- Open Access
Novel swing-assist un-motorized exoskeletons for gait training
© Mankala et al; licensee BioMed Central Ltd. 2009
- Received: 17 November 2008
- Accepted: 03 July 2009
- Published: 03 July 2009
Robotics is emerging as a promising tool for functional training of human movement. Much of the research in this area over the last decade has focused on upper extremity orthotic devices. Some recent commercial designs proposed for the lower extremity are powered and expensive – hence, these could have limited affordability by most clinics. In this paper, we present a novel un-motorized bilateral exoskeleton that can be used to assist in treadmill training of motor-impaired patients, such as with motor-incomplete spinal cord injury. The exoskeleton is designed such that the human leg will have a desirable swing motion, once it is strapped to the exoskeleton. Since this exoskeleton is un-motorized, it can potentially be produced cheaply and could reduce the physical demand on therapists during treadmill training.
A swing-assist bilateral exoskeleton was designed and fabricated at the University of Delaware having the following salient features: (i) The design uses torsional springs at the hip and the knee joints to assist the swing motion. The springs get charged by the treadmill during stance phase of the leg and provide propulsion forces to the leg during swing. (ii) The design of the exoskeleton uses simple dynamic models of sagittal plane walking, which are used to optimize the parameters of the springs so that the foot can clear the ground and have a desirable forward motion during walking. The bilateral exoskeleton was tested on a healthy subject during treadmill walking for a range of walking speeds between 1.0 mph and 4.0 mph. Joint encoders and interface force-torque sensors mounted on the exoskeleton were used to evaluate the effectiveness of the exoskeleton in terms of the hip and knee joint torques applied by the human during treadmill walking.
We compared two different cases. In case 1, we estimated the torque applied by the human joints when walking with the device using the joint kinematic data and interface force-torque sensors. In case 2, we calculated the required torque to perform a similar gait only using the kinematic data collected from joint motion sensors. On analysis, we found that at 2.0 mph, the device was effective in reducing the maximum hip torque requirement and the knee joint torque during the beginning of the swing. These behaviors were retained as the treadmill speed was changed between 1–4 mph. These results were remarkable considering the simplicity of the dynamic model, model uncertainty, non-ideal spring behavior, and friction in the joints. We believe that the results can be further improved in the future. Nevertheless, this promises to provide a useful and effective methodolgy for design of un-motorized exoskeletons to assist and train swing of motor-impaired patients.
- Joint Torque
- Spinal Cord Injury Patient
- Treadmill Walking
- Treadmill Training
- Human Walking
The incidence of spinal cord injury (SCI)in the United States is approximately 11,000 per year, with a prevalence of nearly 250,000 . Damage to the spinal cord often impacts walking functions. Approximately, 52% of this population has motor incomplete lesions , therefore, the potential to regain functional ambulation. Rehabilitation targets restoring these functions. Currently, therapist assisted body-weight supported treadmill training (BWSTT) is used for such patient groups. In this training, a patient walks on a motorized treadmill with a harness that partially unloads the weight of the trunk from the supporting leg, while therapists help the patient in moving the legs and trunk manually [2–4]. Clinical trials with BWSTT in iSCI patients show that it is safe and results in improvements in walking[5, 6]. Despite these benefits, clinical practice of BWSTT is limited because a number of therapists are required to manually facilitate the step training [3, 7]. The duration of such a training is often limited by the rapist fatigue.
MIME, ARM and MIT-MANUS represent early advances in robotic devices for use in upper extremity training and rehabilitation [8–10]. These devices, and a majority of newer rehabilitation machines for the upper extremity, are powered. A second group of upper extremity machines is un-motorized or passive. This group consists of gravity balancing orthoses, which are designed for people with limited strength [11–14]. These un-motorized machines provide benefits similar to motorized machines, in a restricted way, but do not require sophisticated electronics or power sources to run the machine. As a result, they can be more affordable and possibly require less oversight by trained engineering personnel in future.
Lower extremity machines are emerging in recent years for gait training, but they are still not common in rehabilitation clinics. The design of lower extremity machines is more involved compared to those for the upper extremity because issues of posture, balance, and limb movement need to be simulatneously addressed within the design. Lokomat is a motorized bilateral exoskeleton for hip and knee joints, designed for spinal cord injury patients to be used on a treadmill . Mechanized Gait Trainer (MGT) is a single degree-of-freedom powered machine that drives a foot using a crank and rocker system (Hesse and Uhlenbrock, 2000). An active leg exoskeleton (ALEX) was recently developed at the University of Delaware by the author's group which was shown to successfully alter the gait of a healthy and stroke subjects walking on a treadmill [16, 17].
