as low as possible. The forward kinematic system that is used must be a simplified model to decrease computational time, yet allow the exoskeleton to be adjustable according to the patient’s leg dimensions. Keywords Lower limb exoskeleton, rehabilitation, customizable
The 21st century has seen the realization of wearable robots. From their first introduction into the industrial workplace in the 1960s (Craig, 2005), robots have developed at an incredible rate and now encompass almost every aspect of modern society. Wearable robots are defined as “a mechatronic system that is designed
around the shape and function of the human body, with segments and joints corresponding to those of the person it is externally coupled with” (Mohammed and Amirat, 2008). A bio‐mechatronic system is needed for such wearable robots, which is the integration of biology, mechanical, electronic and computer engineering, as shown in Figure 1 (Naidu et al., 2012). Due to technological developments, robotic exoskeleton systems have evolved from rudimentary prototypes with limited application to highly sophisticated devices. These systems have the ability to enhance the performance of humans and enable disabled individuals to perform actions according to the Activities of Daily Living (ADL). There are approximately 250 000 cases of spinal cord injuries per annum in the United States of America alone (Koslowski, 2009). Severe trauma to the spinal cord may result in paraplegia or tetraplegia. Paraplegia is the loss of motor function in the lower extremities, usually with retained upper limb functions. Damage to the central nervous system or spinal cord injuries may result in such a loss of upper or lower limb motor functions (Stokes, 2010). An exoskeleton structure is required for
www.intechopen.com
Int J Adv Robotic Sy,Limb 2012, Vol. 9, 152:2012 Riaan Stopforth: Customizable Rehabilitation Lower Exoskeleton System
1. Introduction
1
individuals who w have lostt their motor functions in their lower limbss. Such an injury could result in the implementation of the low wer limb exosk keleton.
Figure 1. Bio‐M Mechatronics integration concept
The concept of using an exoskeleton for protectio on or enhancementt has been aaround for hu undreds of years. y was only receently that pow However, it w wered exoskeleetons became a reaality. One of the first conttributions to early e developmentt of po owered exo oskeletons was N”. This full b “HARDIMAN body exoskeleeton was desig gned by General E Electric Co. in the 1960s, comprising an iinner and outer ex xoskeleton wh hich operated on a master/sslave control schem me (Pons, 20008). In the lasst five years there t has been major m develop pment in exo oskeleton systems (Mohammed d and Amiratt, 2008). Systtems such ass the Hybrid Assisstive Limb (H HAL), which iis now in the fifth generation, and a Berkley’ss BLEEX, disp play cutting edge modern techn nology. d by researchers r att the University of HAL‐5 was designed Tsukuba, Jap pan. The HAL L‐5 was aimed d at meeting both strength aug gmentation an nd rehabilitattion requirem ments. HAL‐5 is a full f body exosskeleton whicch is controlled by two control schemes, naamely “Bio‐cy ybernetic con ntrol” and “cyberneetic robot conttrol” (Sanaki, 2006). The forrmer control schem me utilizes ellectromyograp phy (EMG) siignal detection forr augmentatio on operation. The latter con ntrol scheme is ussed for repetittive activities or when theree are no viable EM MG signals. This draws on a databasse of predefined m motions for a sspecific operattor (Inc, 2011). r exoskeeleton which was BLEEX is a lower limb robotic y researchers at the Univerrsity of Califo ornia, developed by Berkley, in an effort to o improve the t load beaaring o operato or. BLEEX is controlled c thro ough capabilities of the a highly sen nsitive controll system whicch uses data ffrom sensors on the exoskeleto on to predict the movemen nt of here are no sensors s measu uring the operator. However, th the interactiion force beetween the operator o and the exoskeleton ((Kazerooni, 20005). 2
Int J Adv Robotic Sy, 2012, Vol. 9, 152:2012
The objectives an nd contributio on of the reseearch were to o desiign, simulatee, design aand test a lower limb b exosskeleton sysstem, which would be the initiall consstruction for future reseaarch on adva anced roboticc prossthetics. Thiss system wo ould be main nly used forr reha abilitation of the lower lim mbs by mean ns of motion n with hin the gait cycle. c Future d development and research h coulld then be bassed on this research. ntegration of the t biologicall Thiss paper invesstigates the in and mechanical systems s withiin a body, an nd that of thee bjectives and contributionss desiigned exoskelleton. The ob werre to develop aa forward kineematic model of the system m to allow a for the motions of walking. Thee exoskeleton n systtem is requireed to cost lesss than US$ 3,0 000, thereforee requ uiring a systtem that is rreliable, safe and with a a simp plified contro ol system. Sim mulations for the gait cyclee are shown and d tests are performed to t show thee relationship betw ween actuatorss and the reha abilitation of aa persson in motion.. 2. Biiological and M Mechanical In ntegration The lower limb of o the human n skeleton com mprises threee mary joints, namely n the h hip, knee and d ankle. Thee prim Deg grees Of Freeedom (DOF F) each jointt permits iss illusstrated in Figu ure 2 (Sanaki, 22006).
