Locomotor adaptation to a powered ankle-foot orthosis depends on control method
© Cain et al; licensee BioMed Central Ltd. 2007
Received: 07 March 2007
Accepted: 21 December 2007
Published: 21 December 2007
We studied human locomotor adaptation to powered ankle-foot orthoses with the intent of identifying differences between two different orthosis control methods. The first orthosis control method used a footswitch to provide bang-bang control (a kinematic control) and the second orthosis control method used a proportional myoelectric signal from the soleus (a physiological control). Both controllers activated an artificial pneumatic muscle providing plantar flexion torque.
Subjects walked on a treadmill for two thirty-minute sessions spaced three days apart under either footswitch control (n = 6) or myoelectric control (n = 6). We recorded lower limb electromyography (EMG), joint kinematics, and orthosis kinetics. We compared stance phase EMG amplitudes, correlation of joint angle patterns, and mechanical work performed by the powered orthosis between the two controllers over time.
During steady state at the end of the second session, subjects using proportional myoelectric control had much lower soleus and gastrocnemius activation than the subjects using footswitch control. The substantial decrease in triceps surae recruitment allowed the proportional myoelectric control subjects to walk with ankle kinematics close to normal and reduce negative work performed by the orthosis. The footswitch control subjects walked with substantially perturbed ankle kinematics and performed more negative work with the orthosis.
These results provide evidence that the choice of orthosis control method can greatly alter how humans adapt to powered orthosis assistance during walking. Specifically, proportional myoelectric control results in larger reductions in muscle activation and gait kinematics more similar to normal compared to footswitch control.
Advancements in robotic technology have enabled several research groups around the world to build working robotic exoskeletons for assisting human locomotion [1–8]. The exoskeletons have a range of intended uses including enhancing human performance in healthy individuals, replacing motor capabilities in disabled individuals, and aiding in neurological rehabilitation. In each case, improvements in computer processing, energy efficiency, and sensors and actuators are allowing devices to far surpass previous expectations.
In order for robotic exoskeletons to better assist humans, it is imperative to determine how humans respond to mechanical assistance given by exoskeletons. Most of the published research has focused on hardware and software development. Few studies have actually measured human motor adaptation or physiological responses when using the devices. The human response is a key aspect that determines the success of the exoskeleton. Different exoskeleton control methods could produce extremely different levels of adaptation and adaptation rate, meaning that certain control schemes could prevent a user from effectively using an exoskeleton.
One of the main factors likely affecting how humans respond to mechanical assistance from an exoskeleton is the method of control. A wide range of control algorithms have been used by different research groups. They can rely on kinematic, kinetic, or myoelectric feedback, or some combination of these [3, 7–15]. Because each research group has their own custom-built hardware along with their own control algorithm, it would be difficult to separate the effects of controller from hardware even if human response results were readily available in the literature.
We developed a single-joint ankle exoskeleton (i.e. powered ankle-foot orthosis) that can supply mechanical plantar flexion assistance during walking [14–17]. For this study, we studied locomotor adaptation in healthy subjects walking with the powered ankle-foot orthosis using two different orthosis control methods. By using the same exoskeleton to evaluate each orthosis control method, we can separate the effects of the controller from the hardware. One group of subjects used footswitch control that activated the orthosis when the forefoot made contact with the ground . A second group of subjects used proportional myoelectric control that activated the orthosis based on soleus electromyography amplitude [14, 18]. The two orthosis control methods were chosen based on our previous experience and familiarity with how they could be used with our specific exoskeleton. The footswitch control is a simple and purely kinematic/kinetic orthosis control method, depending only upon the gait kinematics of the subject and the forces acting on the foot during gait. The proportional myoelectric control is an orthosis control method depending only upon the subject's motor commands.
The purpose of this study was to directly compare human responses to a robotic exoskeleton using two different orthosis control methods. The two control methods affect the relationship of the efferent signal to movement in different ways. In footswitch control the supplied exoskeleton torque and the efferent signal are not well related – existence of muscle activation or motor commands does not guarantee that the exoskeleton is producing torque. In proportional myoelectric control, the supplied exoskeleton torque is related directly to the motor command. We hypothesized that different control methods (footswitch versus proportional myoelectric) used to control a powered ankle-foot orthosis would produce differences in how subjects adjusted gait kinematics and muscle activation to adapt to the powered exoskeleton.