Using Lokomat with body weight support, Hornby et al  and others have shown that significant improvements can be achieved in walking of patients with chronic and sub-acute SCI. However, the cost of such a device runs in several hundreds of thousands US $, which make these prohibitive for many rehabilitation facilities and unaffordable by hospitals in under-developed countries. To increase the accessibility and success of BWSTT, costs of the therapy should be minimized.
Gottschall and Kram  suggested simple, non-motorized, devices which can apply forces to assist the limb swing and propel the leg foward during walking. They applied forces using rubber bands at the foot or pelvis by a spring-loaded pulley system. Even though their swing-assist devices need further developments, their results suggest that simple devices can assist those with reduced voluntary force production, such as subjects with iSCI. The non-motorized lower extremity gravity balancing orthosis (GBO), that eliminates or reduces the effects of gravity on the joints, have been used for training studies on chronic stroke patient and yielded favorable results by the author's group [20–22].
However, the design of GBO is fundamentally different from the design philosophy of the swing-assist exoskeleton presented in this paper, as the latter is motivated from providing propulsive forces to the leg during walking. We believe that the design presented in this paper is unique since it presents a simple un-motorized bilateral exoskeleton for swing assistance. In order to scientifically design the orthosis, we use the dynamics of walking to predict and optimize the motion of a leg, once it is strapped into an orthosis. The model of the swing leg provides a framework for optimization of the parameters of the exoskeleton, which are torsion springs at the hip and the knee joint.
The organization of the paper is as follows: In Section, we describe the dynamics of the human leg during swing and provide a framework for optimizing the parameters of the exoskeleton to obtain a feasible gait. In Section, we discuss the physical design of the exoskeleton and its interface with a human subject during treadmill walking. The analysis of the data collected during treadmill walking and their interpretations are also discussed. These are followed by conclusions of the work.
Sagittal Plane Model of Human Walking
The system dynamics depends on the following quantities: m1, m2 – masses of the thigh and shank (leg + device); L1, L2 – lengths of thigh and shank segments; , – location of the center of mass of the thigh and shank (leg + device) measured from their respective joints; I1, I2 – inertia of thigh and shank (leg + device) about their center of mass. Please note that '(leg+ device)' indicates the equivalent quantity based on human leg and device parameters. Simulation results section shows how the equivalent parameters are calculated based on anthropometric data and device mass assumptions.
In our study, we have used two different models for the hip motion: (i) hip is inertially fixed, (ii) hip has only vertical motion, i.e., it is assumed to remain fixed in the horizontal direction. While more complex models could have been made to describe the human hip motion, we believe that pendular motion of the hip in the sagittal plane may be a reasonable first model. A spinal cord injury patient, by himself or herself, has very little residual motion left in the limbs and the sagittal plane motion will be the predominant motion during their treadmill training. In this paper, we only describe the second model, where the hip has only vertical motion (represented by red lines in Figure 1). We believe that this model is more realistic to capture the movement on a treadmill.
In this model, we assume that the foot of the stance leg remains in contact with the treadmill and moves along with it until the swing leg makes contact with the treadmill again. We also assume that the knee in the stance leg remains locked. With these assumptions, using the kinematic model of the stance leg, we compute the up and down motion of the hip. This motion is then used in the dynamics of the swing leg.
Equations of Motion
Note that while finding the device parameters from simulations we assume that the external torque τ i applied is zero and based on the above dynamics we find θ i (t). Whereas while analyzing the experimental results, based on the encoders data we know θ i (t). We use this information to calculate the external torque τ i , more specifically the human applied component. In the later case, external torque τ i can be treated as a summation of device interface torques τ FT (which is known as it is recorded by Force-Torque (F/T) sensors) and the human applied torque τ h . Based on the dynamic equations we can estimate human applied torque τ h .
Knee Lock and Unlock
In the above equation, the first term represents reaction torue due to gravity, the second term represents reaction torque due to torsion spring and the third represents reaction torque due to shank acceleration. Based on day to day observations of healthy subjects walking on a treadmill it is observed that the knee does not unlock until the swing leg touches the ground.
