Comparisons involving the static control law suggest three separate contributions to the success of this approach: discovering a good generic assistance pattern, customizing it to individual users, and facilitating motor learning.
We can learn from static control law (constant torque) when compared to optimized assistance: it can identify similar patterns, it accommodates different individuals, and increase walking performance.
Results from a prior experiment using the same hardware (17), comparing the zero-torque condition with walking in normal shoes (no exoskeleton) or with static assistance.
Normal shoes mean no exoskeleton and static assistance mean exoskeleton and constant torque.
active prostheses
Prostheses are an artificial part of a body. It can play an unconscious or conscious role in the body. For example, a titanium leg can replace a leg and adapt itself to the host walking patterns.
Metabolic energy cost of walking for each condition, tested in validation trials. Optimized assistance resulted in the lowest metabolic rate and a large reduction compared to the zero-torque condition. Variability is primarily due to differences between participants. Bars are means, error bars are standard deviations, and asterisks denote statistical significance (P < 0.05)
Three conditions were compared: zero torque (walking, no torque), optimized (walking, torque applied), and static (constant torque). In Figure 1A, the optimized method saved the most energy, with static coming in second, and zero torque last.
The primary outcome was the energy cost of walking, defined as gross metabolic rate during walking minus the rate measured while standing still
The goal of the optimization method, through the use of control laws governing the exoskeleton, is to reduce the amount of energy walking. The amount of energy lost during a walk is defined as the amount of energy measured walking subtracted by the amount of energy standing still (basal level).
T. K. Uchida, A. Seth, S. Pouya, C. L. Dembia, J. L. Hicks, S. L. Delp, Simulating ideal assistive devices to reduce the metabolic cost of running. PLOS ONE 11, e0163417 (2016). doi:10.1371/journal.pone.0163417pmid:27656901
This paper presents the results of a muscle simulator with a massless device. They show that assisting movements with a device is not always necessary and can reduce muscle activity. The authors hypothesize this can provide insight into future assistive device designs. Models, simulations and software are freely available online.
Static assistance resulted in a 19.3 ± 8.6% reduction in energy cost compared to zero torque (t test, P = 4 × 10−5, n= 11), a larger improvement than in our
Compared with zero torque, static assistance out-performed zero torque (walking and no torque applied) with statistical difference.
The double-reversal validation test
Double-reversal validation test tries to define and remove bias in the data. For example, consider a natural factor that could remove one parameter, and ask if you should counter this evolution to keep the status quo. If so, consider when the natural factor is about to vanish and ask if it is a good thing to act again to reverse the first intervention you made. If not, it is preferable to think that the first intervention is usefull even in the absence of the natural factor.
J. R. Koller, D. H. Gates, D. P. Ferris, C. D. Remy, “Body-in-the-loop optimization of assistive robotic devices: A validation study,” paper presented at Robotics: Science and Systems XII, Ann Arbor, MI, 18 to 22 June 2016; available at http://www.roboticsproceedings.org/rss12/p07.pdf.
This paper introduces a new type of human-machine interaction to calculate datas in real-time.
R. W. Jackson, S. H. Collins, An experimental comparison of the relative benefits of work and torque assistance in ankle exoskeletons. J. Appl. Physiol. 119, 541–557 (2015). doi:10.1152/japplphysiol.01133.2014pmid:26159764
This paper studies how the exoskeleton influence human mobility.
(CMA-ES)
Evolution strategy (ES) is an optimisation technique based on the principle of evolution. The Covariance Matrix Adaptation (CMA) is one method that proved itself to be one of the more effective by recalculate at each iteration the Covariance matrix of the distribution.
transient metabolic data
Transient represents a small time interval, which is the case for the metabolic data represented.
first-order dynamical model
Such a model is based on a first order differential equation (ODE). An ODE represents the slope at one time point, rather than between two time points.
Exoskeletons
An exoskeleton refers to a skeleton outside the body - or external skeleton. It has a role of protection in addition to the support role of the internal skeleton. It is commonly found in the insect kingdom.
The static pattern, based on (17), is similar to the optimized patterns but resulted in higher metabolic rate. Torque was negligible in the zero-torque mode.