Figu ure 2. Lower lim mb degrees of freeedom
The research off the lower limbs initiallly neglected d stab bility and focu used mainly on n the motion o of the system.. The operator’s baalance could b be maintained d through thee a such as crutches. Th he aim of thee use of stability aids, reseearch is to rehaabilitate the m motion of a perrson’s legs. www.intechopen.com
Hip abduction/adduction and internal rotation do not play a significant role during the walking cycle (Hian Kai et al., 2009), and were omitted from the design. The design developed is seen in Figure 3, which permitted walking in a straight line. This straight line walking means that the hip, knee and ankle joints permit articulation of the limbs in the sagittal plane (Naidu et al., 2011b).
selected such that the maximum torque was met, which allows for the operator to be raised or lowered from a seated position. Electric linear actuators from Phoenix Mecano’s LZ60 range were selected as they offered high speed/load capabilities and a less bulky design than direct mounted rotational actuators. 3. Customizable Kinematic Model A kinematics analysis was undertaken for the lower limb exoskeletons. The Denavit‐Hartenberg (D‐H) convention was incorporated for assigning the reference frames. The transformation matrix shown in Equation (1), represents joint i relative to joint i‐1. The exoskeletons are rigid serial mechanisms, which allow for the end‐effecter to be represented relative to the fixed base frames (Craig, 2005).
Figure 3. Lower limb design
The ranges of motion for the joints are constrained such that hyper‐extension and hyper‐flexion do not occur. These ranges are tabulated in Table 1 (Naidu et al., 2011b). Mechanical stops at the extremities act as a failsafe in the event of an electrical or software failure from the safety switches. Lower operational limits can be entered on a graphics user interface (GUI) should a patient need rehabilitation at lower angles.
Table 1. Joint range of motion
Both the hip and knee DOF were actuated, while the ankle joint was designed to be passive. A torsion spring mounted at the ankle was used to return the foot plate to a neutral position during the swing phase of the walking cycle. Data from clinical gait analysis (Riener et al., 2002) were evaluated to determine the joint torques for the actuated DOF. For a 100 kg system, the torque requirement for hip extension was 80 Nm. The torque required for knee extension during stair climbing was 140 Nm and 50 Nm during walking. Actuators were www.intechopen.com
��� ��� ����� �� ��� ����� 0
���� ��� ����� ��� ����� 0
0 ������ ����� 0
���� ������ �� � (1) ����� �� 1
Where: ��−1=distance from ��−1 to �� about ��−1 �� =angle from ��−1 to �� about ��−1 �� =angle from ��−1 to �� about ��� �� =distance from ��−1 to �� along �� The lower limbs have identical kinematic chains, thus the fixed reference frame was defined at the hip, and the transformation matrices relating the ankle to the reference frame were found. These matrices can be seen in Equations (2) ‐ (4), which have been derived from Equation (1) (Naidu et al., 2011b).
��� ��
� ��
�1 ��1 0 0 �1 �1 0 0 � (2) ��� 0 0 1 0 0 0 0 1
� ��
� ��
1 0 ��� 0 0
0 1 0 0
�� ��� �� �� �� 0 0 0 0
0 0 1 0
�1 0 � (3) 0 1
0 0 1 0
�� 0 � (4) 0 1
The forward kinematics of the exoskeleton leg were obtained using Equation (5) (Craig, 2005). This kinematics model relates the end‐effector to the origin of the base frame, which is represented by the GH joint.
� ��
� ��� ��� � ��� �� (5)
Riaan Stopforth: Customizable Rehabilitation Lower Limb Exoskeleton System
3
3.1 Inverse Kinematic Models (Naidu et al., 2012) m existt, of Several inveerse kinematiic iterative methods which the Damped D Leaast Squares (DLS) ( metho od is superior arou und singulariities and com mplex designs. The DLS method d implements the Jacobian n of the positiional vector of thee mechanism m and is requ uired to solvee the change in ߴ (Buss, 2004). Matrix ଷܶ is in the form off Equation 1 aand the X ,Y an nd Z he base framee, are terms of the end‐effector, relative to th represented b by the first thrree rows of the last column.. The end‐effector is considered d to be the foot part off the on of the end d‐effector willl be exoskeleton. This positio n of represented by S (Equattion 6) and is a function dߴସ . ߴଵ ǡ ߴଶ ǡ ߴଷ and
ܬൌ
డௌሾሿ డణ
(7)
The change iin ߴ can be determined by Equation 8 (B Buss, 2004).