Error from the desired final configuration (not the entire gait) was taken as the objective function that the optimization process would minimize. In addition, positive ground clearance (the relative (vertical) position of the foot w.r.t. the treadmill is greater than zero) at a finite number of points during the gait was imposed as a constraint. The optimized parameters were then used to perform forward simulations of the leg. During these forward simulations, locking event was not simplified with the stiff spring but instead the exact model described in knee lock and unlock section was used. Actual values of the desired starting and final configurations are given in the simulations results section.
Device parameters are found based on the following healthy subject's biological data on whom the experimental tests were also conducted.
BodyWt = 72.6 kg
Height = 167 cm
Age = 35 yrs
Lthigh = 0.41 m
Lshank = 0.40 m
The following average anthropometric data for human leg  was used to obtain the other important parameters required for simulations.
mthigh = 0.1000 × Body Wt
mshank = 0.0465 × Body Wt
mfoot = 0.0145 × Body Wt = foot mass
Rthigh = 0.323 × Lthigh (radius of gyration of thigh)
Rshank = 0.302 × Lshank (radius of gyration of shank)
Apart from the thigh and shank mass, in this simulation, we also considered foot mass and device mass. We assumed that the mass of the thigh and shank segments of the deviceis 1 kg each and is distributed such that their center of mass and radius of gyration coincide with center of mass and radius of gyration of human thigh and shank segments repectively. Based on the anthopometric data and the device mass assumptions, the equivalent mass and center of mass parameters to be used in the simulation can be found as follows,
m1 = mthigh + mdevice_thigh
m2 = mshank + mfoot + mdevice_shank
L1 = Lthigh
L2 = Lshank
For the stance leg, we specify the symmetrically opposite initial conditions, i.e., the final configuration of swing leg is taken as the initial configuration of the stance leg and vice versa. With these system parameters and desired configurations, the optimization routine gives the design parameters as c1 = 7.9 Nm/rad, c2 = 5.3 Nm/rad, = 22°, = 0°.
Our results from the simulation resulted in a more natural human walking under the condition when the hip was allowed to move up and down, compared to the case when the hip remains inertially fixed. This is consistent with human walking, where the hip moves up and down. From the perspective of energy flow, the springs get charged during the stance phase by the treadmill and the body-weight support system which allows only a vertical motion to the hip. In swing phase, the potential energy stored in springs is converted to kinetic energy of the swing leg. Some energy flows out at the hip, working against the constraint of only vertical motion, and some energy is lost during knee and heel impact. In human walking, there is a finite-time when the leg is in double support. In this phase, both swing and stance legs are in contact with the ground. In future, if the foot is modeled as a separate limb, this double support phase of human walking can also be accounted.
Experimental Results and Discussion
A pelvic link made of aluminum is attached rigidly to the trunk belt. In order to help the pelvis remain nearly vertical during treadmill walking, a back pack frame is used. This back pack frame is rigidly connected to the pelvic link through aluminum sections. Other links in the device are the telescopic thigh and shank segments, connected successively through revolute joints. All links have slots to adjust the link lengths and match these to the human wearing it. The device thigh is connected to the human thigh with the help of a thigh brace. The device shank is connected to the human foot via a foot piece. Currently, the foot piece only allows sagittal plane ankle motion. At the device hip and knee joints, torsion springs are connected in parallel to obtain a desired stiffness and equilibrium configuration, suggested by the optimization. Encoders are mounted at all revolute joints to measure hip and knee angles. Two force-torque sensors are mounted on each leg of the exoskeleton, one sandwiched between the thigh link and the thigh brace and the other between the shank link and the foot piece. These sensors measure the forces and torques transmitted between the device and the human.
The exoskeleton was first adjusted to match the limb lengths of the subject, a 45 years healthy male of Asian origin, 70 inches tall. The subject's biological data was used to find the optimal spring parameters while walking on the treadmill at a speed of around 2.0 mph (see simulation results section). The appropriate springs were mounted on the exoskeleton. Note that in a clinical setting too, based on test subjects' biological data, device parameters can be found from the simulations. Once the desired stiffness parameters are obtained, the device joints' stiffness can be approximately adjusted based on an existing collection of springs. The equilibrium configurations of the springs can then be suitably adjusted if the parts used to mount the springs have slots or set of holes instead of a single hole that would allow only a single equilibrium configuration.