Although static pattern and optimization both experience similar torque patterns, static condition had a higher metabolic rate, meaning it spent more energy.
quadratic approximations
In Mathematics, approximation means a function can be estimated by a simpler function. The quadratic approximation makes use of the second derivative of the function of interest.
physiological
Physiology is a fiel of study that focus on the interactions between different systems inside one organism and their normal mechanism.
motor learning
Motor learning is the change of response or performence one can do after a new experience or training.
biomechanics observations
Biomechanics observations focuses on the motions and forces in actions while the body is moving.
customization
Customization consists of addapting a devise in regard of the differences between individuals. This for a better accomodation.
benchmarking problems
Benchmark problem is a way to compare several algorithms to check wich one is the more efficient to solve a problem.
parametric interactions
parametric interaction is the way parameters interact or influence each other.
time-varying dynamics
The dynamic of a system is how this system will evolve with time. Time-varying means here that the system will have different patterns according to the time it was recorded.
R. W. Jackson, C. L. Dembia, S. L. Delp, S. H. Collins, Muscle-tendon mechanics explain unexpected effects of exoskeleton assistance on metabolic rate during walking. J. Exp. Biol. jeb.150011 (2017). doi:10.1242/jeb.150011pmid:28341663
The papers examines the different modes of exoskeleton torque applied to different participants, and concludes how different combinations of torque can have such a variety of muscle behaviours.
K. A. Witte, J. Zhang, R. W. Jackson, S. H. Collins, “Design of two lightweight, high-bandwidth torque-controlled ankle exoskeletons,” in 2015 IEEE International Conference on Robotics and Automation (Institute of Electrical and Electronics Engineers, 2015), pp. 1223–1228.
In this paper, authors have created two tethered ankle-foot exoskeletons.
R. M. Alexander, Optimization and gaits in the locomotion of vertebrates. Physiol. Rev. 69, 1199–1227 (1989). pmid:2678167
This paper studies the various ways of locomotion of the vertebrates.
human locomotor performance
Locomotor performances are the variety of movements a human can do related to the energy needed for such a movement.
torque patterns
Torque represents a rotational force. It can be seen as the twist of an object.
B. T. Quinlivan, S. Lee, P. Malcolm, D. M. Rossi, M. Grimmer, C. Siviy, N. Karavas, D. Wagner, A. Asbeck, I. Galiana, C. J. Walsh, Assistance magnitude versus metabolic cost reductions for a tethered multiarticular soft exosuit. Sci. Robot. 2, eaah4416 (2017). doi:10.1126/scirobotics.aah4416
The aim of this study was to characterize the relationship between assistance effectivness and metabolic cost of walking with an exoskeleton by examing walking patterns of test subjects. Results showed that an increase of exosuit assistance relsuts in a net decreas of the metabolic rate.
S. H. Collins, M. B. Wiggin, G. S. Sawicki, Reducing the energy cost of human walking using an unpowered exoskeleton. Nature 522, 212–215 (2015). doi:10.1038/nature14288pmid:25830889
Unpowered ankle exoskeleton equipment were developed by this group that behaved parallel to calf muscles and reduce the energy (metabolic cost) spent on walking over 7% by reducing load on the muscles, which is comparable to powered exoskeletons that are currently used. In addition, authors implied that human locomotion can still be more effecient.
L. M. Mooney, E. J. Rouse, H. M. Herr, Autonomous exoskeleton reduces metabolic cost of human walking. J. Neuroeng. Rehabil. 11, 151 (2014). doi:10.1186/1743-0003-11-151pmid:25367552
Authors enhanced the design of previous autonomous exoskeleton for unloaded walking conditions along with significant reduction of energy (metabolic rate) consumption.
prior study
They studied total exoskeleton work that reduced metabolic cost (energy) by 17% and average exoskeleton torque that increased the energy consumption by 13% on individual human gaits.
Many optimization methods that work well in simulation
These studies consider all types of movement (running, walking, etc.) through computer simulations for optimisation of energy usage which did not yield the same physical results of body movement.
