Lower limb exoskeletons and exosuits are worn in parallel with the body to assist, augment, or otherwise affect mobility. Collectively referred to as “exos,” exoskeletons and exosuits are often used in gait rehabilitation or as a mobility aid.1 Exos deliver mechanical assistance to targeted biological joint(s) and, thus, interact directly with the wearer’s musculoskeletal system. To effectively unload or augment the biological joint, knowledge of joint biomechanics is typically incorporated into the design or control of the exo. Generally speaking, the research field has a good comprehension of lower limb joint-level biomechanics, including: dynamic (quasi-) stiffness2–6; distribution of work across the joints7–9; and how factors such as speed, terrain, and load change the joint dynamics.10–12 However, skeletal muscles are the actual actuators that generate movement.13,14 Therefore, there are clear advantages to movement support technology that interacts at the muscle level.
The quantity of published research that combines muscle or tendon biomechanics with exos is small but exhibits a slow upward trends in recent years. Figure 1 illustrates the long-standing and increasing interest in muscle and tendon biomechanics research. The increase in exo research is comparatively more recent, likely due to technological advancements improving the feasibility of exo research.15 Emerging in 2005, published research combining the fields of exos and muscle or tendon biomechanics has slowly increased, although it still trails behind both of its components.
—The number of studies published over time on exos, muscle or tendon biomechanics, and both exos and muscle or tendon biomechanics from 1965 to 2022. The data were found through the PubMed database (https://pubmed.ncbi.nlm.nih.gov). All data searches were in English and for lower limb extremities (KEYWORD: gait OR walking OR (lower AND (limb OR extremities))) in human (KEYWORD: human OR participant OR patient). Specific data searches were (A) exos (KEYWORD: (exoskelet * OR exosuit)), (B) muscle or tendon biomechanics (KEYWORD: (muscle mechanics) OR ((muscle OR tendon) AND (biomechanics))), and (C) muscle or tendon biomechanics and exos (KEYWORD: (muscle mechanics) OR ((muscle OR tendon) AND (biomechanics)) AND (exoskelet * OR exosuit)).
Citation: Journal of Applied Biomechanics 39, 5; 10.1123/jab.2023-0046
Understanding of muscle-level biomechanics has improved with time, facilitated by improvements in technology and modeling. Well-established techniques, like ultrasound, were able to capture muscle mechanics in intact humans in vivo during stationary tasks.16–18 Improvements in technology allowed capture of muscle mechanics in dynamic situations, such as walking.19–21 Advances in high-density electromyography (HD-EMG) and blind source separation enabled measuring the firing activity of contractile microstructures (ie, motor units) in human muscles in vivo,22,23 which was critical to understanding how whole-body movement is accomplished via fiber contraction at the microscale.24 Computational muscle-level models combined with a rise in open-source platforms and models, and an increase in computing power, also served to advance our understanding of muscle-level biomechanics at the macroscale and microscale. Improvements in implementing trajectory optimization for musculoskeletal modeling facilitated faster modeling and simulations, which allowed researchers to more easily simulate and investigate numerous models.25 In 2015, a modeling toolbox called Calibrated EMG-Informed NeuroMusculoSkeletal integrated EMG-driven and EMG-informed algorithms into the musculoskeletal modeling environment of OpenSim.26 The resulting open-source neuromusculoskeletal modeling platform was able to better simulate, test, and understand how muscle activity controls movement. Researchers have also developed tools to facilitate advanced optimal control (Moco in OpenSim) and customizable control (Simulated Controller Optimization Environment [SCONE]) in neuromuscular simulations.27,28 Furthermore, recent advancements have combined computationally fast musculoskeletal models with reinforcement learning to explore muscle mechanics during real-life tasks. MyoSuite, an open-source framework with reinforcement learning-capable musculoskeletal models, enables in silico design of robotic devices and controllers, thereby speeding up the design of personalized exos for in vivo real-world applications.29,30 With a better understanding of muscle-level biomechanics, we gleaned insights into the mechanisms of mobility and biomechanical principles about which exoskeletons can be designed and controlled.
Traditionally, the key players in exo design and control have been mechatronic and actuation aspects, whereas the human aspects have often been disregarded or minimally incorporated. Many exoskeletons, both old and new, exclusively use techniques other than muscle-level biomechanics for their controllers and design. A number of review articles discuss the state of the art for nonmuscle-level exoskeleton control and design,1,31–33 so we only briefly summarize them here. Generally, exo design and control focuses on the robotic part of the human–robot system, placing high importance on the output torque of the device (Figure 2). Oftentimes, exos use position or torque tracking to enforce a predefined kinematic or kinetic profile, and the human must yield and adapt to the robot’s actuation rather than working symbiotically.34–37 Impedance control considers the interaction between human and robot, resulting in a controller that is responsive to the human while still enforcing predetermined dynamics.37–40 Volitional controllers, such as proportional myoelectric controllers, further shift the controller focus from the robot to the human component of the human–robot system by using human measures to inform the output.41–43 However, proportional myoelectric controllers measure the emergent behavior of the underlying muscle actions and, therefore, do not fully capture the muscle mechanics. Human-in-the-loop optimization acts to tune the exo parameters to optimize for a chosen output parameter, which is often human energy expenditure or muscle activity.44–46 Although a powerful and state-of-the-art technique, human-in-the-loop optimization can be time intensive and risks falling into local minima traps. Some consideration to muscle properties has also been included in design—most noticeably exos that employ contractile elements mimicking musculoskeletal properties, such as the McKibben pneumatic actuator (commonly referred to as an “artificial muscle”).41,47 However, McKibben actuators require a compressed air source, thus limiting the locations that the exo can be used in. In addition to little incorporation of muscle-level biomechanics in exo design and control, no exo has taken into consideration the short- and long-term response of skeletal muscle to robot-delivered mechanical loads. Inclusion of muscle-level biomechanics may overcome some shortfalls in current exo design and control.
—Visual representation of the “focus” of high-level exo controllers in terms of the robot–human system. Controllers on the left of the scale prioritize the robot part of the system, whereas the controllers to the right of the scale focus on the human part.
Citation: Journal of Applied Biomechanics 39, 5; 10.1123/jab.2023-0046
The design and control of exos has been increasingly inspired by muscle-level biomechanics over time. Earlier exoskeletons had heavy and bulky form factors, with a focus on providing enough power to compensate for 100% of biological joint torque-generating capacity.48,49 Time and experience led to smaller, and less-bulky, exos with a wide variety of actuation methods. The development of light-weight yet strong materials, more compact powerful batteries and motors, and innovative design all contributed to reducing the bulk and weight of exos. Artificial pneumatic muscles that attempted to mimic biological muscle properties and were originally developed for prosthesis use became a popular actuation method for exos.41,47,50,51 Linear springs, acting in-line with muscle–tendon units, such as the triceps surae complex, have also been used in exo design as a form of musculotendon mimicry.52,53 Straps and cables, a more recent form of exo actuation, are often inspired by biological tendons.54,55 Although these techniques are guided by muscle properties, none of them rely on muscle-level biomechanics, which is the next logical step for exo development.
Inclusion of muscle-level biomechanics in exo design/control has the potential to break through some of the barriers to widespread exo success. Despite technological advancements, there is substantial room for improvement in exo design/control, particularly for real-world environments. In this narrative review, we will discuss how muscle-level biomechanics has been and could be effectively implemented in exo control and design. We introduce the techniques and hardware used to understand muscle-level biomechanics and then discuss how muscle-level biomechanics can inform design and personalization. Next, we consider control choices and integrating real-time muscle-level biomechanics estimators within robotic control logics. After presenting how muscle-level biomechanics affects the human–machine interaction, we end the review with recommendations and predictions of research within muscle-level biomechanics and lower limb exos. Throughout the manuscript, we classified techniques as “direct” if they measured a muscle property through sensors and “indirect” if the technique used a musculoskeletal model with or without sensor input to estimate a muscle property.
Measuring Muscle Biomechanics
Incorporation of muscle biomechanics into exo design, control, or personalization requires a method to measure or estimate muscle biomechanics. The measurement or estimation technique usually needs to provide relevant biomechanical data in vivo, in real time (not for evaluation purposes), and during dynamic movement. Furthermore, sensors should be compatible with the physical structure of the exo. In this section, we reviewed muscle measurement methods and sensors to measure or estimate muscle biomechanics that meet the requirements for use with exoskeletons.
Direct Measures
Although various sensing methods measure parameters that can provide information about muscle-level kinetics and kinematics, the depth and accuracy of measurements are different. We present the methods in order from the deepest to most superficial measurements.
Sarcomere Microendoscopy
Sarcomere microendoscopy can directly measure the second-harmonic frequencies of light generated in the muscle fibers to visualize muscle sarcomeres and their contractile dynamics. Llewellyn et al56 showed that second-harmonic generation with 920-nm illumination can effectively visualize sarcomeres in human extensor digitorum muscle in vivo. They showed that visualizing sarcomere contractile dynamics in millisecond-scale resolution can overcome cardiac and respiratory motion artifacts, which is possible through high-speed data acquisition.
As sarcomere length determines force production capacity in muscles, microendoscopy can reveal muscle biomechanics in vivo while being minimally invasive. Sanchez et al57 developed a wearable microendoscope to visualize sarcomere twitch dynamics in individual motor units of major skeletal muscles, including soleus and vastus lateralis. Although sarcomere microendoscopy is still an invasive method, and it has never been used in conjunction with exoskeletons, it is a potential measurement method that can deepen the knowledge of muscle biomechanics to inform design, control, and individualization of lower limb exoskeletons. For instance, because many movement disorders exhibit or originate from disruptions in sarcomere structure and performance, rehabilitative exos could be designed based on how they affect motor unit contractile dynamics and the structure of sarcomeres in the short and long term. Similarly, different types of exos can take advantage of the evaluation of user adaptation and how adaptation results in desired or undesired remodeling in the muscle, which is advantageous for real-world applications.
Sarcomere microendoscopy demonstrates how joint angle affects sarcomere length and contractile dynamics. Cromie et al58 measured sarcomere length in carpi radialis brevis while the wrist was in flexion and extension. The results showed substantial sarcomere length variability in an individual fiber. More studies addressed how joint angle changes sarcomere length and affects muscle force-generating capacity in lower limbs.59,60 As the changes in the sarcomere length at different angles were in the range of 2 to 4 μm for the soleus, and the measurement precision of the method is ∼30 nm, the measurement is, therefore, suitable for tracking the length of sarcomeres in lower limb movement.60 In addition, Lichtwark et al61 showed that fascicle length change can represent sarcomere length change.