Twelve healthy subjects [(mean ± standard deviation) 6 male, 6 female, age 25.15 ± 2.5 years, body mass 74.1 ± 11.84 kg] gave informed consent and participated in the study. The University of Michigan Medical School Institutional Review Board approved the protocol.
We attached a pneumatic artificial muscle to the posterior of each AFO. Inflating (pressurizing) the pneumatic muscle created a plantar flexor torque. The artificial pneumatic plantar flexor muscle had a moment arm of approximately 10 centimeters. Air was supplied to the pneumatic muscle by four parallel proportional pressure regulators (MAC Valves, Inc., Wixom, MI) via nylon tubing (0–6.2 bar). An analog-controlled solenoid valve (MAC Valves, Inc., Wixom, MI) was attached in parallel with the air supply to assist in exhausting unwanted air from the pneumatic muscle. Pressurization of the pneumatic muscle and solenoid valve activity produced sounds that were audible to the subject.
Subjects completed two identical sessions of testing wearing the AFO. Each session went as follows: 10 minutes of treadmill walking with the AFO passive (Passive AFO), 30 minutes of treadmill walking with the AFO powered (Active AFO), and finally 15 minutes of walking with the AFO passive (Passive AFO). The transitions from passive to powered, and powered to passive, occurred without stopping. For safety, we gave the subject an oral countdown to when the transition would occur. The second session of testing was completed three days after the first session. This three day rest period was chosen to allow the subjects to recover from any muscle fatigue and soreness that may have occurred during the first session.
All subjects were naive, never experiencing walking with a powered orthosis until the first day of training. Before testing, subjects were told that the powered orthosis would provide "extra push-off force." We instructed subjects to walk in the manner they preferred and that it would take some time to adjust to the powered orthosis.
The pressure in the pneumatic muscle was controlled by one of two real-time control schemes: proportional myoelectric control or foot switch control (Figure 1). Subjects experienced either proportional myoelectric control or foot switch control (six subjects, 3 male and 3 female, in each control scheme).
In the footswitch control scheme, we controlled the pressure in the pneumatic muscle through the use of a forefoot footswitch (B & L Engineering, Tustin, CA). This footswitch control was implemented through a desktop computer and a real-time control board (dSPACE, Inc., Northville, MI). The software was composed in Simulink (The Mathworks, Inc., Natick, MA) and converted to ControlDesk (dSPACE, Inc., Northville, MI). The software sent a 0 to 10 V analog signal to the proportional pressure regulators and solenoid valves to control the activation and deactivation (pressure) of the pneumatic muscles. The software program regulated air pressure in the pneumatic muscle via an on-off or "bang-bang" controller. If the voltage signal from the footswitch was below the threshold value (a threshold was used to ensure a consistent pressure control signal), then the software signaled for zero or minimum pressure in the pneumatic muscle. If the voltage signal was above the threshold, the software signaled for maximum pressure in the pneumatic muscle.
In the proportional myoelectric control scheme, the pressure in the pneumatic muscle was proportional to the processed soleus electromyography (EMG). The EMG signal was processed as follows: It was first high-pass filtered with a second-order Butterworth filter (cutoff frequency 20 Hz) to remove movement artifact, full wave rectified, and low-pass filtered with a second-order Butterworth filter (cutoff frequency 10 Hz) in order to smooth the signal. Setting threshold cutoff values appropriately eliminated background noise in the signal. The amplitude of the control signal was scaled with adjustable gains. The control was implemented in the same way as the footswitch control except that the control signal was proportional. Data from the six subjects who used proportional myoelectric control was previously reported by Gordon and Ferris .
Because the control signal that resulted from the myoelectric control scheme was proportional, it was important to set the gain of the control signal consistently. We tuned the gain separately each day to ensure that the relationship between the soleus EMG and the control signal remained the same. To set the gain, we followed the following procedure: 1) While the subject walked with the AFO passive (the first Passive AFO period), we adjusted the gain without activating the AFO so that a maximum control signal (10 V) was produced at the maximum or peak of the soleus EMG. 2) We then doubled the gain. 3) After doubling the gain, we did not change it for the remainder of the training session.