Initial efforts in this domain have demonstrated the ability to optimize a single gait or device parameter using line search (24) or gradient descent
Earlier proposed devices optimised for limited movement (walking or running) rather than generalising the devices for all kinds of movements.
Torque was applied to one ankle using a versatile exoskeleton emulator system (30) (Fig. 2, C and D, and figs. S2 and S3) with precise low-level torque control (31). The emulator, inspired in part by other laboratory-based testbeds (32–34), allows a wide range of assistive behaviors to be applied in rapid succession, without the need to design or build new hardware
The methods the authors used required changes in the torque. So, instead of building different exoskeletons with different torque values, they decided to use an emulator, a device that will provide the same assistance and results as the exoskeleton, but with easily adjustable peak torque and timing of peak torque. This made it easier to test different generations with different torque values.
The best available estimate of metabolic energy cost, for example, requires about one minute of respiratory data per evaluation
Most optimized experiments performed had constant decay rates of 20 breaths per 60 seconds. For better approximations, we need more evolved noise removal techniques and fitting techniques.
nervous systems
The nervous system is responsible for the coordination of body by sending signals to the other organ systems.
mathematical models (12), biomechanics observations (13), and humanoid robots (14), but each simplifies important aspects of the human-robot system
These studies aims on designing robots that have locomotion similar to humans/animals, and later using the same to study the movement patterns and adaptability. The authors expect these studies to assist in developing the more humanised prosthetic limbs.
Assistance strategies have typically been derived from mathematical models (12), biomechanics observations (13), and humanoid robots
These papers suggest the possible methods to reduce energy consumption. Collins et. al proposed that reducing ankle push-off results in increased metabolic energy expenditure along with increasing muscle damage.
only modest enhancements have been achieved compared to the potential benefits expected on the basis of simulations
Patients having large muscle degeneration can have optimally designed exoskeletons (equipment used to improve walking). These devices can significantly reduce metabolic energy needed for muscle movement.
Few approaches have been successful
Previous work on bionic prosthesis, pneumatic exoskeleton, autonomic exoskeleton, and multiarticular soft exosuit reduced the cost of metabolism (lower energy) within minor ranges as compared to the work by these authors, which will be seen throughout the article.
The double-reversal validation test prevented confounding influences from measurement noise during optimization and trial order during validation (26). Participants were not exposed to any of the validation conditions during optimization, because optimized assistance was the weighted average of the best control laws from the final generation (26). The primary outcome was the energy cost of walking, defined as gross metabolic rate during walking minus the rate measured while standing still
The double reversal test refers to the use of the 'zero torque' and 'static' models, compared to the adaptable one. This comparison is used to eliminate the author's possible bias when assuming his design and adjustments may influence positively his results. Therefore the authors use it to prove their hypothesis. For the static model, different configurations were tested while for the adaptable model, since it is based on an evolutionary framework, only the last generation is taken into account for the comparison.
This parameterization also implicitly allowed adjustment of features such as the timing and amount of positive joint work, which may be important to energy economy (29). Some torque patterns were not possible with this parameterization, such as those with multiple peaks. More complex patterns, defined by additional parameters, might allow better approximations of global optima at the cost of lengthier optimization periods.
Since the assistance (assistive ankle torque) depends on 4 parameters, these can be adjusted to allow different possible configurations. Adjusting the timing and positive joint work gave the authors better results, reducing the energy the user inputs in the exoskeleton to achieve movement. If more parameters could be controlled, more efficient configurations could be achieved, but the time to evaluate all of them will be lengthier.
and the shape and size of the distribution are chosen to increase the likelihood of further improvement in subsequent generations. This optimization strategy is relatively tolerant of both measurement noise and human adaptation, because neither objective function values nor their derivatives are used directly, and each generation is evaluated independently
Using the CAM-ES method, they adapt the control laws for optimization. First they run an algorithm to adjust the pattern of assistance that uses a set of control laws for a determined time, they gather the information for a determined period of time, make random changes and adaptions with CAM-ES and start over. Unlike other type of experiments, this one is designed to make up for common situations that cause errors, such as sudden abnormal values or human adaption to the machine (which reduces the energy input).