Furthermore, sarcomere microendoscopy can equip research in the field of lower limb exos to investigate muscle adaptation. Pincheira et al62 determined how 3 weeks of eccentric exercise training changes sarcomeres in biceps femoris long head. They estimated sarcomere length and number as well as fascicle length before and after the training. The results showed an increase in the sarcomere length, although the sarcomere numbers in series did not change, as well as a heterogeneous change in fascicle length.
Ultrasound Imaging
Ultrasonography enables deriving muscle architectural (eg, volume, pennation angle, physiological cross-sectional area) and functional parameters (eg, fascicle kinematics), which are critical to understanding muscle biomechanics. As muscles are in series with tendons, other common biological measurements, like joint kinematics and kinetics or EMG, cannot provide enough details to measure the separate kinetics and kinematics of muscles and tendons without many assumptions and models. Muscle–tendon biomechanics can provide useful information for individualization, fatigue recognition, real-time and overground joint torque estimation, and so on, which can enhance control and design in lower limb exos.
Fukunaga et al63 used real-time ultrasonography to determine fascicle length and pennation angle of human vastus lateralis muscle in vivo and noninvasively at rest and during static contraction. Ultrasonography enabled in vivo measuring of parameters that can inform muscle-level kinetics and kinematics, such as physiological cross-sectional area of muscles,64 investigating differences in longitudinal strain of the Achilles tendon65 as well as studying force–velocity behavior of the medial gastrocnemius muscle–tendon unit.66 Manual tracking of muscle architectures with ultrasonography is subjective and time consuming, which has led to algorithms developed for automated tracking of fascicle length67 for which the accuracy and repeatability is assessed,68 and the source code and standalone version of the semiautomated fascicle tracking algorithm was made open source.69
Recently, studies used muscle ultrasonography during gait to estimate parameters like volitional motion or ability and muscle power, which can be useful for exo control. For instance, Nuckols et al70 used ultrasound imaging to estimate the onset of soleus concentric contraction just before push off, when the soleus begins to generate positive power, in real time during various gait conditions. They showed that an exo control strategy based on muscle positive power could adapt to the individuality of contraction timing (eg, healthy and poststroke patient) and how it changed with changes in gait (eg, speed and incline). Nuckols et al70 also compared gait segmentation based on automated ultrasound detection of the soleus contraction onset with ground reaction forces and found that the error was within 1% of the gait cycle. Other approaches to link human mechanics to exo control strategies include optimization and grid search (ie, parameter sweeping), which explore and determine appropriate exo control parameters for a specific gait condition.71 However, both techniques are time consuming and not necessarily generalizable to differences in gaits and terrains. Direct EMG or matching to kinetics and kinematics of joints are also approaches that try to intuitively adapt to human gait, but their performance is not as good as optimization or grid sweep. One possible reason is that joint kinetics and kinematics and EMG cannot completely determine the dynamic state of muscles that are the real actuators of human gait.
Moreover, muscle architecture can be input into data-driven (eg, using machine learning) or mechanistic (eg, musculoskeletal modeling) models that can estimate motion intention. Jahanandish et al72 performed image enhancement and model fitting to extract ultrasound features of rectus femoris muscle, namely thickness, angle between aponeuroses, pennation angle, fascicle length, and image echogenicity. The echogenicity, defined as the ability to send and receive an ultrasound wave, was measured by averaging the intensity of all pixels in the region of interest. Then, they estimated knee joint angle and angular velocity during nonweight-bearing knee flexion/extension based on the trained model with an average root mean square error value of 7.45° and 0.262 rad/s, which is close to a similar study that used surface electromyography (sEMG). Muscle thickness and image echogenicity showed the highest correlation with both angle and angular velocity. Also, some studies showed that inputs from both sEMG and ultrasound images (fascicle length and pennation angle) can increase accuracy of ankle joint moment prediction by machine learning73 models or neural and musculoskeletal74 models.
Although studies used ultrasound B-mode probes to extract fascicle kinematics based on 2D images, Yan et al75 developed a measurement system to use ultrasound wearable A-mode probes (1D measurement) to estimate the acoustic nonlinearity parameter of skeletal muscles in vivo. The acoustic nonlinearity parameter changed in tissues due to diseases or residual stress, which suggested that similar changes may occur with contraction in muscles. Yan et al75 showed high correlation between the acoustic nonlinearity parameter of biceps brachii muscle and elbow joint torque, with an average coefficient of determination (R2) of 0.861. Hence, they proposed the acoustic nonlinearity parameter as supplementary information for force control of exoskeletons.
Shear Wave Tensiometry
An emerging research technique to directly and noninvasively measure superficial tendon and ligament kinetics is shear wave tensiometry.76 The noninvasive device is placed on the skin directly over the target tendon or ligament. A vibrational stimulus is delivered to the tendon (via the skin), which propagates through the tendon at a speed dependent on the tendon axial load due to real-time characteristics of the tendon modulating the propagation of the vibration. The shear wave speed can be calculated knowing the distance between 2 miniature accelerometers placed along the tendon direction and measuring the arrival time of the propagated signal.
Shear wave tensiometry shows promising capabilities for applications in lower limb exos. First, it is applicable to lower limb exos as the impulses are detectable throughout the gait cycle and can track dynamically varying wave speed.77 Several studies used shear wave tensiometry to track the dynamics of standing balance and measure gait kinetics in clinical and able-bodied populations, including children, adults, and older adults.78–82 Second, in more dynamic tasks like running and jumping, Schmitz et al83 proposed adding redundant accelerometers and using a Kalman filter to mitigate random sensor noise due to the high rate of loading and impact events. Harper et al84 developed a wearable shear wave tensiometer with dynamic range to track unconstrained locomotion. A later work combined this with inertial measurement units85 to measure the work and power output of triceps surae outdoors. Schneebeli et al86 used intraclass correlation coefficient (ICC3.1) to test the test–retest (ICC3.1 .87–.99), intersection (ICC3.1 .75–.93), and intrasession (ICC3.1 .85–.96) reliability of shear wave tensiometry, which shows promising use in clinical and research settings.
High-Density Electromyography
Noninvasive and flexible or woven textile bidimensional grids of electrodes can measure muscle high-density electromyograms (HD-EMG).87 HD-EMG is an interferent electrical signal generated by the superposition of action potentials generated by hundreds of skeletal muscle fibers during contraction. Because skeletal muscle fibers are directly innervated by alpha motor neurons (neural cells residing in the spinal cord), the HD-EMGs carry information about alpha motor neuron activity, which is directly associated to the control of movements.88 Blind source separation can be used to disentangle the interferent HD-EMG and decode both discharge timings of motor neurons as well as the resulting action potentials traveling along innervated skeletal fibers. HD-EMG holds great potential to explore the neuromechanics of movement in individuals wearing exoskeletons. Many studies have addressed HD-EMG in upper limb gesture recognition and exoskeleton controllers, but the literature in lower limb exoskeletons is scarce.
Some studies investigated HD-EMG during locomotion and demonstrated that spatial patterns of electromyograms can reveal how muscles adapt to different locomotion modes. Schlink et al23 compared different signal processing methods to reduce motion artifacts in HD-EMG during human locomotion and proposed canonical correlation analysis filtering during fast walking and running. Schlink et al89 also showed that spatial patterns of electromyograms are heterogeneous and differ among lower limb muscles and locomotion speeds, which could be due to preferential recruitment of faster motor units under greater loads. Moreover, they showed that fatigue alters spatial myoelectric patterns in the medial gastrocnemius during locomotion, whereas lower limb biomechanics remains similar, a potential strategy to avoid overuse injuries.90 Exos can use HD-EMG to recognize fatigue or locomotion mode to tune assistance (eg, timing and magnitude of torque). Also, HD-EMG can reveal muscle recruitment strategies, impairments, and adaptations of users as valuable information for exoskeleton design.
Force Myography
Force myography (FMG) refers to a broad category of methods that noninvasively measure muscle biomechanics by quantifying how muscle external geometry changes. FMG usually includes an array of sensors to detect muscle deformation or stiffness due to contraction as well as methods that process the collected data to extract the desired parameters. FMG can vary in type of sensors (eg, force or pressure sensors, strain sensors, bending sensors); sensor arrangement; and data processing method (eg, signal processing and feature extraction, machine learning techniques), with specific limitations associated with each technique.37 However, sensor characteristics and device configuration can adversely affect the reliability of FMG signals. For instance, many researchers have calibrated the preload forces when the device is donned based on the user’s oral feedback.91 Moreover, some methods are based on cross-sectional area increase of the muscles.92 These methods can be used in some applications, like gesture recognition, but cannot estimate individual muscle force without making assumptions on the contribution of the individual muscles to the cross-sectional area increase. Also, some techniques are only applicable to static situations as they cannot handle motion artifacts.93
Some recent studies developed and used soft and wearable strain sensors to monitor muscle contractions.92,94 Alvarez et al94 used sensors based on strain-mediated contact in anisotropically resistive structures to measure changes in muscle deformation, which correlate with muscle input and knee torque.
Many studies include FMG in hand rehabilitative and assistive exoskeletons to enhance user intention in the upper body,95–99 but FMG in lower limb applications is limited. Jiang et al100 proposed a wearable gait phase determination system based on FMG. The proposed force myography band could correctly detect more than 99.9% of gait phases over 12,965 gait phase segments with an average temporal error of 55.2 milliseconds. With upper limb applications motivating hardware and processing improvements in FMG, there is an increased likelihood of interesting applications for lower limb exos in the future.
Indirect Measures
Aside from the aforementioned direct measurements, real-time musculoskeletal models have been used to estimate biomechanical parameters in vivo, such as muscle–tendon forces,101 kinematics, joint torques,102,103 joint stiffness,5,6 and compressive loads,104 which can be used in exoskeleton controls105–108 without using invasive measurement sensors.
One research direction is to use optimization to estimate individual muscle forces from the net joint torques (often got from inverse dynamics109,110), named static optimization.111–114 Specifically, researchers developed real-time computing platforms that utilize the static optimization method in estimating muscle forces.115,116 However, optimization-based muscle force estimation for lower limbs often requires force plates to measure ground reaction forces, which limits this potential in out-of-lab applications. Trajectory optimization can avoid this issue by modeling the ground contacts; however, the objective function is normally movement-type specific and may not reflect an individual person or patient.25,117 Furthermore, the computation time is long and prevents it from being used in the real-time control of wearable robotic devices.
