Therapeutic and assistive exoskeletons and exosuits show promise in both clinical and real-world settings. Improving their autonomy can enhance usability, effectiveness, and cost efficiency. This Review presents a generic control framework for autonomous operation of upper and lower limb devices and reviews current advancements and future directions. We highlight how data-driven machine learning aids in intention recognition, synchronization, patient assessment, and task-agnostic control. In addition, we discuss how reinforcement learning optimizes control policies through digital human twins and how generative AI supports therapy planning and patient engagement. Richer patient-specific data and more accurate digital twins are needed for clinical validation and widespread deployment.

Exoskeleton balance support can improve the ability to counteract perturbations. The process of human adaptation to this support, however, remains unclear. Here, we assessed how able-bodied individuals adapted to balance support provided by an ankle exoskeleton during walking, specifically when counteracting forward-directed pushes at the pelvis. Activation of the balance support led to immediate and clear reductions in both Center Of Mass (COM) displacement and soleus EMG activity. Further adaptations were observed across the first 35 perturbations for COM displacement and only across the first 5 perturbations for EMG activity before reaching a stable value. These findings demonstrate that adaptation to balance support is a rapid process. These results indicate that minimal training time is required for an individual to effectively utilize exoskeleton balance support.

Background: Spasticity, i.e. stretch hyperreflexia, increases joint resistance similar to symptoms like hypertonia and contractures. Botulinum neurotoxin-A (BoNT-A) injections are a widely used intervention to reduce spasticity. BoNT-A effects on spasticity are poorly understood, because clinical measures, e.g. modified Ashworth scale (MAS), cannot differentiate between the symptoms affecting joint resistance. This paper distinguishes the contributions of the reflexive and intrinsic pathways to ankle joint hyper-resistance for participants treated with BoNT-A injections. We hypothesized that the overall joint resistance and reflexive contribution decrease 6 weeks after injection, while returning close to baseline after 12 weeks. Methods: Nine participants with spasticity after spinal cord injury or after stroke were evaluated across three sessions: 0, 6 and 12 weeks after BoNT-A injection in the calf muscles. Evaluation included clinical measures (MAS, Tardieu Scale) and motorized instrumented assessment using the instrumented spasticity test (SPAT) and parallel-cascade (PC) system identification. Assessments included measures for: (1) overall resistance from MAS and fast velocity SPAT; (2) reflexive resistance contribution from Tardieu Scale, difference between fast and slow velocity SPAT and PC reflexive gain; and (3) intrinsic resistance contribution from slow velocity SPAT and PC intrinsic stiffness/damping. Results: Individually, the hypothesized BoNT-A effect, the combination of a reduced resistance (week 6) and return towards baseline (week 12), was observed in the MAS (5 participants), fast velocity SPAT (2 participants), Tardieu Scale (2 participants), SPAT (1 participant) and reflexive gain (4 participants). On group-level, the hypothesis was only confirmed for the MAS, which showed a significant resistance reduction at week 6. All instrumented measures were strongly correlated when quantifying the same resistance contribution. Conclusion: At group-level, the expected joint resistance reduction due to BoNT-A injections was only observed in the MAS (overall resistance). This observed reduction could not be attributed to an unambiguous group-level reduction of the reflexive resistance contribution, as no instrumented measure confirmed the hypothesis. Validity of the instrumented measures was supported through a strong association between different assessment methods. Therefore, further quantification of the individual contributions to joint resistance changes using instrumented measures across a large sample size are essential to understand the heterogeneous response to BoNT-A injections.