We tested our method by optimizing the pattern of assistive torque applied by an exoskeleton worn on one ankle during walking. We applied assistance at one ankle to allow comparisons to a prior study that used the same hardware
The authors decided to apply their method in an exoskeleton in only one ankle because the same equipment was used in another study in only one ankle. So they tried to replicate the same conditions so that any differences will be a result of their pattern of assistive torque. Therefore, they could see if their method (pattern of assistive torque) could get better results than the authors of reference #17.
These tests hint at the potential for a new type of biomechanics study, in which human-in-the-loop optimization can be leveraged to compare the best possible outcomes for different devices or gait conditions or to test how various features of gait change with optimized performance.
The authors suggest that human-in-the-loop technology will birth new biomechanic studies, where the proposed technology can be expanded to a large range of studies to improve performance.
activity in the optimized condition was reduced by 41% compared to the zero-torque mode and 36% compared to walking in normal shoes
With the optimization method, muscle performance was improved in addition to reducing metabolic energy compared to normal conditions without walking assistance.
Soleus
A major muscle in the calf.
The evolutionary strategy that we used was more effective than other methods that we tried, but it seems likely that improved techniques could be developed.
Compared to previous experiments, incorportating a mathematical model that integrates evolution strategies work best. However, the authors agree that there is room for improvement.
These approaches have scope to improve mobility for people with a wide range of distinct physiological needs, from individuals with chronic stroke to athletes.
The optimzation methods proposed in the paper can be used as a foundation to improve mobility for people with different illness backgrounds.
suggesting that longer-time-scale adaptation might be at play.
Here is an example related to longer-time-scale adaptation:
*http://bjsm.bmj.com/content/bjsports/early/2017/07/19/bjsports-2016-097325.full.pdf
New candidate algorithms should tolerate high measurement noise, facilitate human adaptation, and require very few evaluations before converging.
Now that an optimization method has been produced to reduce energy consumption when walking, future methods should be able to do the same job much faster.
we found only small changes in optimized parameters (table S3) and no further reduction in metabolic rate (fig. S9).
The convergence test was done to see if continuing the generations could lead to differences in the adaption. The small number of differences in the results indicates that the number of generations used at first was optimal.
Successfully reducing both metabolic rate and muscle activity suggests that alternate objective functions with similar properties could be optimized—for example, related to speed (40), endurance (41), balance, or overall satisfaction
The goal of the paper is to discuss control laws that can adapt to individual walking patterns, with the goal of improving walking and reducing energy consumption. Since the optimization method was successful, the authors suggest other properties that can be improved using similar optimization methods, such as endurance and balance.
Optimizing a similar number of parameters in a feedback control structure, such as a neuromuscular model (38), or switching between optimized modes (39) could enhance performance under changing locomotor conditions.
The optimization method is able to accomodate unique walking patterns.
Methods for automatically discovering, customizing, and continuously adapting assistance could overcome these challenges, allowing exoskeletons and prostheses to achieve their potential.
Several methods are being tested by different groups of scientists. Here are some examples:
We performed a test of convergence with a subset of participants in the main study (n = 8) by continuing the optimization for an additional four generations
To demonstrate that the number of generations chosen in the original test was optimal, they continued the experiment to see if the results continued to improve or varied in a significant way. The initial test was done with 11 participants. From this group, 8 were chosen to continue with the adjustments using the same algorithm and methodology to adjust the control laws.
In daily life, a proxy measure such as heart rate or muscle activity
The entertainment industry has begun to take part in the health care market, providing tools to collect, interpret, and analyze personal data. Other methods for collecting data are emerging. Here are a few examples:
Optimized ankle exoskeleton torque pattern for each participant. Patterns varied widely and spanned a large portion of the allowable range. Lines are measured torque, normalized to stride time and body mass, averaged across strides.
These are the torque patterns (amount of applied torque) that improve the walking performance for each participant. These curves have been normalized, meaning they have been numerically manipulated so that a generalization can be seen across participants.
such as a neuromuscular model
Neuromuscular junctions can have effets in spinal muscular atrophy for example:
http://journal.frontiersin.org/article/10.3389/fnana.2016.00006/full
(9–11)
Handford and Srinivasan suggested that properly designed robotic prosthetic devices for amputees can have much lower energy consumption than non-amputee's ankle torque.