Δϑ =(inv v(J’J + λ2I)*J’)*eerror (8)
mping J’, λ and I represent thee Jacobian traanspose, dam i matrrix respectively. The dam mping factor and identity factor comp pensates the computationaal problems as a gularities (Na eet al., 2008). result of sing w its transspose The multipliication of the Jacobian with produces a square maatrix which allows for the computation of the inverrse of this multiplication. The m error value is i the differen nce between the t initial possition and the target position. This T is updateed in the iteraation. The algorith hm of the DLS D method is illustrated d in Figure 4. Thee iteration pro ocess is depen ndent on the error e value which is initially calculated. Thee angle chang ge is then iterativ vely solved and a the errorr value and joint angles are updated. u Thiss process is repeated r untill the error value iss reduced to a desired toleraance. The inverse kinematics were w derived d analytically and verified throu ugh simulation ns on Matlab® ®. The worksp pace, seen in Figurre 5, was mod delled with th he use of Matllab®, which used a random nu umber generaation method, and m ran nge of motion n for each leg. For depicts the maximum 4
Erro or = target – curr ent position
Wh hile error > value
Δ� =(inv(J’J = + λ²I))*J’)*error
ܺሺߴ ߴଵ ǥ ߴ ሻ ߴଵ ǥ ߴ ሻ (6) ܵ ൌ ܻሺߴ ܼሺߴ ߴଵ ǥ ߴ ሻ
n of small ang gle changes iss a function of the The iteration Jacobian and d results in a linear derivattion of the inv verse kinematics. n is representted by Equatiion 7 (Buss, 2004). The Jacobian This will result in an m x x n matrix; in this model m m = 3 and n = 4.
simu ulation purposses, limb lengtths were set to o L1 = 500 mm m and L2 = 430 mm ffor the thigh an nd shank respeectively.
Int J Adv Robotic Sy, 2012, Vol. 9, 152:2012
Update error
Upd date theta
Figu ure 4. Flowchartt illustrating thee iteration proceess of the DLS meth hod
The plane in Figu ure 5 represen nts the sagitta al plane, with h the anterior of th he model facin ng the positiv ve x direction.. The workspace d depicts the ran nge of motion n of the swing g leg (blue), while the stance leeg (green) is included forr ual reference only. The fulll scope of th he workspacee visu inclu w standing g udes the jointt ranges both for motion while upriight and in a sseated position n.
Figu ure 5. Range of m motion for loweer limbs
4. C Control Architeecture Instructions such h as joint chan nges or co‐orrdinate pointss n a GUI. Thee data is sen nt to Matlab® ® are performed on whiich calculates the relevant k kinematic calcculations. Thee putted joint an ngles are then sent to thee calculated or inp relevant microcon ntrollers whicch will carry out o the motorr www.intechopen.com
control operaations. The Ro obotic Commu unication Prottocol (Stopforth et al., 2011) wass used to send d data between n the devices. A PD P control system was imp plemented on n the system. A GUI G interface allows the user to controll the exoskeleton system through Matlab® ®. The kinem matic models are calculated an nd these ang gles are used d as references within w the conttrol system. Feedback F enco oders allow for thee positional co ontrol model, w which is show wn in Figure 6. Thee microcontro oller relays th he position con ntrol back to Matlaab®.
The LZ60P0150 linear actuato ors from Pho oenix Mecano o werre used in thee design of th he exoskeleton legs. Thesee actu uators operatee at 36 V and ccan exert a forrce of 2000 N.. Testts indicated th hat there is a llinear relation nship between n the actuator feed d‐rate and th he gait cycle, as shown in n Figu ure 9.