Spatiotemporal gait characteristics change during very slow walking, a relevant speed considering individuals with movement disorders or using assistive devices. However, we lack insights in how very slow walking affects human balance control. Therefore, we aimed to identify how healthy individuals use balance strategies while walking very slow. Ten healthy participants walked on a treadmill at an average speed of 0.43 m s⁻¹, while being perturbed at toe off right by either perturbations of the whole-body linear momentum (WBLM) or angular momentum (WBAM). WBLM perturbations were given by a perturbation on the pelvis in forward or backward direction. The WBAM was perturbed by two simultaneous perturbations in opposite directions on the pelvis and upper body. The given perturbations had magnitudes of 4, 8, 12 and 16 % of the participant's body weight, and lasted for 150 ms. After perturbations of the WBLM the centre of pressure placement was modulated using the ankle joint, while keeping the moment arm of the ground reaction force (GRF) with respect to the centre of mass (CoM) small. After the perturbations of the WBAM a quick recovery was initiated, using the hip joint and adjusting the horizontal GRF to create a moment arm with respect to the CoM. These findings suggest no fundamental differences in the use of balance strategies at very slow walking compared to normal speeds. Still as the gait phases last longer, this time was exploited to counteract perturbations in the ongoing gait phase.

Series elastic actuators (SEA) with their inherent compliance offer a safe torque source for robots that are interacting with various environments, including humans. These applications have high requirements for the SEA torque controllers, both in the torque response as well as interaction behavior with its environment. To differentiate state of the art torque controllers, this work introduces a unifying theoretical and experimental framework that compares controllers based on their torque transfer behavior, their apparent impedance behavior, and especially the passivity of the apparent impedance (i.e., their interaction stability) as well as their sensitivity to sensor noise. We compare classical SEA control approaches such as cascaded PID controllers and full state feedback controllers with advanced controllers using disturbance observers, acceleration feedback and adaptation rules. Simulations and experiments demonstrate the trade-off between stable interactions, high bandwidths and low noise levels. Based on these trade-offs, an application-specific controller can be designed and tuned, based on desired interaction with the respective environment.

Background: Balance control is important for mobility, yet exoskeleton research has mainly focused on improving metabolic energy efficiency. Here we present a biomimetic exoskeleton controller that supports walking balance and reduces muscle activity. Methods: Humans restore balance after a perturbation by adjusting activity of the muscles actuating the ankle in proportion to deviations from steady-state center of mass kinematics. We designed a controller that mimics the neural control of steady-state walking and the balance recovery responses to perturbations. This controller uses both feedback from ankle kinematics in accordance with an existing model and feedback from the center of mass velocity. Control parameters were estimated by fitting the experimental relation between kinematics and ankle moments observed in humans that were walking while being perturbed by push and pull perturbations. This identified model was implemented on a bilateral ankle exoskeleton. Results: Across twelve subjects, exoskeleton support reduced calf muscle activity in steady-state walking by 19% with respect to a minimal impedance controller (p < 0.001). Proportional feedback of the center of mass velocity improved balance support after perturbation. Muscle activity is reduced in response to push and pull perturbations by 10% (p = 0.006) and 16% (p < 0.001) and center of mass deviations by 9% (p = 0.026) and 18% (p = 0.002) with respect to the same controller without center of mass feedback. Conclusion: Our control approach implemented on bilateral ankle exoskeletons can thus effectively support steady-state walking and balance control and therefore has the potential to improve mobility in balance-impaired individuals.

Humans prioritize regulation of the whole-body angular momentum (WBAM) during walking. When perturbed, modulations of the moment arm of the ground reaction force (GRF) with respect to the centre of mass (CoM) assist in recovering WBAM. For sagittal-plane perturbations of the WBAM given at toe off right (TOR), horizontal GRF modulations and not centre of pressure (COP) modulations were mainly responsible for these moment arm modulations. In this study, we aimed to find whether the instant of perturbations affects the contributions of the GRF and/or CoP modulations to the moment arm changes, in balance recovery during very slow walking. Perturbations of the WBAM were applied at three different instants of the gait cycle, namely at TOR, mid-swing (MS), and heel strike right (HSR). Forces equal to 16% of the participant's body weight were applied simultaneously to the pelvis and upper body in opposite directions for a duration of 150 ms. The results showed that the perturbation onset did not significantly affect the GRF moment arm modulation. However, the contribution of both the CoP and GRF modulation to the moment arm changes did change depending on the perturbation instant. After perturbations resulting in a forward pitch of the trunk a larger contribution was present from the CoP modulation when perturbations were given at MS or HSR, compared to perturbations at TOR. After backward pitch perturbations given at MS and HSR the CoP modulation counteracted the moment arm required for WBAM recovery. Therefore a larger contribution from the horizontal GRF was needed to direct the GRF posterior to the CoM and recover WBAM. In conclusion, the onset of WBAM perturbations does not affect the moment arm modulation needed for WBAM recovery, while it does affect the way CoP and GRF modulation contribute to that recovery.