We also applied the approach to running with exoskeletons on both ankles (subject 2; table S1) and found a 27% improvement in energy cost compared to the zero-torque mode and 13% energy savings compared to normal running shoes (Fig. 4F).
After putting a exoskeleton on both legs (two exoskeleton total) and having the participant run, the optimization method out-performed zero torque and normal shoes.
At a slow walking speed (0.75 m s−1), the algorithm drove torque to its lower limit, resulting in a small reduction in energy cost compared to zero torque (Fig. 4A) and a 19% reduction compared to the initial control law (fig. S6).
At a slow speed, the amount of torque reduced to its minumum, causing it to burn more energy than zero torque. This suggests limitations of the software after a specific level (or threshold).
Experiments have primarily been conducted using specialized prototypes that embed a single intuited functionality, with each prototype requiring years of development,
All the experiments performed were not optimised for having generic assistance patterns along with not having motor learning function which limited their use.
Optimized assistance reduced the metabolic cost of walking at a typical speed (1.25 m s−1; 33% reduction versus zero torque, 25% versus normal shoes; Fig. 4B)
Optimized assistance out-performed walking in normal shoes, meaning that optimized assistance resulted in lower energy consumption.
These wide-ranging control laws, some of which participants noted were initially uncomfortable, may have forced them to explore new motor control strategies, which has been shown to be a necessary part of skill acquisition in some interventions
The participants experienced difficulty at first, which suggests that their mode of walking was not 'energy effecient'. Thus, the optimization method noted this mode of walking and attempted to improve stride patterns with the goal of reducing the amount of energy spent.
Participants had a similar duration of exposure to the exoskeleton in both studies, but in the prior study, participants were trained with a narrow range of eight static control laws, whereas during optimization, they experienced 32 diverse control laws.
Both experiments lasted the same duration, however, optimzation has many more control laws, the functions that accomodate unique exoskeleton behavior.
We optimized assistance for 11 participants (subjects 1 to 11; table S1) as they walked on a treadmill at a normal speed (1.25 m s−1). After optimization, we performed validation tests comparing optimized assistance with a fully passive “zero-torque” mode and with a “static” assistance condition. Static assistance approximated the best hand-tuned torque pattern for this device, which had previously resulted in a 6% reduction in energy cost compared to zero torque (17).
Two systems were used as a comparison for the adaptable model. The first was the 'zero torque' in which the device did not exert any force to assist the person walking. The second was the device with a static assisting force, that did not change or adapt with time.
The primary differences between these studies relate to the conditions during adaptation
The major difference between the results for static assistance and zero torque are the method of exoskeleton adapting to human.
Eight of 11 participants had a lower metabolic rate with optimized assistance, with the difference in rate ranging from a 3.3% increase to a 16.5% reduction
Eight people experienced better performance with the optimied assistance (stride improved, energy reduced).
Individually optimized assistance resulted in 5.8 ± 6.2% lower metabolic rate than with the static control law (t test, P = 0.01, n = 11)
When compared to static condition, the optimized condition was better at reducing the amount of energy spent walking (improved stride).
Static assistance (Fig. 3E) was similar to the average of the optimized control laws, and in our prior study (17), it delivered a 6% reduction in energy cost.
In previous studies, the static performance was able to reduce the energy cost which displays how it can be beneficial. In addition, since the optimized assistance out-performed the static condition, which means optimization is a success.
musculoskeletal
Musculoskeletal system (or locomotal system) is the system that allow a body to move. It is mainly composed of skeleton and muscles.
Human-in-the-loop optimization accommodates this complexity.
Although their are large, biological differences between individuals, the optimization assistance can find similarities among the participants.
Optimized torque patterns (Fig. 3D) did share some qualitative features with each other, such as a peak torque that occurred at about 50% of stride, suggesting qualities that may be beneficial for most people and useful initial parameters for future optimizations.
Generally, participants share similar torque at 50% stride, this is the highest point in the curve displayed in Figure 3D. The authors believe parameters produced at 50% stride can be used for future experiments.