Figure 6. Conttrol architecturee of the PD contrrol implemented d in the exoskeleton n system
The microcontroller that w was used for the control of the nsory network k is the ATM Mega linear actuattors and sen 1280 on an Arduino A boarrd. Limit swittches were pllaced on the actuaators to act ass a fail safe sy ystem, should d the feedback fro om the enco oders on the actuators give inaccurate reesults. The limit switches acted as hom ming positions, wh hich reset thee counters on the encoderss and therefore prevented p t the person from inju uring themselves. The electroniic integration n of the systeem is gure 7. shown in Fig
Figu ure 8. Constructiion of the lowerr limb exoskeletton that is bein ng tested for com mfort and motio on
Figure 7. Systeem integration o of the exoskeleto on legs
5. Tests and R Results The design o of the lower lim mb exoskeleto on was kept sim mple so as to havee a platform to o perform testts and observee the interaction off the human‐m machine interaaction. The deesign of the lowerr limb exoskeeleton that was w constructeed is shown in Fig gure 8. A stand was developed to o mount the ex xoskeleton leg gs to, allowing for free motion as if the person n were walkin ng on air. This elim minated the pro oblematic areaa of instability y and injuring of an ny person duee to the exoskeeleton tipping g and falling.
The graph in Figu ure 9 shows th hat the actuato or with a feed‐‐ rate of approximaately 28 mm/s would allow the knee jointt to rotate r 90° in 1 second. Reesults shown in Figure 100 indiicate the relatiionship betweeen the actuattor length and d flexiion of the leg.
www.intechopen.com
Riaan Stopforth: Customizable Rehabilitation Lower Limb Exoskeleton System
Figu ure 9. Graph of ffeed rate (x 10‐1 m/s) vs.gait cyccle
5
Figu ure 12. Average hip flexionjointt angle within th he gait cycle
Figure 10. Graph of actuator llength (mm) vs.knee flexion
The graph in n Figure 10 sh hows that for a 120° revolu ution, the actuator m moves from 270 mm down to 118 mm, giiving the required stroke length of approximaately 150 mm.. Full motion of the actuators alllow for the fu ull gait cycle tto be completed, aas required forr rehabilitation.The final deesign of the lowerr limb exoskelleton being teested is show wn in Figure 11.
Figu ure 13. Average knee extensionjjoint angle with hin the gait cyclee
The maximum po ower of the acctuators for th he hip flexion n and knee extensions were reco orded as bein ng 110 W and d 50 W W respectively y. 6. C Conclusion
Figure 11. The lower limb exo oskeleton being tested
The control system was tested t for the weight of 1000 kg person. The walking motiion was initiaated to compleete a gait cycle within w 3 secon nds. The hip flexion and knee extensionjoin nt angles weree monitored for f the gait cy ycles, and the averrage values were w recorded d as illustrateed in Figure 12 and Figure 13 reespectively. C Constraint limiits of d for the hip p and knee joint 55° and 60°° were placed respectively. 6
Int J Adv Robotic Sy, 2012, Vol. 9, 152:2012
The lower limb p prototype exosskeleton was designed thatt wed for hip, knee and ank kle joint articu ulation in thee allow sagiittal plane. Other O DOF were elimin nated on thee exosskeleton fram me as these mo otions were used less often n for lleg rehabilitattion. The exosk keleton legs w were designed d to allow a for adjjustability an nd the rehabiilitation of a a persson with a weight w less tthan 100 kg. This weightt restrriction alloweed for a safety factor to be in ncluded. d to derive the forward d The D‐H notation was used kineematics of thee mechanism. The D‐H parrameters weree used d to derive th he relative join nt transformation matrices.. The workspace off the design w was simulated d to determinee mped Least Sq quare method d the range of mottion. The Dam ve the inversee kinematics of o the system,, wass used to solv allow wing determin ning the anglee of each joint. ound to be 288 The feed‐rate of the linear actuators was fo m/s allowing th he knee joint tto rotate 90° in n 1 second, ass mm requ uired for the g gait cycle. Forr a 120° revolu ution, a strokee leng gth of 150 mm m is requireed. The resullts that weree obta ained for walk king motion h have shown th hat the controll algo orithm and architecture hav ve caused an overshoot forr the knee joint of o 1° above the constrain nt limit (60°)) www.intechopen.com
indicated on the GUI. An undershoot was observed by the hip joint, which is possibly due to the load on the actuators that damped the chance of an overshoot. The operation of the biological leg and previous lower limb exoskeletons were researched. The mechanical properties of the biological leg were correlated to the design and development of the exoskeleton legs to allow rehabilitation in the sagittal plane. The objectives explained in the introduction were achieved. The integration of the electronic system to control and operate the mechanical system was explained in the paper. Safety implementation of the system was integrated mechanically, electronically and by means of software. The research has developed a prototype system that allows for the rehabilitation of a person’s lower limbs, which came to a total cost of under US$ 3,000.Satisfactory results were obtained to allow future work to be performed on the system. 6.1. Future Work The actuators that were used in the prototype system had a low torque and speed which could be increased to allow rehabilitation of people with greater weight. Higher torque actuators that have a low weight ratio would be more beneficial, but would increase the cost. The prototype system could be improved and expanded for different types of applications. Adaptive control architecture could be implemented into the GUI model that will take into account the weight of the person. These variables could be determined by sensors placed on the exoskeleton lower limb system. The designed lower limb exoskeleton system will allow for rehabilitation in an up‐right position. Investigation of a lower limb rehabilitation system in a seated position could be considered, with the use of an impedance control system. 7. References [1] Buss, S.R. (2004), ʺIntroduction to Inverse Kinematics with Jacobian Transpose, Pseudoinverse and Damped Least Squares methodsʺ, IEEE Journal of Robotics and Automation, 17 April 2004. [2] Craig, J.J. (2005), Introduction to Robotics ‐Mechanics and Control 3rd ed. Upper Saddle River: Pearson Prentice Hall. [3] Hian Kai, K., Missel, M., Craig, T., Pratt, J.E., Neuhaus, P.D. (2009), ʺDevelopment of the IHMC Mobility Assist Exoskeleton”, IEEE International Conference in Robotics and Automation (ICRA 2009), pp. 2556‐2562.