To solve this problem, muscle force estimation using sEMG-driven musculoskeletal models has been developed, which requires joint angles and sEMG signals as input. At first, due to computational complexity, studies were focused on only one joint and its corresponding muscles,118,119 and an infinitely stiff tendon model was also used in the muscle models to save in computational time.120 Later, these types of studies were expanded to multiple joints and muscles as well as incorporating an elastic tendon to cover more complex movements.102,121 Recently, the open-source Calibrated EMG-Informed NeuroMusculoSkeletal Modeling Toolbox was developed to lower the entry to this field.26,103 Furthermore, experimental EMG paired with inverse dynamics allowed for the personalization of muscle–tendon models to the user as they represent the inputs/outputs pair.26 Not only can this method estimate muscle forces, but studies also demonstrated that it could accurately estimate joint torque and extrapolate results for unknown tasks (not used for the muscle–tendon model parameters personalization) and even other degrees of freedom.122 As musculoskeletal models were directly driven by the sEMG signals and there is an electromechanical delay between sEMG signal and muscle force generation, muscle forces and joint torques could be calculated prior to the actual output of the musculoskeletal system.106,107 Predicting muscle forces and joint torques provides a substantial advantage in exoskeleton control as it gives a windows of opportunity wherein optimal assistance can be provided to the user. By knowing the force of each muscle that wraps over a joint, joint stiffness5,6,123,124 and compressive loads can be estimated.104,125–129 As these parameters are more closely relevant to either human–robot interactions or injury risks, they are more suitable to be directly used as control parameters in exoskeletons. Even though the aforementioned studies have shown very promising results on reproducing joint torques, validating the actual muscle forces in vivo on human subjects poses ethical challenges, making it difficult to directly measure and validate these forces. However, experimental studies conducted on cats have provided valuable insights into the effectiveness of various approaches, such as static optimization, trajectory optimization, and surface EMG.130–133 The similarities in muscle structures and types between humans and cats support the belief that these methods can be reliable in interpreting human movements as well. Although there may be some species-specific differences, the fundamental principles underlying muscle mechanics and control are expected to be comparable. In general, indirect measurement methods provide very useful muscle biometric information using very simple sensor setups, compared with direct methods, which has advantages in real-life applications.
Muscle and Tendon Biomechanics for Exoskeleton Design or Personalization
A main objective of exos is to interact with the human wearing them. Thus, this interaction will influence the design and control of the exo. For a long time, the human neuromusculoskeletal system was ignored in the conception of exoskeletons, resulting in devices that were powerful but bulky and with little to no benefit for the user.49 To overcome this challenge, exoskeleton emulators for iterative design were established, the most established of which is the human-in-the-loop technique.46 Although emulators can result in the creation of optimal design and control,52 they require a large amount of human data (3600 recorded conditions in the cited study).
Direct Measurements
Direct measurements of muscle biomechanics can provide useful information to inspire novel designs or to help tune exoskeletons. Ultrasound imaging can reveal the effects of exo stiffness on force production in muscles and tendons. Farris et al134 evaluated the effect of parallel elastic assistance of ankle exoskeleton in hopping on soleus muscle–tendon unit mechanics using in vivo ultrasound imaging. They showed that parallel assistance to soleus muscle reduces muscle force, but average positive fascicle power does not significantly change. This can help with tuning the stiffness of lower limb exos to optimize metabolic cost. Takahashi et al135 investigated foot–ankle interplay during walking by adding stiffness to the foot through shoes and insoles. They used ultrasound imaging to show that a stiffer foot results in decreased shortening velocity and increased force output in soleus muscle. The results suggested added foot/shoe stiffness as a potential design parameter to improve locomotion economy in tasks where muscles should generate more muscle force (eg, walking with load carriage) or operate with less economy (eg, fast walking speeds close to the walk-to-run transition). Also, Nuckols et al136 showed how exoskeleton stiffness alters soleus muscle contractile dynamics and affects metabolic rate during walking. They used ultrasound imaging to show that exoskeletons with higher rotational stiffness increase fascicle length and velocity and decrease fascicle force. However, the change in contractile dynamics results in a bowl-shape metabolic cost. Hence, measuring contractile behavior can help to design or tune exos to steer more economical force production in the muscles. Moreover, Beck et al137 showed how artificially fast balance-correcting exoskeleton torque can improve balance by 9% and how it affects fascicle mechanics.
Moreover, direct measurements can help exo personalization through design and evaluation of assistance profile. Nuckols et al138 demonstrated that ultrasound imaging can measure muscle dynamics to develop exosuit assistance profiles that are tailored to the individual and adaptive to dynamic walking tasks. Also, Schmitz et al139 used a shear wave tensiometer to directly measure force in the Achilles tendon during walking assisted with ankle exosuit while carrying a load. They performed a pilot experiment in an unconstrained outdoor environment to evaluate 9 different exosuit assistance profiles based on reduction in the peak force in Achilles tendon. The most efficient exosuit assistance profile could reduce metabolic cost by 9.6%.
Indirect Measurements
Human exoskeleton simulation can acquire the same benefits of emulators without requiring a high volume of human data. As previously presented, there is a wealth of literature on neuromusculoskeletal modeling based on cadaveric studies or human experiments. Nevertheless, one of the main and active challenges is the simulation of how human and mechanical devices interact. One of the main axes of simulation research is based on predictive simulation using optimal control, which can reproduce the interaction between the human exoskeleton system and the environment. Optimal control can also predict the effect of exos on humans or human biomechanical variables in unknown conditions. Fournier et al140 began with a biomechanical simulation of healthy and spinal cord injury (SCI) kinematics at varied walking speeds, then added a simulated exoskeleton to the human neuromusculoskeletal system, which allowed them to predict ground reaction forces. Predictive simulation and optimal control can also be used to predict the effect of exos with different mechanical properties; for example, Sreenivasa et al141 looked into the optimal stiffness of an ankle–foot orthosis for reducing muscle effort in children with gait abnormality, and Febrer-Nafría et al142 optimized parameters such as angle shape for the kinematic assistance delivered through an active knee exoskeleton to an SCI patient. In both cases, after testing their simulation with prerecorded human data, they created an optimization routine that tracked real kinematics or torque parameters and/or tried to minimize power, muscle activation, or jerk.
Simulation can also be used to create a set of requirements for exos. Afschrift et al143 investigated a capability gap (torque that a weakened muscle can produce against the torque task requirement computed using inverse dynamics), which allowed identification of the minimum level of assistance required for people with muscle weakness (reduction of maximal force capability in the muscle) to realize certain tasks. This is an effective technique to determine the minimum power of the actuator in the exo for people with muscle weakness. Following the same philosophy, a study simulated ideal actuators to find the best single degree of freedom (DOF) or multi-DOF joint to actuate to minimize metabolic cost during running.144 Results showed that at low running speeds (2 m/s), all joints provided the same metabolic reduction, but at higher speeds, the hip could create better metabolic savings. For multi-DOF, a combination of hip, knee, and ankle offered the highest metabolic reduction. Another interesting simulation study145 showed how coupled joint assistance (using the same actuator/assistance on multiple DOF) had the same or similar effect on metabolic savings as multiactuator assistance and provided better metabolic reduction than assisting only one DOF. This provided important guidelines in exoskeleton design to save on hardware components, that is, going from one actuator per DOF to one actuator crossing multiple joints, thus providing cheaper devices.
There remain 3 main issues and challenges that delay the broad adoption of these simulation tools. The first one, which is common to not only neuromusculoskeletal modeling simulation but also simulators in general, is the gap between simulation and reality, which limits the transferability of simulation results to the real system. The issue of transferability was observed in a study146 where results obtained in simulation promised a higher metabolic reduction (69%) than what was obtained in the real system (25.9%). Using simplified musculoskeletal models (3 DOF 9 muscles model146) can contribute to poor transferability. The second challenge is the idealization of exo components and structures. For instance, exoskeletons are usually modeled as a simplified mechanical structure that is rigidly connected to the human skeletal system, with the force produced by the actuator ideally transmitted directly to the skeletal system without loss due to friction, soft tissue connection, and shear forces.146,147 An interesting study148 tried to tackle part of this problem by implementing a spring and damper contact model between the human and the skeletal system, which resulted in accurately reproduced kinematics of the real exoskeleton system. Nevertheless, their model needed to be calibrated with multiple experimental recordings of the real human and exoskeleton. The third shortcoming is the current inability to simulate the change (or learning) of the neural system due to the exos’ assistance. The lack of modeled learning limits the ability to effectively design assistance for neurorehabilitation of patients such as stroke survivors. An interesting solution would be to replace optimization for simulating the neural command by a machine learning algorithm wherein a neural network trained on the simulation could adapt to the different assistances.149
Muscle Biomechanics for Exoskeleton Control
Muscle biomechanics can provide useful information to estimate valuable control signals for different applications, including fatigue, volitional ability, and joint torque. Understanding generation of joint torques is useful for exoskeleton control. Direct or indirect measurement of joint torque can allow creation of a symbiotic system between the human and exo by providing assistance proportional to the force produced by the user (Figure 3). Although joint torque can be estimated through inverse dynamics, estimation through muscle biomechanics may be more suitable for exoskeleton applications, particularly applications in uncontrolled (nonlaboratory) environments.
—A general exoskeleton controller using a biomechanical model. EMG indicates electromyography.
Citation: Journal of Applied Biomechanics 39, 5; 10.1123/jab.2023-0046
Direct Measurements
Despite attempts and innovations to directly measure muscle biomechanics during gait or other lower limb movements to estimate useful control signals for muscle in the loop controllers (eg, estimating joint torque or fatigue), we could find only one study that actually used a direct measurement of muscle biomechanics to estimate a parameter for the control parameter. Sheng et al150 used real-time ultrasound-based muscle fatigue assessment to robustly control their hybrid (functional electrical stimulation and electrical motor) knee exoskeleton to switch between its modes and avoid extensive stimulation of the fatigued muscle.