For example, the timing of optimized torque onset ranged from onset at 17% to onset at 37% of the stride period (Fig. 3D), or about half the testable range in this and prior studies (3)
The stride (heel up, see Figure 1C) begins at 17% and ends (toe firmly placed on treadmill, Figure 1C). Thus, image 3D represents a range of strides for the 11 participants.
suggesting about a 14% net improvement with optimized assistance compared to normal shoes
Although no torque is worst than wearing normal shoes, the overall effect of including the optimized assistance, torque applied to individual, is better (14% better) at reducing the amount of energy spent walking.
N. Hansen, “The CMA evolution strategy: A comparing review” in Towards a New Evolutionary Computation, J. A. Lozano, P. Larrañaga, I. Inza, E. Bengoetxea, Eds. (Springer, 2006), pp. 75–102.
This article aims to study the differents use of the covariance matrix adaptation, depending on the size of differents sytems.
Wearing the exoskeleton in zero-torque mode is about 10% more costly than walking in normal shoes without the exoskeleton (Fig. 3B)
With the exoskeleton attached to a leg (see Figure 1C and 1D) and no torque applied, it cost an invidual more energy than not wearing the exoskeleton at all. This is not preferred, since the goal is to reduce the amount of energy.
Optimized parameters were identified after four generations
One generation, which is computed from covariance matrix adaptation evolution strategy (CMA-ES), is performed by the control law. Each generation is better than the previous generation, which is the evolutionary feature of the control law, see Figure 1A.
Optimized assistance substantially improved energy economy for all participants, confirming the effectiveness of the method.
Optimized assistance is the condition when torque is appled to improve stride, thus reducing the amount of energy spent. This method out-performed all other methods in reducing the amount of energy spent.
Static assistance approximated the best hand-tuned torque pattern for this device, which had previously resulted in a 6% reduction in energy cost compared to zero torque (17)
In previous experiments, the static condition performed better than having no torque applied to stride.
(A) Slow walking (0.75 m s−1). (B) Normal walking (1.25 m s−1). (C) Fast walking (1.75 m s−1). (D) Uphill walking (10% grade). (E) Loaded walking (load equal to 20% of body mass).
Under different walking conditions, the energy expenditure was measured for normal shoes, zero torque, and optimized.
forming one generation, a covariance matrix adaptation evolution strategy (CMA-ES) (28) is used to calculate the next generation of control laws to be tested
After applying the formulas and algorithms (control laws) that regulate the changes in the assistance the exoskeleton gives the user, all the results are taken into account to improve the efficiency of the control laws. The method they use is CAM-ES, which imitates biological evolution. In a broad sense, it tends to keep the control laws that are more efficient and it randomly changes or adjusts those which don't produce the expected results of energy consumption by the user.
Optimized control law parameter values. Optimized values varied widely across participants. Values are normalized to their allowable range (26). Lines are medians, boxes cover the 25th to 75th percentiles, and whiskers show the range.
Control parameters are a computed and customized for each individual participant, see Figure 1. The box-whisper plots for peak torque and peak time indicate that patients have similar optimal parameters (the boxes are much tighter), while rise time and fall time have larger ranges which displays the diversity between participants.
Steady-state metabolic energy cost is estimated for each control law by fitting a first-order dynamical model to 2 min of transient metabolic data (fig. S1)
Using the frequency and volume of respiration, CO2 production and oxygen consumption, the authors formulated a model (with first order differential equations) that estimates the metabolic rate.
gradient descent
Gradient descent is an optimization strategy to find the minimum of a function.
Energy cost reductions ranged from 14.2 to 37.9% (fig. S4 and table S3), with an average reduction of 24.2 ± 7.4% (t test, P = 1 × 10−6, n = 11; Fig. 3A).
The amount of energy saved with the effective control laws ranges between 14.2% to 37.9%. The average amount of energy saved with the new, optimized, stride (a result of the control laws, see Figure 1) is 24.2 +/- 7.4%. The reported values are statistically significant, meaning that there is a significant difference between having torque applied to stride and having no torque at all.
Optimized assistance reduced the metabolic cost of walking to 2.16 ± 0.38 W kg−1, down from 2.84 ± 0.40 W kg−1 with zero torque (mean ± standard deviation).