Data for a Range of Treadmill Speeds
In this paper, we presented a simple un-motorized bilateral exoskeleton for swing assistance and training of motor impaired patients. This exoskeleton is aimed at reducing the physical and financial costs associated with therapist assisted training. The device consists of two segments – thigh and shank with torsion springs at hip and knee joints. Stiffness of the springs and their equilibrium configurations were the design parameters, which were optimized based on the required performance of the exoskeleton. We modeled the human leg with two links, thigh and shank segments, moving in the sagittal plane. The foot was modeled as a point mass and the hip had the motion of an inverted pendulum. The dynamics are developed when the device is strapped to the leg. In the simulation, we observed that the device helps the leg during swing to clear the ground and go to a desired final configuration. We also performed simulations with change in leg mass to evaluate the robustness of the design to variation of system parameters. We found that the system was robust for up to 50% change in leg mass.
An exoskeleton was fabricated based on the optimized parameters from simulations. This device was tested on a healthy subject at different treadmill speeds. To show the effectiveness of the device, we compare two different cases. In case 1, we estimated the torque applied by the human joints when walking with the device using the joint kinematic data and interface force-torque sensors. In case 2, we calculated the required torque to perform a similar gait only using the kinematic data collected from joint motion sensors. On analysis, we found that at 2.0 mph, the device was effective in reducing the maximum hip torque requirement and the knee joint during the beginning of the swing. These behaviors were retained as the treadmill speed was changed between 1–4 mph. These results were remarkable considering the simplicity of the dynamic model, model uncertainty, non-ideal spring behavior, and friction in the joints. We believe that the results can be further improved in the future. Nevertheless, this promises to provide a useful and effective methodolgy for design of un-motorized exoskeletons to assist and train swing of motor-impaired patients.
We acknowledge the following sources of support for this work: NIH R24 and NIDRR Model Systems Center subcontracts from Rehabilitation Institute of Chicago, NIH National Center for Medical Rehabilitation Research under Grant HD38582. We also acknowledge Vivek Sangwan for valuable inputs.
Support of WCU (World Class University) program through the Korea Science and Engineering Foundation funded by the Ministry of Education, Science and Technology (No. R32-2008-000-10022-0) is also gratefully acknowledged.
- National Spinal Cord Injury Statistical Center: Annual Report. Birmingham, AL 2005.Google Scholar
- Barbeau H, Wainberg M, Finch L: Description and application of a system for locomotor rehabilitation. Med Biol Eng Comput 1987,25(3):341-344. 10.1007/BF02447435View ArticlePubMedGoogle Scholar
- Behrman AL, J HS: Locomotor training after human spinal cord injury: a series of case studies. Phys Ther 2000,80(7):688-700.PubMedGoogle Scholar
- Wernig A, Muller SL: Locomotion with body weight support improved walking in persons with severe spinal cord injuries. Paraplegia 1992,30(4):229-238.View ArticlePubMedGoogle Scholar
- Barbeau H, Norman K, Fung J, Visintin M, Ladouceur M: Does neurorehabilitation play a role in the recovery of walking in neurological populations? Ann N Y Acad Sci 1998, 860: 377-392. 10.1111/j.1749-6632.1998.tb09063.xView ArticlePubMedGoogle Scholar
- Barbeau H, Fung J: The role of rehabilitation in the recovery of walking in the neurological population. Curr Opin Neurol 2001,14(6):735-740. 10.1097/00019052-200112000-00009View ArticlePubMedGoogle Scholar