It is important to note that there is not a simple linear relationship between the control signal amplitude (whether it is from electromyography or a footswitch) and the force developed by the muscle/torque provided by the orthosis. The control signal directly controlled the pressure supplied to the pneumatic muscle. Increasing pressure in the muscle increases the force developed by the muscle. However the force that the muscle actually develops is affected by its activation (pressure), the muscle length, and the bandwidth . In isometric conditions, a pneumatic muscle is able to develop 1700 N of force. As the muscle shortens, less force is developed. When the muscle reaches its minimum length (~71% of its resting length), the force developed drops to zero. The force bandwidth of the artificial muscle is approximately 2.4 Hz, which is very similar to the 2.2 Hz force bandwidth of human muscle . Approximately a 50 ms electromechanical delay existed between onset of the control signal and the initial rise in the artificial muscle tension. A more detailed description of the pneumatic muscle performance can be found in Gordon et al.. There is no direct relationship between the control signal and the force/torque provided by the AFO. Therefore, a bang-bang control signal does not result in an applied bang-bang torque or power at the ankle joint.
We recorded kinematic, kinetic, and electromyography data from each subject during the first 10 seconds of every minute as they walked on a treadmill at 1.25 m/s. Kinematic data was sampled at 120 Hz. All other signals were sampled at 1200 Hz. Three-dimensional kinematic data was recorded using a 6-camera video system (Motion Analysis Corporation, Santa Rosa, CA) and twenty-nine reflective markers placed on each subject's pelvis and lower limbs. Step cycle data was collected using footswitches (B & L Engineering, Tustin, CA), which were placed in each shoe. Artificial pneumatic muscle force was measured using a compression load cell (Omega Engineering, Stamford, CT) mounted in series with the pneumatic muscle. We recorded lower limb surface EMG (Konigsberg Instruments, Inc., Pasadena, CA) from the left soleus, tibialis anterior, medial gastrocnemius, lateral gastrocnemius, vastus lateralis, vastus medialis, rectus femoris, medial hamstring and lateral hamstring muscles using bipolar surface electrodes. The EMG was bandpass filtered with a lower bound of 12.5 Hz and an upper bound of 920 Hz. We minimized crosstalk by visually inspecting the EMG signals during manual muscle tests prior to treadmill walking, moving electrode placement if needed. We marked the position of the electrodes on each subject's skin using a permanent marker to ensure the same electrode placement for the second session of testing. The sound of the pneumatic muscle inflating and deflating was audible to the subjects for both control signals. No distinguishable difference between the noises associated with each controller could be identified.
We created average step cycle profiles of each minute of walking for EMG, kinematic, and kinetic variables for each subject. Each minute's average step cycle was calculated from the complete step cycles that occurred during the first 10 seconds of that minute. To examine how EMG amplitude changed over time, we calculated the normalized root mean squared (RMS) EMG values for each minute of walking for each subject. RMS EMG values were calculated from high pass filtered (cutoff frequency 40 Hz) and rectified EMG data for the complete gait cycle, stance phase, and swing phase. All RMS EMG values were normalized to the last minute of walking with the passive AFO before activating the pneumatic muscle (the last pre-passive minute), or what we called the Baseline condition. We also made average step profiles for the joint angles that were created from the marker data (low-pass filtered, cutoff frequency 6 Hz). In order to examine the changes in the kinematics over time, we calculated joint angle correlations between the average step cycle profiles of each minute and the average joint profile from the last pre-passive minute for the same session. We created average step cycle torque and power profiles for the AFO only (torque and power that the AFO was producing). From these, we calculated the positive and negative work performed by the AFO during a step cycle. Foot and shank parameters were adjusted to account for added AFO mass and inertia.
We used a general linear model (GLM), or multiple regression, to test for significant effects between controllers, effects of minute within footswitch control group, and effects of minute within proportional myoelectric control group for the four outcome parameters (soleus EMG RMS, ankle angle correlation common variance, positive orthosis work, and negative orthosis work). The equation for the general linear model is of the form y = β0 + β1x1 + β2x2 + ... + βnxn + ε, where Y is the response variable, βn are model parameters, and ε is the error. Our previous study examining subjects using proportional myoelectric control found that subjects were at steady state walking dynamics for the last 15 minutes of powered orthosis walking on the second day of training . As a result, we used only the last 15 minutes of data on day 2 to test for significant differences between controllers during steady state. A general linear model was also used to test the effect of controller on post-adaptation, or the period of walking after turning the power to the AFO off. The entire 15 minutes of post-powered orthosis walking was used for the post-adaptation analysis.