Neuromusculoskeletal Modeling
One indirect way of accessing joint torque is by using a neuromusculoskeletal model fed by EMG. Using neuromusculoskeletal modeling in combination with an exoskeleton is a recent technique, which was primarily delayed by the limitation of achieving real-time computation of the complex model (multi-DOF and multimusculotendon unit).109,122 Fleischer and Hommel151 were the first to present the possibility of using a real-time neuromusculoskeletal model for the control of a unilateral knee exoskeleton. Durandau et al105 further developed the methods to apply it to a lower limb exoskeleton for the knee and ankle and tested it on patients (SCI and stroke). The results showed the possibility of using this kind of system for rehabilitation purposes, wherein muscle effort is reduced while also reducing neural control variability. Furthermore, the same system is used for a bilateral ankle exoskeleton during diverse walking conditions,106 showing the possibility to reduce the muscular effort in variable walking speed and showing better adaptability than the current state-of-the-art speed adaptative controller,52,152 although the metabolic reduction is less. Finally, neuromusculoskeletal models offer the possibility to compute biomechanical variables other than joint torque, such as stiffness, which can also benefit robotic control. Yao et al153 combined real-time muscle stiffness computation (tibialis anterior) and joint torque to control a nonambulatory ankle exoskeleton. The joint torque estimation is used as an input command to the torque controller, and the muscle stiffness is used to modulate the stiffness coefficient of the admittance controller used for the torque controller. Nevertheless, this method relies on EMG sensors, which are subject to noise and muscle crosstalk as well as interference from electromechanical devices (ie, an exoskeleton).
Neural control models can be used to drive musculoskeletal models for exoskeletons, which reduces reliance on EMG sensors. Ruiz Garate et al154–156 explored the possibility of using a motor primitive-based neural control model for controlling leg exoskeletons together with musculoskeletal models. They conducted experiments to evaluate the performance of the neuromusculoskeletal model-based controller. A full-leg exoskeleton that had motors at the ankle, knee, and hip was used as the hardware platform. Participants were asked to perform a locomotion track involving ground-level walking, ascending stairs, and descending stairs and several transitions between these tasks. They showed that the assistance significantly decreased time to perform tasks. Dzeladini et al157 showed that a reflex-based neuromuscular controller for an ankle orthosis can reduce the net metabolic cost compared with the transparent mode without disturbing the walking dynamics at slow and normal speeds. Later, they extended the neuromuscular control framework into a hip and knee robotic exoskeleton and tested it on SCI patients.158,159 Their results showed that neuromuscular controller enabled SCI subjects to walk at several speeds, including near healthy speeds, in a healthy-like manner. Shafer et al160 implemented a simple reflex-based neuromuscular controller for an ankle exoskeleton to study the effects of the reflex control parameters, such as the reflex gains and reflex time delay on users. They found that the reflex-based assistance could systematically reduce users’ biological ankle moment; however, it did not reduce their overall metabolic cost. Until now, most studies have looked at the immediate effect on the neuromusculoskeletal system, but long-term effects have not yet been investigated. The direct effect of assistance on the neural system has also been overlooked.
Changes to Muscle Biomechanics When Using Lower Limb Exoskeletons
In this section, we summarize the changes to muscle biomechanics as a result of using exoskeleton-assisted movement. Studies have investigated these changes during actuation of lower limb joints during either walking, seated knee flexion/extension, hopping, squatting, or sit-to-stand tasks. The influence on plantar flexion and dorsiflexion is commonly studied when using ankle exoskeletons and is summarized in Table 1, whereas the following sections address actuation of either the knee or the hip joints. We also briefly discuss the studies that have used commercial lower limb exoskeletons or robotic gait trainers, such as Lokomat and Lopes.
Summary of Changes to Muscle–Tendon Biomechanics When Participants Wore an Ankle Exoskeleton
| Actuation type | Task | Muscle(s) or tendon(s) | Condition varied | Changes to muscle biomechanics |
|---|---|---|---|---|
| Passive | Hopping | Plantar flexors | Ankle joint stiffness | Decreased activity, force, and force rate134,161,162 Operating at less optimal fiber length and increased fiber velocity134,163 Increased fascicle excursion and overall no change in averaged positive fascicle power134 |
| Active | Walking | Soleus | Ankle joint stiffness | Reduced muscle force, increased fascicle length, and velocity136 |
| Plantar flexors | Inclination of walking surface | Increase in inclination results in lower reductions in muscle activity164 | ||
| Plantar flexors | Exoskeleton work | Increasing work reduced activity and increased muscle synergy weights165 | ||
| Soleus | Exoskeleton torque | Increasing torque reduced activity and increased muscle synergy activations165 | ||
| Plantar flexors | Adaptation to walking with exoskeleton | Decreased activity166 | ||
| soleus | Controller type | Gait timing controllers are better than myoelectric controller proportional to soleus activity,167 myoelectric controller proportional to gastrocnemius medialis reduces soleus activity168 | ||
| — | Speed, step length, and walking incline | Increased apparent efficacy with increasing speed or step length,169 which drops when the surface incline increases164 | ||
| Walking with load | Achilles tendon | With and without exo, carrying additional load | Increase in tendon force with increase in load carried but reduced with exo assistance139 | |
| Walking in participants with chronic stroke | Bilateral latissimus dorsi, erector spinae, external oblique, hip flexors and extensors, knee flexors and extensors, and plantar flexors and dorsiflexors | With and without exo | Improved similarity of synergies in either side170 |
Active Actuation of Knee or Hip Joint
Many studies have shown that knee exoskeletons can reduce activity of associated muscles. Some studies that actuated the knee joint provided additional passive supports at either ankle or hip.171–178 The results showed reductions when walking with the powered mode as compared with the minimal impedance mode for the soleus in stance phase,173 the biceps femoris in swing phase,171 and rectus femoris, tibialis anterior, gastrocnemius lateralis, and semitendinosus172 during stance and swing phase. In minimal impedance mode, also called “free” or “transparent” mode, the exo attempts to be transparent to the user and minimally impact walking biomechanics either by providing no torque or actively controlling for minimal interaction torque. Active knee assistance with passive hip support reduced muscle activity of hip and knee extensors.176 This general decrease in EMG activity was further associated with a decreasing trend over the time with assisted walking due to adaptation.176
Studies also evaluated knee exoskeletons in other tasks, such as squatting177 or knee flexion–extension.175,178 A knee exoskeleton also reduced knee extensor muscle activity when using a controller that was capable of injecting the minimal amount of energy needed to support oscillations of the knee.175
Passive springs placed anteriorly on the hip stored and released energy, thereby reducing plantar flexor activity during walking.179 Actuated hip exoskeletons increased muscle activity of tibialis anterior, rectus femoris, and gastrocnemius medialis in the minimal impedance mode, whereas that of the semitendinosus reduced.174 However, providing assistance reduced the muscle activity of the tibialis anterior, rectus femoris, and gastrocnemius medialis but increased activity of the semitendinosus.174
Actuation Across Multiple Joints and Use of Robotic Gait Trainers and Commercial Exoskeletons
An exoskeleton with active knee and ankle actuation reduced muscle activity for both healthy participants and those with stroke in the knee flexors and extensors and in the ankle plantar and dorsiflexors.105 Increased assistance reduced the variability of muscle activity in this study. Lower limb exoskeletons with hip and knee actuation and passive ankle support reduced muscle activity of the vastus medialis, gastrocnemius medialis,180,181 tibialis anterior, rectus femoris, and soleus180 when compared with walking with a minimal impedance mode. Activity of semitendinosus and biceps femoris increased when walking slowly with the exoskeleton.180 Furthermore, muscle synergy patterns were shown to be altered when wearing such exoskeletons.182 Otálora et al183 showed that actuation of the hip, knee, and ankle joints reduced activity in the knee flexors and extensors and in tibialis anterior.
Gait training devices, such as the LOPES and Lokomat, provide different degrees of actuation of the lower limb and, unlike exoskeletons, can also provide body weight support. Walking in the LOPES in zero-impedance mode showed a decrease in muscle activity of the muscles involved in push off but an increase in activity of muscles that contribute to acceleration and deceleration of the swing leg.184,185 The activation timing was rather unchanged.184 The virtual pivot point model implemented in LOPES II through admittance control also showed decreased muscle activity in rectus femoris, hamstring, medial gastrocnemius, and gluteus maximus muscles.185 Literature about how the Lokomat impacted posttraining muscle activity has been inconclusive. The Lokomat imposes able-bodied joint trajectories through impedance control, with varying levels of exo guidance force. At higher levels of guidance force, the joint trajectories are more rigid, and it is increasingly difficult for the muscles of the user to influence or affect the trajectory of the Lokomat. The results depended on guidance level,186,187 speed,186,188 or body weight support.187,189 Walking on the Lokomat increased the activity of biceps femoris, vastus lateralis,186,189 erector spinae, and tibialis anterior189 in healthy participants and increased tibalis anterior, biceps femoris, and rectus femoris activity in participants with stroke.187 However, another study identified a general lowering of muscle activity compared with treadmill walking in participants with stroke as well as healthy walkers.190 For participants with SCI, Lokomat training reduced vastus lateralis and rectus femoris activity during stance phase and increased gastrocnemius medialis during swing phase.191 The Lokomat was also used to differentially assist either side during gait and showed an inverse relation to muscle activity on the other side.188 The activity in the trunk muscles during training with the Lokomat was similar to quiescent supine lying.192 Using a robotic trainer that only assisted the hip increased activity of knee flexors.193
Commercially available exoskeletons have been rigorously studied in literature.192,194–204 For instance, studies reported reduction in rectus femoris,196,197 gluteus medius, medial hamstring, tibialis anterior, soleus,197 gluteus maximus, rectus femoris, vastus lateralis, and gastrocnemius and an increase in biceps femoris196 when walking with the Ekso GT (Ekso Bionics Holdings, Inc), which assists the user in a predefined trajectory. However, the change in muscle activity depended on gait phase, whether the user had voluntary control in the exoskeleton, and the speed of reference walking without exoskeleton.196 Voluntary control increased burst duration compared with fast walking.196
The Ekso has often been used to train participants with neurological impairments.192,194–196,198–203 Participants with multiple sclerosis showed a new muscle synergy when walking with the exoskeleton.203 Incomplete SCI participants showed nonreciprocal firing patterns and reduced muscle activations, especially that of the rectus femoris200 and tibialis anterior.196 A reduced variability in muscle activity was seen when walking with the exoskeleton.200 Alamro et al192 observed an increase in trunk EMG when walking Ekso assisted compared with the Lokomat for SCI participants. Trunk EMG activation remained similar between Ekso overground and treadmill walking.