- Dobkin B, Apple D, Barbeau H, Basso M, Behrman A, Deforge D: Weight-supported treadmill vs over-ground training for walking after acute incomplete SCI. Neurology 2006,66(4):484-493. 10.1212/01.wnl.0000202600.72018.39PubMed CentralView ArticlePubMedGoogle Scholar
- Krebs HI, Hogan N: Robot-aided neurorehabilitation. IEEE Trans Rehabil Eng 1998, 6: 75-87. 10.1109/86.662623PubMed CentralView ArticlePubMedGoogle Scholar
- Lum PS, Burgar CG, Shor PC, Majmundar M, Loos M: Robot-assisted movement training compared with conventional therapy techniques for the rehabilitation of upper limb motor function following stroke. Arch Phys Med Rehabil 2002, 83: 952. 10.1053/apmr.2001.33101View ArticlePubMedGoogle Scholar
- Reinkensmeyer DJ, Kahn LE, Averbuch M, McKenna-Cole AN, Schmit BD, Rymer WZ: Understanding and treating arm movement impairment after chronic brain injury: Progress with the arm guide. J Rehabil Res Dev. 2000,37(6):653-662.PubMedGoogle Scholar
- Rahman T, Ramanathan R, Stroud S, Sample W, Seliktar R, Harwin W, Alexander M, Scavina M: Towards the control of a powered orthosis for people with muscular dystrophy. Proceedings of the Institution of Mechanical Engineers, Part H: Journal of Engineering in Medicine 2001,215(3):267-274. 10.1243/0954411011535858View ArticleGoogle Scholar
- Agrawal SK, Gardener G, Pledgie S: Design and Fabrication of a Gravity Balanced Planar Mechanism Using Auxiliary Parallelograms. Journal of Mechanical Design, Transactions of the ASME 2001,123(4):525-528. 10.1115/1.1413771View ArticleGoogle Scholar
- Cardoso LF, Tomazio S, Herder JL: Conceptual design of a passive arm orthosis. Proceedings, ASME Design Engineering Technical Conferences 2002.Google Scholar
- Sanchez RJ, Liu J, Rao S, Shah P, Smith R, Rahman T, Cramer SC, Bobrow JE, Reinkensmeyer DJ: Automating Arm Movement Training Following Severe Stroke: Functional Exercises With Quantitative Feedback in a Gravity-Reduced Environment. IEEE Transactions on Neural systems and Rehabilitation Engineering 2006,14(3):378-389. 10.1109/TNSRE.2006.881553View ArticlePubMedGoogle Scholar
- Colombo G, Joerg M, Schreier R, Dietz V: Treadmill training of paraplegic patients using a robotic orthosis. J Rehabil Res Dev 2000,37(6):693-700.PubMedGoogle Scholar
- Banala S, Kulpe A, Agrawal SK: A Powered Leg Orthosis for Gait Rehabilitation of Motor-Impaired Patients. Proceedings, IEEE International Conference on Robotics and Automation 2007.Google Scholar
- Banala S, Kim S, Agrawal SK, Scholz JP: Robot Assisted Gait Training with Active Leg Exoskeleton (ALEX). IEEE Trans Neural Syst Rehabil Eng. 2009,17(1):2-8. 10.1109/TNSRE.2008.2008280View ArticlePubMedGoogle Scholar
- Hornby TG, Zemon DH, Campbell D: Robotic-assisted, body-weight-supported treadmill training in individuals following motor incomplete spinal cord injury. Phys Ther 2005, 85: 52-66.PubMedGoogle Scholar
- Gottschall JS, Kram R: Energy cost and muscular activity required for leg swing during walking. J Appl Physiol. 2005,99(1):20-23. 10.1152/japplphysiol.01190.2004View ArticleGoogle Scholar
- Banala S, Agrawal SK, Fattah A, Krishnamoorthy V, Hsu WL, Scholz JP, Rudolph K: Gravity-Balancing Leg Orthosis and Its Performance Evaluation. IEEE Transcation on Robotics 2006,22(6):1228-1237. 10.1109/TRO.2006.882928View ArticleGoogle Scholar
- Agrawal SK, Banala S, Fattah A, Sangwan V, Krishnamoorthy V, Hsu WL, Scholz JP: Assessment of Motion of a Swing Legand Gait Rehabilitation with a Gravity Balancing Exoskeleton. IEEE Trans Neural Syst Rehabil Eng. 2007,15(3):410-420. 10.1109/TNSRE.2007.903930View ArticlePubMedGoogle Scholar
- Krishnamoorthy V, Hsu WL, Kesar TM, Benoit DL, Banala SK, Perumal R, Sangwan V, Binder-Macleod S, Agrawal SK, Scholz JP: Gait Training following stroke: A pilot study combining a gravity-balanced orthosis device, functional electrical stimulation and visual feedback. Journal of Neurologic Physical Therapy 2008, 232: 102-202.Google Scholar
- Winter DA: Biomechanics and Motor Control of Human Movement. John Wiley & Sons, Inc; 1990.Google Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.