To test for differences in adaptation rate between controllers, we used the methodology of Noble and Prentice . This method defines a band of normal variation within steady state dynamics and then calculates the amount of time required to reach and stay within that band. As mentioned above, we used data from the last 15 minutes of powered walking on day two for the steady state period. The band of steady state variation for each outcome parameter was calculated as the mean ± two standard deviations from the steady state period. Time to steady state was defined as the time it took for a measure to enter the steady state range and remain there for three consecutive minutes without any two consecutive minutes outside of the steady state range afterwards. This analysis was performed for each subject individually. Differences in learning rate (time to steady state) were assessed using a repeated measures ANOVA.
An overground testing session was used to measure the amount of work and power that each subject produced without the AFO. This let us estimate the amount of assistance that the powered AFO was providing the subjects. During the overground collection, a subject would walk without wearing an AFO over two force plates at a speed of 1.25 m/s (± 0.06 m/s). Subjects completed ten trials. Force plate data and kinematic marker data were used to calculate net torques and work performed about the ankle joint by using commercial software (Visual3D, C-Motion, Inc., Rockville, MD).
Effects and responses
Muscle activation patterns were modified as the subjects trained with the powered AFO. Examples of these changes can be seen in Figures 4 and 5. By the end of the second day of training, differences in the muscle activation patterns compared to passive orthosis walking were very subtle. The exception to this was the soleus muscle activation amplitude in the subjects using proportional myoelectric control (Figure 3). There were no significant differences in stride time between orthosis control methods, condition, or day. Footswitch subjects had a stride time of 1.26 ± 0.10 seconds (mean ± standard deviation) and proportional myoelectric subjects had a stride time of 1.24 ± 0.12 seconds. The artificial plantar flexor produced a peak torque that was approximately 47% of the peak torque generated at the ankle when walking overground (Figure 3). As subjects trained with the powered AFOs, the torque and power produced by the AFO became more focused at toe-off (Figure 3).
Subjects using both controllers walked with kinematics different from baseline (GLM, p < 0.03). Only subjects using proportional myoelectric control reduced EMG amplitudes of the soleus, medial gastrocnemius, and lateral gastrocnemius below baseline (GLM, p < 0.03). It is important to note that Gordon and Ferris  only found that the soleus EMG amplitude was significantly different from baseline for subjects (n = 10) using proportional myoelectric control.
No significant differences in post-passive adaptation rate were found between the two controllers.
Subjects using proportional myoelectric control returned closer to their normal (Baseline) kinematic patterns by the end of the second day compared to subjects using footswitch control. There are several aspects of the proportional myoelectric control that could have contributed to this difference. First, proportional control allows for a more graded response in orthosis dynamics than the bang-bang nature of footswitch control used in this study. With step-to-step variability in orthosis output, it would likely be easier for the nervous system to determine the relationship between soleus activation and orthosis assistance using proportional myoelectric control than using footswitch control. Second, proportional myoelectric control put the orthosis under a control mode that is more similar to the normal physiologic control that the nervous system uses to generate motion. It is likely that the nervous system has some representation of the transfer function from soleus motor neuron recruitment to ankle movement. Wearing the orthosis with proportional myoelectric control would likely be interpreted as a relatively minor change in the transfer function. Wearing the orthosis with footswitch control would likely be a more non-natural modification to lower limb movement control.
The artificial pneumatic plantar flexor produced a peak torque 47% of the maximum ankle plantar flexor torque produced when walking (Figure 3). We did not expect the powered orthosis to provide all of the torque needed at the ankle during gait. In a previous study by Gordon et al. the powered orthosis was only able to generate a peak plantar flexor torque that was 57% of the peak net ankle plantar flexor moment, regardless of the potential force generation capabilities of the artificial pneumatic plantar flexor. Gordon et al. also found that the net ankle moment remained approximately the same regardless of the assistance given to the subjects; the sum of the AFO produced torque plus the physiological torque was approximately equal to the physiological torque produced when walking without a powered orthosis. A good estimate of what torque the ankle is producing is the difference between overground biological torque and the torque produced by the powered orthosis (Figure 3). Previously, the powered orthosis was found to produce about 70% of the positive plantar flexor work done during normal walking .
It is possible that the footswitch control signal was producing too much torque (more than required for normal walking). Reducing the magnitude of the bang-bang control signal used for the footswitch control method could allow a new dynamic equilibrium point closer with normal or baseline kinematics and reduced plantar flexion activation.