Stroke participants were studied at all 3 phases—acute, subacute, and chronic—with the Ekso.194,195,198,199,201,202 Participants with acute stroke showed reduction in soleus and rectus femoris activity on the affected side during stance phase with the exoskeleton.194 Vastus lateralis and rectus femoris showed the largest dissimilarity in activation with the exoskeleton on the affected side.194 Using the Ekso for training in subacute stroke improved bilateral symmetry of tibialis anterior and decreased cocontractions of the proximal muscles, suggesting improvement in proximal muscle activity.199 Activation timing for the semitendinosus was improved in the paretic leg after training.201 The Ekso-trained group also showed more physiological motor control in the semitendinosus and rectus femoris.201 Chronic stroke participants who trained with the Ekso showed decreased rectus femoris and increased soleus activity of both sides, reduced activity of unaffected biceps femoris, and an increase in affected biceps femoris during gait.195 Furthermore, during swing phase the rectus femoris activity of the affected side was reduced.198 The synergy modules necessary to reconstruct lower limb muscle activities were more similar to healthy walking post Ekso training.202
This section has overviewed the impact on muscles as a result of assistance provided by exos. In summary, we see a reduction in muscle activity across joints, and tasks, as well as changes to muscle synergies. This understanding must be accounted for when designing novel controllers that interact with the muscle directly.
Discussion and Future Predictions
Exos have been designed with the goal of enhancing human movement. However, current technologies have shown only modest results in healthy individuals and limited clinical impact. A central element hampering progress is that exos do not interact directly with underlying skeletal muscles but, instead, deliver external mechanical loads that directly influence the biological skeletal system. As such, current systems do not consider how biological muscles, tendons, and joints react to mechanically delivered torques. In this context, research in exos has overlooked the effect of robotic assistance, especially at the extreme ends of the spatiotemporal scale (eg, cell growth over months or years).88,205 This is a critical aspect of human–exo physical interaction as skeletal muscles will remodel new structural properties if exposed long enough to exo-delivered torques. The current inability to predict how exo assistance strategies would remodel skeletal muscle structurally hampers progress in longitudinal neurorehabilitation and in day-to-day adoption of wearable assistive technologies.
In the future, we envision research advances aimed at “closing the loop” between exos and human skeletal muscle biology.88 In this context, we envision that future exos should be capable of delivering coordinated mechanical torques to alter, in a controlled way, skeletal muscle form and function over time scales ranging from seconds (eg, a movement cycle) to months (eg, the recovery stage following a neuromuscular injury) and beyond (eg, across aging stages).
We envision this will require developments in 3 key directions: (1) recording both neural and mechanical function underlying muscle contraction (eg, motor neuron discharges and innervated skeletal fiber force) in the intact moving human in vivo; (2) fusing recorded neuromechanical data with numerical models to predict skeletal muscle adaptation in response to exos-delivered torques over time; and (3) developing predictive controllers to steer skeletal muscle form and function with enough certainty to induce a targeted, positive change in the future. This will open up a new class of muscle-centered exos that directly respond to biological cues to maintain integrity of skeletal muscles over the life span.
Acknowledgments
This work has received funding from the European Research Council under the European Union’s Horizon 2020 research and innovation program as part of the European Research Council Starting Grant INTERACT (Grant agreement No. 803035) as well as by the EU RIA Project SOPHIA (Grant agreement No. 871237).
References
- 3.
Shamaei K, Sawicki GS, Dollar AM. Estimation of quasi-stiffness and propulsive work of the human ankle in the stance phase of walking. PLoS One. 2013;8(3):e59935. doi:10.1371/journal.pone.0059935
- 4.
Rouse EJ, Gregg RD, Hargrove LJ, Sensinger JW. The difference between stiffness and quasi-stiffness in the context of biomechanical modeling. IEEE Trans Biomed Eng. 2013;60(2):562–568. PubMed ID: 23212310 doi:10.1109/TBME.2012.2230261
- 5.↑
Cop CP, Cavallo G, van ’t Veld RC, et al. Unifying system identification and biomechanical formulations for the estimation of muscle, tendon and joint stiffness during human movement. Prog Biomed Eng. 2021;3(3):033002. doi:10.1088/2516-1091/ac12c4
- 8.
Zelik KE, Takahashi KZ, Sawicki GS. Six degree-of-freedom analysis of hip, knee, ankle and foot provides updated understanding of biomechanical work during human walking. J Exp Biol. 2015;218(6):876–886. doi:10.1242/jeb.115451
- 9.↑
Montgomery JR, Grabowski AM. The contributions of ankle, knee and hip joint work to individual leg work change during uphill and downhill walking over a range of speeds. R Soc Open Sci. 2018;5(8):180550. doi:10.1098/rsos.180550
- 11.
Voloshina AS, Kuo AD, Daley MA, Ferris DP. Biomechanics and energetics of walking on uneven terrain. J Exp Biol. 2013;216(Pt 21):3963–3970. PubMed ID: 23913951 doi:10.1242/jeb.081711
- 13.↑
Zajac FE. Muscle coordination of movement: a perspective. J Biomech. 1993;26(suppl 1):109–124. doi:10.1016/0021-9290(93)90083-Q
- 15.↑
Sawicki GS, Beck ON, Kang I, Young AJ. The exoskeleton expansion: improving walking and running economy. J Neuroeng Rehabil. 2020;17(1):1–9. doi:10.1186/s12984-020-00663-9
- 17.
Pillen S, van Alfen N. Skeletal muscle ultrasound. Neurol Res. 2013;33(10):1016–1024. doi:10.1179/1743132811Y.0000000010
- 19.↑
Hodson-Tole EF, Lai AKM. Ultrasound-derived changes in thickness of human ankle plantar flexor muscles during walking and running are not homogeneous along the muscle mid-belly region. Sci Rep. 2019;9(1):1–11. doi:10.1038/s41598-019-51510-4
- 20.
Fukunaga T, Kubo K, Kawakami Y, Fukashiro S, Kanehisa H, Maganaris CN. In vivo behaviour of human muscle tendon during walking. Proc R Soc Lond B Biol Sci. 2001;268(1464):229–233. doi:10.1098/rspb.2000.1361
- 25.↑
De Groote F, Falisse A. Perspective on musculoskeletal modelling and predictive simulations of human movement to assess the neuromechanics of gait. Proc Biol Sci. 2021;288(1946):20202432. doi:10.1098/rspb.2020.2432
- 26.↑
Pizzolato C, Lloyd DG, Sartori M, et al. CEINMS: a toolbox to investigate the influence of different neural control solutions on the prediction of muscle excitation and joint moments during dynamic motor tasks. J Biomech. 2015;48(14):3929–3936. PubMed ID: 26522621 doi:10.1016/j.jbiomech.2015.09.021
- 27.↑
Dembia CL, Bianco NA, Falisse A, Hicks JL, Delp SL. OpenSim moco: musculoskeletal optimal control. PLoS Comput Biol. 2020;16(12):e1008493. doi:10.1371/journal.pcbi.1008493
- 28.↑
Geijtenbeek T. SCONE: open source software for predictive simulation of biological motion. J Open Source Softw. 2019;4(38):1421. doi:10.21105/joss.01421
- 29.↑
Caggiano V, Wang H, Durandau G, Sartori M, Kumar V. MyoSuite. A Contact-Rich Simulation Suite for Musculoskeletal Motor Control. Published 2022. Accessed August 8, 2023. https://arxiv.org/abs/2205.13600v1
- 30.↑
Wang H, Caggiano V, Durandau G, Sartori M, Kumar V. MyoSim: fast and physiologically realistic MuJoCo models for musculoskeletal and exoskeletal studies. 2022 International Conference on Robotics and Automation. IEEE; 2022:8104–8111. doi:10.1109/icra46639.2022.9811684
- 31.↑
Anam K, Al-Jumaily AA. Active exoskeleton control systems: state of the art. Procedia Eng. 2012;41:988–994. doi:10.1016/j.proeng.2012.07.273
- 32.
Chen G, Chan CK, Guo Z, Yu H. A review of lower extremity assistive robotic exoskeletons in rehabilitation therapy. Crit Rev Biomed Eng. 2013;41(4–5):343–363. PubMed ID: 37608181 doi:10.1615/CritRevBiomedEng.2014010453
- 33.↑
Ma Y, Wu X, Yi J, Wang C, Chen C. A review on human-exoskeleton coordination towards lower limb robotic exoskeleton systems. Int J Rob Autom. 2019;34(4):431–451. doi:10.2316/j.2019.206-0193
- 34.↑
Kwa HK, Noorden JH, Missel M, Craig T, Pratt JE, Neuhaus PD. Development of the IHMC mobility assist exoskeleton. Proceedings of IEEE International Conference on Robotics and Automation. IEEE; 2009:2556–2562. doi:10.1109/robot.2009.5152394
- 35.
Zhang J, Cheah CC, Collins SH. Experimental comparison of torque control methods on an ankle exoskeleton during human walking. Proceedings of IEEE International Conference on Robotics and Automation. IEEE; 2015;2015-June(June):5584–5589. doi:10.1109/icra.2015.7139980
- 36.
Witte KA, Zhang J, Jackson RW, Collins SH. Design of two lightweight, high-bandwidth torque-controlled ankle exoskeletons. Proceedings of IEEE International Conference on Robotics and Automation. IEEE; 2015;2015-June(June):1223–1228. doi:10.1109/icra.2015.7139347
- 37.↑
Anam K, Al-Jumaily AA. Active exoskeleton control systems: state of the art. Procedia Eng. 2012;41:988–994. doi:10.1016/j.proeng.2012.07.273
- 38.
Aguirre-Ollinger G, Colgate JE, Peshkin MA, Goswami A. Active-impedance control of a lower-limb assistive exoskeleton. IEEE; 2007 IEEE 10th International Conference on Rehabilitation Robotics, ICORR’07; 2007:188–195. doi:10.1109/icorr.2007.4428426
- 39.
Veneman JF, Kruidhof R, Hekman EEG, Ekkelenkamp R, Van Asseldonk EHF, Van Der Kooij H. Design and evaluation of the LOPES exoskeleton robot for interactive gait rehabilitation. IEEE Trans Neural Syst Rehabil Eng. 2007;15(3):379–386. PubMed ID: 17894270 doi:10.1109/TNSRE.2007.903919
- 40.↑
Maggioni S, Reinert N, Lünenburger L, Melendez-Calderon A. An adaptive and hybrid end-point/joint impedance controller for lower limb exoskeletons. Frontiers Robotics AI. 2018;5(OCT):379651. doi:10.3389/frobt.2018.00104/bibtex
- 41.↑
Ferris DP, Lewis CL. Robotic lower limb exoskeletons using proportional myoelectric control. Proceedings of the 31st Annual International Conference of the IEEE Engineering in Medicine and Biology Society: Engineering the Future of Biomedicine, EMBC 2009. Vol 2009. IEEE Computer Society; 2009:2119–2124. doi:10.1109/iembs.2009.5333984
- 42.