Knowledge on joint impedance during walking in various conditions is relevant for clinical decision-making and the development of robotic gait trainers, leg prostheses, leg orthotics and wearable exoskeletons. Whereas ankle impedance during walking has been experimentally assessed, knee and hip joint impedance during walking have not been identified yet. Here we developed and evaluated a lower limb perturbator to identify hip, knee and ankle joint impedance during treadmill walking. The lower limb perturbator (LOPER) consists of an actuator connected to the thigh via rods. The LOPER allows to apply force perturbations to a free-hanging leg, while standing on the contralateral leg, with a bandwidth of up to 39 Hz. While walking in minimal impedance mode, the interaction forces between LOPER and the thigh were low (<5N) and the effect on the walking pattern was smaller than the within-subject variability during normal walking. Using a non-linear multibody dynamical model of swing leg dynamics, the hip, knee and ankle joint impedance were estimated at three time points during the swing phase for nine subjects walking at a speed of 0.5 m/s. The identified model was well able to predict the experimental responses for the hip and knee, since the mean variance accounted (VAF) for was 99% and 96%, respectively. The ankle lacked a consistent response and the mean VAF of the model fit was only 77%, and therefore the estimated ankle impedance was not reliable. The averaged across-subjects stiffness varied between the three time points within 34-66 and 0-3.5 Nm/rad Nm/rad for the hip and knee joint respectively. The damping varied between 1.9-4.6 and 0.02-0.14 Nms/rad Nms/rad for hip and knee respectively. The developed LOPER has a negligible effect on the unperturbed walking pattern and allows to identify hip and knee impedance during the swing phase.

Healthy individuals highly regulate their whole body angular momentum (WBAM) during walking. Since WBAM regulation is essential in maintaining balance, a better understanding is required on how healthy individuals recover from WBAM perturbations. We therefore studied how healthy individuals recover WBAM in the sagittal plane. WBAM can be regulated by adjusting the moment arm of the ground reaction force (GRF) vector with respect to the whole-body centre of mass (CoM). In principle this can be done by centre of pressure (CoP) modulation and/or adjustments of the GRF direction. Two simultaneous perturbations of the same magnitude were applied in opposite direction to the pelvis and upper body (0.34m apart) to perturb WBAM but not the whole body linear momentum (WBLM), while participants walked on a treadmill. The perturbations were given at toe off right, had a magnitude of 4, 8, 12 and 16% of the participant's body weight, and lasted for 150ms. A recovery of the WBAM was seen directly after the perturbations, induced by adapting the moment arm of the GRF with respect to the CoM. The hip joint of the stance leg played an important role in achieving the WBAM recovery. A change in the direction of the GRF vector and not a contributing CoP modulation, caused the change in moment arm. However, the change in GRF direction came from a change in the horizontal GRF, which also affects the WBLM. This suggest that regulating WBAM may take precedence over the WBLM in early recovery.