[4] Inc., C. (2011). “Hybrid Assistive Limb”, Available: http://www.cyberdyne.jp/english/index.html, 18 May 2011 [5] Kazerooni, H., Racine, J.‐L., Lihua, H., Steger, R. (2005), ʺOn the Control of the Berkeley Lower Extremity Exoskeleton (BLEEX),ʺ IEEE International Conference of Robotics and Automation (ICRA 2005), pp. 4353‐4360. [6] Koslowski, H.M. (2009), ʺSpinal Cord Injury: Functional Outcomes in 2009 and Beyond,ʺ Northeast Florida Medicine, vol. 60, pp. 32‐35. [7] Mohammed, S. and Amirat, Y. (2008), ʺTowards intelligent lower limb wearable robots: Challenges and perspectives ‐ State of the artʺ, IEEE International Conference on Robotics and Biomimetics, 2008, pp. 312‐317. [8] Na, M., Yang, B., and Jia, P. (2008), ʺImproved damped least squares solution with joint limits, joint weights and comfortable criteria for controlling human‐like figures.,ʺ IEEE Conference on Robotics, Automation and Mechatronics, pp. 1090‐1095. [9] Naidu, D., Stopforth, R., Bright G., Davrajh, S. (2011a), ʺA 7 DOF exoskeleton arm: Shoulder, elbow, wrist and hand mechanism for assistance to upper limb disabled individuals,ʺ AFRICON, 2011, Livingstone, Zambia; IEEE, pp. 1‐6, 13‐15 Sept. 2011 [10] Naidu, D., Cunniffe, C., Stopforth, R., Bright, G., Davrajh, S. (2011), “Upper and Lower exoskeleton limbs for Assistive and Rehabilitative Applications”, 4th Conference of Robotics and Mechatronics (RobMech), Pretoria, South Africa, November 2011 [11] Naidu, D., Stopforth R., Davrajh S., Bright G. (2012), “A Portable Passive Physiotherapeutic Exoskeleton”, International Journal of Advanced Robotic Systems, InTech, Vol 9 [12] Pons, J.L. (2008), “Wearable Robots: Biomechatronic Exoskeletons”, Chichester, West Sussex: John Wiley & Sons Ltd, 2008. [13] Riener, R., Rabuffetti, M., Frigo, C. (2002), ʺStair ascent and descent at different inclinationsʺ, Gait & Posture, vol. 15, pp. 32‐44. [14] Sankai, Y. (2006), ʺLeading Edge of Cybernics: Robot Suit HAL,ʺ International Joint Conference (SICE‐ ICASE 2006), pp. 1‐2. [15] Stopforth, R., Bright, G., Davrajh, S., Walker, A., (2011), ʺImproved communication between manufacturing robots, ʺSouth African Journal of Industrial Engineering, vol. 22, pp. 99 ‐ 107. [16] Stroke, N. I. o. N. D. a. (2010), “NINDS Brachial Plexus Injuries Information Page”, Available: http://www.ninds.nih.gov/disorders/brachial_plexus/ brachial_plexus.htm, 31 May 2010.
www.intechopen.com
Riaan Stopforth: Customizable Rehabilitation Lower Limb Exoskeleton System
7