Koller JR, Jacobs DA, Ferris DP, Remy CD. Learning to walk with an adaptive gain proportional myoelectric controller for a robotic ankle exoskeleton. J Neuroeng Rehabil. 2015;12(1):1–14. doi:10.1186/s12984-015-0086-5
- 43.↑
Hybart RL, Ferris DP. Neuromechanical adaptation to walking with electromechanical ankle exoskeletons under proportional myoelectric control. IEEE Open J Eng Med Biol. 2023;4:119–128. doi:10.1109/ojemb.2023.3288469
- 44.↑
Xu L, Liu X, Chen Y, et al. Reducing the muscle activity of walking using a portable hip exoskeleton based on human-in-the-loop optimization. Front Bioeng Biotechnol. 2023;11:1006326. doi:10.3389/fbioe.2023.1006326/bibtex
- 45.
Diaz MA, Voss M, Dillen A, et al. Human-in-the-loop optimization of wearable robotic devices to improve human-robot interaction: a systematic review. IEEE Trans Cybern. IEEE; 2022. doi:10.1109/tcyb.2022.3224895
- 46.↑
Zhang J, Fiers P, Witte KA, et al. Human-in-the-loop optimization of exoskeleton assistance during walking. Science. 2017;356(6344):1280–1284. doi:10.1126/science.aal5054
- 47.↑
Klute GK, Czerniecki JM, Hannaford B. McKibben artificial muscles: Pneumatic actuators with biomechanical intelligence. IEEE/ASME International Conference on Advanced Intelligent Mechatronics, AIM. IEEE; 1999:221–226. doi:10.1109/aim.1999.803170
- 48.↑
Dollar AM, Herr H. Lower extremity exoskeletons and active orthoses: challenges and state-of-the-art. IEEE Trans Rob. 2008;24(1):144–158. doi:10.1109/TRO.2008.915453
- 49.↑
Zoss AB, Kazerooni H, Chu A. Biomechanical design of the Berkeley lower extremity exoskeleton (BLEEX). IEEE/ASME Trans Mechatron. 2006;11(2):128–138. doi:10.1109/TMECH.2006.871087
- 50.↑
Wehner M, Quinlivan B, Aubin PM, et al. A lightweight soft exosuit for gait assistance. Proceedings of International Conference on Robotics and Automation. IEEE; 2013:3362–3369. doi:10.1109/icra.2013.6631046
- 51.↑
Zhang JF, Yang CJ, Chen Y, Zhang Y, Dong YM. Modeling and control of a curved pneumatic muscle actuator for wearable elbow exoskeleton. Mechatronics. 2008;18(8):448–457. doi:10.1016/j.mechatronics.2008.02.006
- 53.↑
Etenzi E, Borzuola R, Grabowski AM. Passive-elastic knee-ankle exoskeleton reduces the metabolic cost of walking. J Neuroeng Rehabil. 2020;17(1):1–15. doi:10.1186/s12984-020-00719-w
- 54.↑
Asbeck AT, De Rossi SMM, Holt KG, Walsh CJ. A biologically inspired soft exosuit for walking assistance. Int J Rob Res. 2015;34(6):744–762. doi:10.1177/0278364914562476
- 66.↑
Farris DJ, Sawicki GS. Human medial gastrocnemius force–velocity behavior shifts with locomotion speed and gait. Proc Natl Acad Sci. 2012;109(3):977–982. doi:10.1073/pnas.1107972109
- 68.↑
Gillett JG, Barrett RS, Lichtwark GA. Reliability and accuracy of an automated tracking algorithm to measure controlled passive and active muscle fascicle length changes from ultrasound. Comput Methods Biomech Biomed Engin. 2013;16(6):678–687. PubMed ID: 22235878 doi:10.1080/10255842.2011.633516
- 70.↑
Nuckols RW, Swaminathan K, Lee S, Awad L, Walsh CJ, Howe RD. Automated detection of soleus concentric contraction in variable gait conditions for improved exosuit control. 2020 IEEE International Conference on Robotics and Automation (ICRA). IEEE; 2020:4855–4862. doi:10.1109/icra40945.2020.9197428
- 73.↑
Zhang Q, Clark WH, Franz JR, Sharma N. Personalized fusion of ultrasound and electromyography-derived neuromuscular features increases prediction accuracy of ankle moment during plantarflexion. Biomed Signal Process Control. 2022;71:103100. doi:10.1016/j.bspc.2021.103100
- 75.↑
Yan J, Yang X, Zeng J, Liu H. Acoustic nonlinearity parameter estimation for exoskeleton control. IEEE Trans Med Robot Bionics. 2021;3(4):1002–1010. doi:10.1109/TMRB.2021.3121570
- 76.↑
Martin JA, Brandon SCE, Keuler EM, et al. Gauging force by tapping tendons. Nat Commun. 2018;9(1):1592. doi:10.1038/s41467-018-03797-6
- 78.↑
Acuña SA, Ebrahimi A, Pomeroy RL, Martin JA, Thelen DG. Achilles tendon shear wave speed tracks the dynamic modulation of standing balance. Physiol Rep. 2019;7(23):e14298. doi:10.14814/phy2.14298
- 79.
Ebrahimi A, Martin JA, Schmitz DG, Thelen DG. Shear wave tensiometry reveals an age-related deficit in triceps surae work at slow and fast walking speeds. Front Sports Act Living. 2020;2:69. doi:10.3389/fspor.2020.00069
- 80.
Ebrahimi A, Loegering IF, Martin JA, Pomeroy RL, Roth JD, Thelen DG. Achilles tendon loading is lower in older adults than young adults across a broad range of walking speeds. Exp Gerontol. 2020;137:110966. doi:10.1016/j.exger.2020.110966
- 81.
Ebrahimi A, Kuchler RL, Pomeroy RL, Loegering IF, Martin JA, Thelen DG. Normative Achilles and patellar tendon shear wave speeds and loading patterns during walking in typically developing children. Gait Posture. 2021;88:185–191. PubMed ID: 34098404 doi:10.1016/j.gaitpost.2021.05.023
- 84.↑
Harper SE, Roembke RA, Zunker JD, Thelen DG, Adamczyk PG. Wearable tendon kinetics. Sensors. 2020;20(17):4805. doi:10.3390/s20174805
- 85.↑
Harper SE, Schmitz DG, Adamczyk PG, Thelen DG. Fusion of wearable kinetic and kinematic sensors to estimate triceps surae work during outdoor locomotion on slopes. Sensors. 2022;22(4):1589. doi:10.3390/s22041589
- 86.↑
Schneebeli A, Falla D, Cescon C, Barbero M. Measurement of Achilles tendon loading using shear wave tensiometry: a reliability study. Musculoskelet Sci Pract. 2022;62:102665. doi:10.1016/j.msksp.2022.102665
- 87.↑
Simonetti D, Koopman B, Sartori M. Automated estimation of ankle muscle EMG envelopes and resulting plantar-dorsi flexion torque from 64 garment-embedded electrodes uniformly distributed around the human leg. J Electromyogr Kinesiol. 2022;67:102701. doi:10.1016/j.jelekin.2022.102701
- 88.↑
Sartori M, Sawicki G. Closing the loop between wearable technology and human biology: a new paradigm for steering neuromuscular form and function. Prog Biomed Eng. 2021;3(2):023001. doi:10.1088/2516-1091/abe3e0
- 89.↑
Schlink BR, Nordin AD, Ferris DP. Human myoelectric spatial patterns differ among lower limb muscles and locomotion speeds. Physiol Rep. 2020;8(23):e14652. doi:10.14814/phy2.14652
- 91.↑
Xiao ZG, Menon C. A review of force myography research and development. Sensors. 2019;19(20):4557. doi:10.3390/s19204557
- 92.↑
Zahid M, Zych A, Dussoni S, et al. Wearable and self-healable textile-based strain sensors to monitor human muscular activities. Compos B Eng. 2021;220:108969. doi:10.1016/j.compositesb.2021.108969
- 93.↑
Islam MRU, Bai S. Payload estimation using forcemyography sensors for control of upper-body exoskeleton in load carrying assistance. Model Identif Control. 2019;40(4):189–198. doi:10.4173/mic.2019.4.1
- 95.↑
Dwivedi A, Gerez L, Hasan W, Yang CH, Liarokapis M. A soft exoglove equipped with a wearable muscle-machine interface based on forcemyography and electromyography. IEEE Robot Autom Lett. 2019;4(4):3240–3246. doi:10.1109/LRA.2019.2925302
- 96.
Fajardo J, Neto AR, Silva W, Gomes M, Fujiwara E, Rohmer E. A wearable robotic glove based on optical FMG driven controller. 2019 IEEE 4th International Conference on Advanced Robotics and Mechatronics (ICARM). IEEE; 2019:81–86.
- 97.
Islam MRU, Bai S. Effective multi-mode grasping assistance control of a soft hand exoskeleton using force myography. Front Robot AI. 2020;7:567491. doi:10.3389/frobt.2020.567491
- 98.
Ribas Neto A, Fajardo J, da Silva WHA, et al. Design of tendon-actuated robotic glove integrated with optical fiber force myography sensor. Automation. 2021;2(3):187–201. doi:10.3390/automation2030012
- 99.↑
Song X, van de Ven SS, Chen S, et al. Proposal of a wearable multimodal sensing-based serious games approach for hand movement training after stroke. Front Physiol. 2022;13:811950. doi:10.3389/fphys.2022.811950
- 100.↑
Jiang X, Chu K, Khoshnam M, Menon C. A wearable gait phase detection system based on force myography techniques. Sensors. 2018;18(4):1279. doi:10.3390/s18041279
- 102.↑
Sartori M, Reggiani M, Farina D, Lloyd DG. EMG-Driven forward-dynamic estimation of muscle force and joint moment about multiple degrees of freedom in the human lower extremity. PLoS One. 2012;7(12):e52618. doi:10.1371/journal.pone.0052618
- 104.↑
Moya-Esteban A, van der Kooij H, Sartori M. Robust estimation of lumbar joint forces in symmetric and asymmetric lifting tasks via large-scale electromyography-driven musculoskeletal models. J Biomech. 2022;144:111307. doi:10.1016/j.jbiomech.2022.111307
- 105.↑
Durandau G, Farina D, Asín-Prieto G, et al. Voluntary control of wearable robotic exoskeletons by patients with paresis via neuromechanical modeling. J Neuroeng Rehabil. 2019;16(1):91. doi:10.1186/s12984-019-0559-z
- 106.↑
Durandau G, Rampeltshammer WF, Kooij H, Sartori M. Neuromechanical model-based adaptive control of bilateral ankle exoskeletons: biological joint torque and electromyogram reduction across walking conditions. IEEE Trans Rob. 2022;38(3):1380–1394. doi:10.1109/TRO.2022.3170239
- 107.↑
Lotti N, Xiloyannis M, Durandau G, et al. Adaptive model-based myoelectric control for a soft wearable arm exosuit: a new generation of wearable robot control. IEEE Robot Autom Mag. 2020;27(1):43–53. doi:10.1109/MRA.2019.2955669
- 108.↑
Shah C, Fleming A, Nalam V, Liu M, Huang HH. Design of EMG-driven musculoskeletal model for volitional control of a robotic ankle prosthesis. IEEE International Conference on Intelligent Robots and Systems. IEEE; 2022;2022-October:12261–12266. doi:10.1109/iros47612.2022.9981305
- 112.