The amount of energy spent was reduced (2.16 +/- 0.40 W/kg ) when torque was applied as compared to energy spent with no torque (2.84+/- 0.40 W/kg). The "+/- 0.40" shows the range of energy values where the actual energy value must exist. For example, for no torque (2.84), the actual energy value is between 2.44 (minus portion) and 3.24.
line search
Line search is an optimization strategy to find the minimum of a function.
We optimized assistance for 11 participants (subjects 1 to 11; table S1) as they walked on a treadmill at a normal speed (1.25 m s−1).
Eleven human volunteers were exposed to the flow chart and procedures in Figure 1 and the respective data was used to optimize the performance per patient. The treadmill speed at which they walked (and data was collected) was 1.25 meters per second (or approximately 4.1 feet per second).
defined by a control law, while metabolic rate is measured
The control laws are formulas that can be used to calculate and correct for the difference within the expected results and the actual results obtained. In this case, the difference between the expected energy consumption (metabolic rate) and the results obtained are considered in the control law that determines the changes in the pattern in which the device will assist the user.
metabolic energy cost,
Metabolism is how the body will use "fuel", such as sugar, for the cell activity. Metabolic energy cost referes to the amount of energy needed for a specific metabolic task in an organism.
Ankle torque was determined by four parameters: peak torque, timing of peak torque, and rise and fall times (Fig. 2A) (26)
As determined by the authors, four parameters were used to measure optimal torque applied to stride. These parameters use metabolic data from respiratory measurements and incorporate them into control laws to determine the best stride for each individual, with the goal of reducing energy expense.
Ankle exoskeleton. Drive rope tension caused the device to push on the shank, heel, and toe contact, generating an ankle torque.
Real experimental set up: tether is linked to controller, shank strap attaches to human calf, drive rope, heel rope, and load cell provide torque to leg. The joint encoder receives update from controller to provide optimal torque to human leg.
Exoskeleton emulator system. Off-board motor and control hardware actuated a tethered exoskeleton worn on one ankle while participants walked on a treadmill.
Figure C displays the experimental procedure to measure stride. Respirometry measures the metabolic rate, tether connects the device to the motor and eventually updates the control (optimizer algorithm is located here), control feeds back to exoskeleton and provides torque to test sample (human) to optimize posture. Finally, the treadmill is where human performance is measured.
Examples of possible torque patterns.
The figure displays various test subjects (human strides) that are displayed in different colors with respective strides. For example, the orange line displays small torque and long stride period. Green displays high torque and short stride period.
humanoid robots
Humanoid robot is a robot with a body shaped like the human body.
Parameterization of ankle torque. Each control law determined applied torque as a function of time, normalized to stride period, as a cubic spline defined by peak time, rise time, fall time, and peak torque.
The amount of torque (a type of force) as a function of time. The figure measures four parameters (peak time, rise time, fall time, and peak torque) that the authors believe to be important to determine the optimal stride posture, thus improving the amount of energy spent walking.
By using indirect calorimetry to measure metabolic rates, the authors were able to adjust the torque provided by the device while users were walking, running, and carrying a load
The authors used gas exchange rate of oxygen and CO2 of the users breathing (indirect calorimetry) to measure how much energy was used when walking with their device, the ankle exoskeleton. The authors adjusted the device to reduce this energy input from the users for three activities: walking, running and carrying a load.
A method for minimizing the energy cost of human walking, in which various control laws are applied, metabolic (met.) rate is quickly estimated (est.) for each, costs are compared, and an evolution strategy is used to generate a new set of control laws to be tested, all during walking.
The goal of the optimizer is to reduce the amount of energy spent walking. This is accomplished by measuring respiratory data (metabolic data) and posture from the test subject (human) and updating the device to improve posture when walking.
Measurements of human performance are used to update device control so as to improve performance in the human portion of the system.
A human test is measured for metabolic rate, the minimal energy expenditure that mammals burn at rest, and the results are fed into the optimizing coder (exoskeleton). The optimizer chooses the best measurements and updates the device, which then feeds back into the human.