The differences in soleus activation between the two controllers (Figure 8) suggest that proportional myoelectric control may lead to a lower metabolic cost of transport than the footswitch control. Muscle activation requires the use of metabolic energy. Although nonlinear factors such as muscle length and velocity will affect the relationship between muscle recruitment and metabolic cost , the larger reductions in plantar flexor muscle recruitment for proportional myoelectric control compared to footswitch control may override the differences in muscle-tendon kinematics. This is an important possibility to consider given recent findings from Norris et al.. They showed that the metabolic cost of transport decreased by about 13% when subjects walked with two powered AFOs similar to the design used in this study . However, Norris et al. used a bang-bang control algorithm that started and stopped orthosis activation based on the angular velocity of the foot. Thus, this type of control was similar to our footswitch control; it depended on motion and not neurological signals. It seems feasible that proportional myoelectric control might reduce the metabolic cost of transport during walking more than 13%.
The two controllers produced similar adaptation rates for most parameters. The only significant difference in adaptation rates between controllers was for negative orthosis work. Subjects using footswitch control reached steady state faster on both days of training compared to subjects using proportional myoelectric control. Regardless of control mode, subjects adapted to the powered orthosis much more quickly on the second day. This indicates that subjects were able to store a motor memory of how to walk with the orthosis and then recall that motor memory on a later date. The controller used did not seem to affect this formation or recall of the motor memory.
The results from this study may have been altered if subjects had been allowed to practice using the orthosis for a longer time period. Additional days of training might have resulted in further adaptation to the walking pattern. However, given the relative steady state nature of the outcome parameters during the last 15 minutes of day two (Figure 6), any additional changes would have likely required multiple days.
The choice of controller for a robotic exoskeleton can have a substantial effect on human-machine performance. If the goal of the exoskeleton is to walk with relatively normal joint kinematics, proportional myoelectric control should be preferred over footswitch control. In addition, it seems reasonable to suggest that proportional myoelectric control may provide metabolic savings greater than those from footswitch control as well.
The findings of this study also have important implications for rehabilitation. While rate of motor adaptation was not affected by controller, the steady state walking dynamics were more similar for proportional myoelectric control than footswitch control. This suggests that robotic devices designed to facilitate adaptive training may benefit from more direct nervous system control. Proportional myoelectric control may also have the benefit of amplifying movement errors during practice. Patton et al. found that practice with error-enhancing mechanical forces was more effective in improving movement ability of stroke subjects compared to practice with error-reducing mechanical forces. It would be very interesting to examine how patients with neurological deficits responded to walking practice with a powered orthosis under proportional myoelectric control. It could improve motor learning by enhancing errors in neuromuscular activation patterns in a manner to that found by Patton et al.. Future studies are needed to examine this possibility.
The authors would like to thank Catherine Kinnaird and other members of the Human Neuromechanics Laboratory for assistance with data collection and analysis. This study was supported by NIH grant R01 NS045486 and NSF GRFP.
- Ferris DP, Sawicki GS, Domingo AR: Powered lower limb orthoses for gait rehabilitation. Topics in Spinal Cord Injury Rehabilitation 2005, 11: 34-49.PubMed CentralView ArticlePubMedGoogle Scholar
- Kawamoto H, Sankai Y: Power assist system HAL-3 for gait disorder person. In Computer Helping People with Special Needs: 8th International Conference, ICCHP 2002. Lecture Notes in Computer Science. Edited by: Miesenberger K, Klaus J, Zagler W. Berlin: Springer Verlag; 2002:196-203.View ArticleGoogle Scholar