Happee R. Inverse dynamic optimization including muscular dynamics, a new simulation method applied to goal directed movements. J Biomech. 1994;27(7):953–960. PubMed ID: 8063845 doi:10.1016/0021-9290(94)90267-4
- 113.
Tsirakos D, Baltzopoulos V, Bartlett R. Inverse optimization: functional and physiological considerations related to the force-sharing problem. Crit Rev Biomed Eng. 1997;25(4–5):371–407. PubMed ID: 9505137 doi:10.1615/CritRevBiomedEng.v25.i4-5.20
- 114.↑
Erdemir A, McLean S, Herzog W, van den Bogert AJ. Model-based estimation of muscle forces exerted during movements. Clin Biomech. 2007;22(2):131–154. doi:10.1016/j.clinbiomech.2006.09.005
- 116.↑
Stanev D, Filip K, Bitzas D, et al. Real-time musculoskeletal kinematics and dynamics analysis using marker- and IMU-based solutions in rehabilitation. Sensors. 2021;21(5):1804. doi:10.3390/s21051804
- 120.↑
Sartori M, Lloyd DG, Reggiani M, Pagello E. Fast operation of anatomical and stiff tendon neuromuscular models in EMG-driven modeling. 2010 IEEE International Conference on Robotics and Automation. IEEE; 2010:2228–2234. doi:10.1109/robot.2010.5509932
- 121.↑
Sartori M, Reggiani M, Lloyd D. An EMG-driven musculoskeletal model of the human lower limb for the estimation of muscle forces and moments at the hip, knee and ankle joints in vivo. Simulation and Programming for Autonomous Robots (SIMPAR). IEEE; 2010:137–146. Accessed October 24, 2022. https://simtk.org/home/nmblmodels/
- 124.↑
Cop CP, Durandau G, Esteban AM, van ’t Veld RC, Schouten AC, Sartori M. Model-based estimation of ankle joint stiffness during dynamic tasks: A validation-based approach. 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC). IEEE; 2019:4104–4107. doi:10.1109/embc.2019.8857391
- 126.
Bazrgari B, Shirazi-Adl A, Arjmand N. Analysis of squat and stoop dynamic liftings: muscle forces and internal spinal loads. Eur Spine J. 2007;16(5):687–699. PubMed ID: 17103232 doi:10.1007/s00586-006-0240-7
- 127.
Kim HK, Zhang Y. Estimation of lumbar spinal loading and trunk muscle forces during asymmetric lifting tasks: application of whole-body musculoskeletal modelling in OpenSim. Ergonomics. 2017;60(4):563–576. PubMed ID: 27194401 doi:10.1080/00140139.2016.1191679
- 128.
Stanev D, Moustakas K. Stiffness modulation of redundant musculoskeletal systems. J Biomech. 2019;85:101–107. PubMed ID: 30709554 doi:10.1016/j.jbiomech.2019.01.017
- 129.↑
von Arx M, Liechti M, Connolly L, Bangerter C, Meier ML, Schmid S. From stoop to squat: a comprehensive analysis of lumbar loading among different lifting styles. Front Bioeng Biotechnol. 2021;9:1070. doi:10.3389/fbioe.2021.769117
- 131.
Karabulut D, Dogru SC, Lin YC, Pandy MG, Herzog W, Arslan YZ. Direct validation of model-predicted muscle forces in the cat hindlimb during locomotion. J Biomech Eng. 2020;142(5):051014. doi:10.1115/1.4045660
- 132.
Kaya M, Leonard T, Herzog W. Coordination of medial gastrocnemius and soleus forces during cat locomotion. J Exp Biol. 2003;206(20):3645–3655. doi:10.1242/jeb.00544
- 135.↑
Takahashi KZ, Gross MT, van Werkhoven H, Piazza SJ, Sawicki GS. Adding stiffness to the foot modulates soleus force-velocity behaviour during human walking. Sci Rep. 2016;6(1):29870. doi:10.1038/srep29870
- 136.↑
Nuckols RW, Dick TJM, Beck ON, Sawicki GS. Ultrasound imaging links soleus muscle neuromechanics and energetics during human walking with elastic ankle exoskeletons. Sci Rep. 2020;10(1):3604. doi:10.1038/s41598-020-60360-4
- 137.↑
Beck ON, Shepherd MK, Rastogi R, Martino G, Ting LH, Sawicki GS. Exoskeletons need to react faster than physiological responses to improve standing balance. Sci Robot. 2023;8(75):eadf1080. doi:10.1126/scirobotics.adf1080
- 138.↑
Nuckols RW, Lee S, Swaminathan K, Orzel D, Howe RD, Walsh CJ. Individualization of exosuit assistance based on measured muscle dynamics during versatile walking. Sci Robot. 2021;6(60):eabj1362. doi:10.1126/scirobotics.abj1362
- 139.↑
Schmitz DG, Nuckols RW, Lee S, et al. Modulation of Achilles tendon force with load carriage and exosuit assistance. Sci Robot. 2022;7(71):eabq1514. doi:10.1126/scirobotics.abq1514
- 141.↑
Sreenivasa M, Millard M, Felis M, Mombaur K, Wolf SI. Optimal control based stiffness identification of an ankle-foot orthosis using a predictive walking model. Front Comput Neurosci. 2017;11:23. doi:10.3389/fncom.2017.00023
- 142.↑
Febrer-Nafría M, Fregly BJ, Font-Llagunes JM. Evaluation of optimal control approaches for predicting active knee-ankle-foot-orthosis motion for individuals with spinal cord injury. Front Neurorobot. 2022;15:748148. doi:10.3389/fnbot.2021.748148
- 143.↑
Afschrift M, De Groote F, De Schutter J, Jonkers I. The effect of muscle weakness on the capability gap during gross motor function: a simulation study supporting design criteria for exoskeletons of the lower limb. Biomed Eng Online. 2014;13(1):111. doi:10.1186/1475-925X-13-111
- 144.↑
Uchida TK, Seth A, Pouya S, Dembia CL, Hicks JL, Delp SL. Simulating ideal assistive devices to reduce the metabolic cost of running. PLoS One. 2016;11(9):e0163417. doi:10.1371/journal.pone.0163417
- 145.↑
Bianco NA, Franks PW, Hicks JL, Delp SL. Coupled exoskeleton assistance simplifies control and maintains metabolic benefits: a simulation study. PLoS One. 2022;17(1):e0261318. doi:10.1371/journal.pone.0261318
- 146.↑
Franks PW, Bianco NA, Bryan GM, Hicks JL, Delp SL, Collins SH. Testing simulated assistance strategies on a hip-knee-ankle exoskeleton: A case study. 2020 8th IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics (BioRob). IEEE; 2020:700–707. doi:10.1109/biorob49111.2020.9224345
- 147.↑
Nguyen VQ, LaPre AK, Price MA, Umberger BR, Sup FC. Inclusion of actuator dynamics in simulations of assisted human movement. Int J Numer Method Biomed Eng. 2020;36(5):e3334. doi:10.1002/cnm.3334
- 149.↑
Wang H, Caggiano V, Durandau G, Sartori M, Kumar V. MyoSim: Fast and physiologically realistic MuJoCo models for musculoskeletal and exoskeletal studies. 2022 International Conference on Robotics and Automation (ICRA). IEEE; 2022:8104–8111. doi:10.1109/icra46639.2022.9811684
- 150.↑
Sheng Z, Iyer A, Sun Z, Kim K, Sharma N. A hybrid knee exoskeleton using real-time ultrasound-based muscle fatigue assessment. IEEE/ASME Trans Mechatron. 2022;27(4):1854–1862. doi:10.1109/TMECH.2022.3171086
- 151.↑
Fleischer C, Hommel G. A human—Exoskeleton interface utilizing electromyography. IEEE Trans Rob. 2008;24(4):872–882. doi:10.1109/TRO.2008.926860
- 152.↑
Kim J, Lee G, Heimgartner R, et al. Reducing the metabolic rate of walking and running with a versatile, portable exosuit. Science. 2019;365(6454):668–672. doi:10.1126/science.aav7536
- 153.↑
Yao S, Zhuang Y, Li Z, Song R. Adaptive admittance control for an ankle exoskeleton using an EMG-driven musculoskeletal model. Front Neurorobot. 2018;12:16. doi:10.3389/fnbot.2018.00016
- 154.↑
Ruiz Garate V, Parri A, Yan T, et al. Walking assistance using artificial primitives: a novel bioinspired framework using motor primitives for locomotion assistance through a wearable cooperative exoskeleton. IEEE Robot Autom Mag. 2016;23(1):83–95. doi:10.1109/MRA.2015.2510778
- 155.