- Suzuki K, Kawamura Y, Hayashi T, Sakurai T, Hasegawa Y, Sankai Y: Intention-based walking support for paraplegia patient. 2005, 2703: 2707-2713.Google Scholar
- Zoss AB, Kazerooni H, Chu A: Biomechanical Design of the Berkeley Lower Extremity Exoskeleton (BLEEX). IEEE/ASME Transactions on Mechatronics 2006, 11: 128-138.View ArticleGoogle Scholar
- Kazerooni H, Steger R: The Berkeley Lower Extremity Exoskeleton. Journal of Dynamic Systems Measurement and Control-Transactions of the Asme 2006, 128: 14-25.View ArticleGoogle Scholar
- Jacobsen SC, Olivier M, Smith FM, Knutti DF, Johnson RT, Colvin GE, Scroggin WB: Research robots for applications in artificial intelligence, teleoperation and entertainment. International Journal of Robotics Research 2004, 23: 319-330.View ArticleGoogle Scholar
- Pratt JE, Krupp BT, Morse CJ, Collins SH: The RoboKnee: an exoskeleton for enhancing strength and endurance during walking. In IEEE International Conference on Robotics and Automation; New Orleans, LA. IEEE Press; 2004:2430-2435.Google Scholar
- Blaya JA, Herr H: Adaptive control of a variable-impedance ankle-foot orthosis to assist drop-foot gait. IEEE Transactions on Neural Systems and Rehabilitation Engineering 2004, 12: 24-31.View ArticlePubMedGoogle Scholar
- Kazerooni H, Steger R, Huang L: Hybrid Control of the Berkeley Lower Extremity Exoskeleton (BLEEX). International Journal of Robotics Research 2006, 25: 561-573.View ArticleGoogle Scholar
- Ghan J, Steger R, Kazerooni H: Control and system identification for the Berkeley lower extremity exoskeleton (BLEEX). Advanced Robotics 2006, 20: 989-1014.View ArticleGoogle Scholar
- Kawamoto H, Sankai Y: Power assist method based on Phase Sequence and muscle force condition for HAL. Advanced Robotics 2005, 19: 717-734.View ArticleGoogle Scholar
- Hayashi T, Kawamoto H, Sankai Y: Control method of robot suit HAL working as operator's muscle using biological and dynamical information. 2005, 3063-3068.Google Scholar
- Kawamoto H, Lee S, Kanbe S, Sankai Y: Power assist method for HAL-3 using EMG-based feedback controller. International Conference on Systems, Man and Cybernetics; October 5–8. IEEE 2003, 1648-1653.Google Scholar
- Ferris DP, Czerniecki JM, Hannaford B: An ankle-foot orthosis powered by artificial pneumatic muscles. Journal of Applied Biomechanics 2005, 21: 189-197.PubMed CentralPubMedGoogle Scholar
- Ferris DP, Gordon KE, Sawicki GS, Peethambaran A: An improved powered ankle-foot orthosis using proportional myoelectric control. Gait Posture 2006,23(4):425-428.View ArticlePubMedGoogle Scholar
- Gordon KE, Sawicki GS, Ferris DP: Mechanical performance of artificial pneumatic muscles to power an ankle-foot orthosis. Journal of Biomechanics 2006, 39: 1832-1841.View ArticlePubMedGoogle Scholar
- Sawicki GS, Domingo A, Ferris DP: The effects of powered ankle-foot orthoses on joint kinematics and muscle activation during walking in individuals with incomplete spinal cord injury. J Neuroeng Rehabil 2006, 3: 3.PubMed CentralView ArticlePubMedGoogle Scholar
- Gordon KE, Ferris DP: Learning to walk with a robotic ankle exoskeleton. J Biomech 2007,40(12):2636-2644.View ArticlePubMedGoogle Scholar
- Browning RC, Modica JR, Kram R, Goswami A: The effects of adding mass to the legs on the energetics and biomechanics of walking. Med Sci Sports Exerc 2007, 39: 515-525.View ArticlePubMedGoogle Scholar
- Aaron SL, Stein RB: Comparison of an EMG-controlled prosthesis and the normal human biceps brachii muscle. Am J Phys Med 1976, 55: 1-14.PubMedGoogle Scholar
- Noble JW, Prentice SD: Adaptation to unilateral change in lower limb mechanical properties during human walking. Exp Brain Res 2006,169(4):482-495.View ArticlePubMedGoogle Scholar
- Wolpert DM, Ghahramani Z: Computational principles of movement neuroscience. Nat Neurosci 2000, 3: 1212-1217.View ArticlePubMedGoogle Scholar
- Rall JA: Energetic aspects of skeletal muscle contraction: implications of fiber types. Exerc Sport Sci Rev 1985, 13: 33-74.PubMedGoogle Scholar
- Norris JA, Granata KP, Mitros MR, Byrne EM, Marsh AP: Effect of augmented plantarflexion power on preferred walking speed and economy in young and older adults. Gait Posture 2007,25(4):620-627.View ArticlePubMedGoogle Scholar
- Patton JL, Stoykov ME, Kovic M, Mussa-Ivaldi FA: Evaluation of robotic training forces that either enhance or reduce error in chronic hemiparetic stroke survivors. Exp Brain Res 2006, 168: 368-383.View ArticlePubMedGoogle 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.