Ruiz Garate V, Parri A, Yan T, et al. Motor primitive-based control for lower-limb exoskeletons. 2016 6th IEEE International Conference on Biomedical Robotics and Biomechatronics (BioRob). Vol 2016-July. IEEE; 2016:655–661. doi:10.1109/biorob.2016.7523700
- 156.↑
Ruiz Garate V, Parri A, Yan T, et al. Experimental validation of motor primitive-based control for leg exoskeletons during continuous multi-locomotion tasks. Front Neurorobot. 2017;11(MAR):15. doi:10.3389/fnbot.2017.00015
- 157.↑
Dzeladini F, Wu AR, Renjewski D, et al. Effects of a neuromuscular controller on a powered ankle exoskeleton during human walking. 2016 6th IEEE International Conference on Biomedical Robotics and Biomechatronics (BioRob). Vol 2016-July. IEEE; 2016:617–622. doi:10.1109/biorob.2016.7523694
- 158.↑
Wu AR, Dzeladini F, Brug TJH, et al. An adaptive neuromuscular controller for assistive lower-limb exoskeletons: a preliminary study on subjects with spinal cord injury. Front Neurorobot. 2017;11(JUN):30. doi:10.3389/fnbot.2017.00030
- 159.↑
Wu AR. Human biomechanics perspective on robotics for gait assistance: challenges and potential solutions. Proc Biol Sci. 2021;288(1956):20211197. doi:10.1098/rspb.2021.1197
- 160.↑
Shafer BA, Philius SA, Nuckols RW, McCall J, Young AJ, Sawicki GS. Neuromechanics and energetics of walking with an ankle exoskeleton using neuromuscular-model based control: a parameter study. Front Bioeng Biotechnol. 2021;9:210. doi:10.3389/fbioe.2021.615358
- 163.↑
Farris DJ, Hicks JL, Delp SL, Sawicki GS. Musculoskeletal modelling deconstructs the paradoxical effects of elastic ankle exoskeletons on plantar-flexor mechanics & energetics during hopping. J Exp Biol. 2014;217(22):4018–4028. doi:10.1242/jeb.107656
- 164.↑
Sawicki GS, Ferris DP. Mechanics and energetics of incline walking with robotic ankle exoskeletons. J Exp Biol. 2009;212(1):32–41. doi:10.1242/jeb.017277
- 167.↑
Koller JR, Remy CD, Ferris DP. Biomechanics and energetics of walking in powered ankle exoskeletons using myoelectric control versus mechanically intrinsic control. J Neuroeng Rehabil. 2018;15(1):42. doi:10.1186/s12984-018-0379-6
- 169.↑
Sawicki GS, Ferris DP. Powered ankle exoskeletons reveal the metabolic cost of plantar flexor mechanical work during walking with longer steps at constant step frequency. J Exp Biol. 2009;212(1):21–31. doi:10.1242/jeb.017269
- 170.↑
Rinaldi L, Yeung LF, Lam PCH, Pang MYC, Tong RKY, Cheung VCK. Adapting to the mechanical properties and active force of an exoskeleton by altering muscle synergies in chronic stroke survivors. IEEE Trans Neural Syst Rehabil Eng. 2020;28(10):2203–2213. PubMed ID: 32804652 doi:10.1109/TNSRE.2020.3017128
- 172.↑
Thangavel P, Vidhya S, Li J, Chew E, Bezerianos A, Yu H. Biomechanical effects of robot assisted walking on knee joint kinematics and muscle activation pattern. 2017 International Conference on Rehabilitation Robotics (ICORR). IEEE; 2017:252–257. doi:10.1109/icorr.2017.8009255
- 173.↑
Malcolm P, Galle S, Derave W, De Clercq D. Bi-articular knee-ankle-foot exoskeleton produces higher metabolic cost reduction than weight-matched mono-articular exoskeleton. Front Neurosci. 2018;12(March):69. doi:10.3389/fnins.2018.00069
- 174.↑
Grazi L, Crea S, Parri A, Molino Lova R, Micera S, Vitiello N. Gastrocnemius myoelectric control of a robotic hip exoskeleton can reduce the user’s lower-limb muscle activities at push off. Front Neurosci. 2018;12(FEB):71. doi:10.3389/fnins.2018.00071
- 175.↑
Tagliamonte NL, Valentini S, Sudano A, et al. Switching assistance for exoskeletons during cyclic motions. Front Neurorobot. 2019;13(June):41. doi:10.3389/fnbot.2019.00041
- 176.↑
Haufe FL, Kober AM, Wolf P, Riener R, Xiloyannis M. Learning to walk with a wearable robot in 880 simple steps: a pilot study on motor adaptation. J Neuroeng Rehabil. 2021;18(1):157. doi:10.1186/s12984-021-00946-9
- 177.↑
Wang Z, Wu X, Zhang Y, et al. A semi-active exoskeleton based on EMGs reduces muscle fatigue when squatting. Front Neurorobot. 2021;15(April):625479. doi:10.3389/fnbot.2021.625479
- 178.↑
Soma Y, Mutsuzaki H, Yoshioka T, et al. Single-joint hybrid assistive limb in knee rehabilitation after ACL reconstruction: an open-label feasibility and safety trial. Prog Rehabil Med. 2022;7:20220036. doi:10.2490/prm.20220036
- 179.↑
Haufe FL, Wolf P, Riener R, Grimmer M. Biomechanical effects of passive hip springs during walking. J Biomech. 2020;98:109432. doi:10.1016/j.jbiomech.2019.109432
- 180.↑
Sylos-Labini F, La Scaleia V, d’Avella A, et al. EMG patterns during assisted walking in the exoskeleton. Front Hum Neurosci. 2014;8(JUNE):423. doi:10.3389/fnhum.2014.00423
- 181.↑
Sado F, Yap HJ, Ghazilla RAR, Ahmad N. Exoskeleton robot control for synchronous walking assistance in repetitive manual handling works based on dual unscented Kalman filter. PLoS One. 2018;13(7):e0200193. doi:10.1371/journal.pone.0200193
- 182.↑
Li Z, Liu H, Yin Z, Chen K. Muscle synergy alteration of human during walking with lower limb exoskeleton. Front Neurosci. 2019;12(January):1050. doi:10.3389/fnins.2018.01050
- 183.↑
Otálora S, Ballen-Moreno F, Arciniegas-Mayag L, Cifuentes CA, Múnera M. Biomechanical effects of adding an ankle soft actuation in a unilateral exoskeleton. Biosensors. 2022;12(10):873. doi:10.3390/bios12100873
- 184.↑
van Asseldonk EHF, Veneman JF, Ekkelenkamp R, Buurke JH, van der Helm FCT, van der Kooij H. The effects on kinematics and muscle activity of walking in a robotic gait trainer during zero-force control. IEEE Trans Neural Syst Rehabil Eng. 2008;16(4):360–370. PubMed ID: 18713676 doi:10.1109/TNSRE.2008.925074
- 185.↑
Zhao G, Sharbafi M, Vlutters M, van Asseldonk E, Seyfarth A. Template model inspired leg force feedback based control can assist human walking. 2017 International Conference on Rehabilitation Robotics (ICORR). IEEE; 2017:473–478. doi:10.1109/icorr.2017.8009293
- 186.↑
Moreno JC, Barroso F, Farina D, et al. Effects of robotic guidance on the coordination of locomotion. J Neuroeng Rehabil. 2013;10(1):79. doi:10.1186/1743-0003-10-79
- 188.↑
Weiland S, Smit IH, Reinders-Messelink H, van der Woude LHV, van Kammen K, den Otter R. The effect of asymmetric movement support on muscle activity during Lokomat guided gait in able-bodied individuals. PLoS One. 2018;13(6):e0198473. doi:10.1371/journal.pone.0198473
- 189.↑
Van Kammen K, Boonstra A, Reinders-Messelink H, den Otter R. The combined effects of body weight support and gait speed on gait related muscle activity: a comparison between walking in the Lokomat exoskeleton and regular treadmill walking. PLoS One. 2014;9(9):e107323. doi:10.1371/journal.pone.0107323
- 190.↑
van Kammen K, Boonstra AM, van der Woude LHV, Reinders-Messelink HA, den Otter R. Differences in muscle activity and temporal step parameters between Lokomat guided walking and treadmill walking in post-stroke hemiparetic patients and healthy walkers. J Neuroeng Rehabil. 2017;14(1):32. doi:10.1186/s12984-017-0244-z
- 191.↑
Hornby TG, Kinnaird CR, Holleran CL, Rafferty MR, Rodriguez KS, Cain JB. Kinematic, Muscular, and metabolic responses during exoskeletal-, elliptical-, or therapist-assisted stepping in people with incomplete spinal cord injury. Phys Ther. 2012;92(10):1278–1291. PubMed ID: 22700537 doi:10.2522/ptj.20110310
- 192.↑
Alamro RA, Chisholm AE, Williams AMM, Carpenter MG, Lam T. Overground walking with a robotic exoskeleton elicits trunk muscle activity in people with high-thoracic motor-complete spinal cord injury. J Neuroeng Rehabil. 2018;15(1):109. doi:10.1186/s12984-018-0453-0
- 193.↑
Son C, Lee A, Lee J, et al. The effect of pelvic movements of a gait training system for stroke patients: a single blind, randomized, parallel study. J Neuroeng Rehabil. 2021;18(1):185. doi:10.1186/s12984-021-00964-7
- 194.↑
Androwis GJ, Pilkar R, Ramanujam A, Nolan KJ. Electromyography assessment during gait in a robotic exoskeleton for acute stroke. Front Neurol. 2018;9(AUG):630. doi:10.3389/fneur.2018.00630
- 195.↑
Calabrò RS, Naro A, Russo M, et al. Shaping neuroplasticity by using powered exoskeletons in patients with stroke: a randomized clinical trial. J Neuroeng Rehabil. 2018;15(1):35. doi:10.1186/s12984-018-0377-8
- 197.↑
Swank C, Wang-Price S, Gao F, Almutairi S. Walking with a robotic exoskeleton does not mimic natural gait: a within-subjects study. JMIR Rehabil Assist Technol. 2019;6(1):e11023. doi:10.2196/11023
- 199.↑
Infarinato F, Romano P, Goffredo M, et al. Functional gait recovery after a combination of conventional therapy and overground robot-assisted gait training is not associated with significant changes in muscle activation pattern: an EMG preliminary study on subjects subacute post stroke. Brain Sci. 2021;11(4):448. doi:10.3390/brainsci11040448
- 201.↑
Longatelli V, Pedrocchi A, Guanziroli E, Molteni F, Gandolla M. Robotic exoskeleton gait training in stroke: an electromyography-based evaluation. Front Neurorobot. 2021;15(November):150. doi:10.3389/fnbot.2021.733738
- 202.↑
Zhu F, Kern M, Fowkes E, et al. Effects of an exoskeleton-assisted gait training on post-stroke lower-limb muscle coordination. J Neural Eng. 2021;18(4):046039. doi:10.1088/1741-2552/abf0d5
- 204.↑
Tan CK, Kadone H, Watanabe H, et al. Differences in muscle synergy symmetry between subacute post-stroke patients with bioelectrically-controlled exoskeleton gait training and conventional gait training. Front Bioeng Biotechnol. 2020;8(July):770. doi:10.3389/fbioe.2020.00770