Joint impedance plays an important role in postural control and movement. However, current experimental knowledge on lower limb impedance during gait is limited to the ankle joint. We designed the LOwer limb PERturbator (LOPER) aimed to assess knee and hip joint impedance during gait. The LOPER applies force perturbations with a 39 Hz bandwidth, tested on a free-hanging leg. In minimal impedance mode, peak interaction forces during walking are low (&lt; 5 N). Also, this mode has a negligible effect on the gait pattern, as it is smaller than the within-subject variability during normal walking. In short, the LOPER is a transparent device able to elicit a clear response at both hip and knee joints to investigate lower limb dynamics. A second motor added to the LOPER could improve isolation of the perturbation contribution to knee and hip dynamics. People with neurological disorders can benefit from knowledge of joint impedance during gait through improved biomimetic devices and clinical decision making.

BACKGROUND: In the last two decades, lower-limb exoskeletons have been developed to assist human standing and locomotion. One of the ongoing challenges is the cooperation between the exoskeleton balance support and the wearer control. Here we present a cooperative ankle-exoskeleton control strategy to assist in balance recovery after unexpected disturbances during walking, which is inspired on human balance responses. METHODS: We evaluated the novel controller in ten able-bodied participants wearing the ankle modules of the Symbitron exoskeleton. During walking, participants received unexpected forward pushes with different timing and magnitude at the pelvis level, while being supported (Exo-Assistance) or not (Exo-NoAssistance) by the robotic assistance provided by the controller. The effectiveness of the assistive strategy was assessed in terms of (1) controller performance (Detection Delay, Joint Angles, and Exerted Ankle Torques), (2) analysis of effort (integral of normalized Muscle Activity after perturbation onset); and (3) Analysis of center of mass COM kinematics (relative maximum COM Motion, Recovery Time and Margin of Stability) and spatio-temporal parameters (Step Length and Swing Time). RESULTS: In general, the results show that when the controller was active, it was able to reduce participants' effort while keeping similar ability to counteract and withstand the balance disturbances. Significant reductions were found for soleus and gastrocnemius medialis activity of the stance leg when comparing Exo-Assistance and Exo-NoAssistance walking conditions. CONCLUSIONS: The proposed controller was able to cooperate with the able-bodied participants in counteracting perturbations, contributing to the state-of-the-art of bio-inspired cooperative ankle exoskeleton controllers for supporting dynamic balance. In the future, this control strategy may be used in exoskeletons to support and improve balance control in users with motor disabilities.

In this paper, we present the design, control, and preliminary evaluation of the Symbitron exoskeleton, a lower limb modular exoskeleton developed for people with a spinal cord injury. The mechanical and electrical configuration and the controller can be personalized to accommodate differences in impairments among individuals with spinal cord injuries (SCI). In hardware, this personalization is accomplished by a modular approach that allows the reconfiguration of a lower-limb exoskeleton with ultimately eight powered series actuated (SEA) joints and high fidelity torque control. For SCI individuals with an incomplete lesion and sufficient hip control, we applied a trajectory-free neuromuscular control (NMC) strategy and used the exoskeleton in the ankle-knee configuration. For complete SCI individuals, we used a combination of a NMC and an impedance based trajectory tracking strategy with the exoskeleton in the ankle-knee-hip configuration. Results of a preliminary evaluation of the developed hardware and software showed that SCI individuals with an incomplete lesion could naturally vary their walking speed and step length and walked faster compared to walking without the device. SCI individuals with a complete lesion, who could not walk without support, were able to walk with the device and with the support of crutches that included a push-button for step initiationOur results demonstrate that an exoskeleton with modular hardware and control allows SCI individuals with limited or no lower limb function to receive tailored support and regain mobility.

Background: People with brain or neural injuries, such as cerebral palsy or spinal cord injury, commonly have joint hyper-resistance. Diagnosis and treatment of joint hyper-resistance is challenging due to a mix of tonic and phasic contributions. The parallel-cascade (PC) system identification technique offers a potential solution to disentangle the intrinsic (tonic) and reflexive (phasic) contributions to joint impedance, i.e. resistance. However, a simultaneous neurophysiological validation of both intrinsic and reflexive joint impedances is lacking. This simultaneous validation is important given the mix of tonic and phasic contributions to joint hyper-resistance. Therefore, the main goal of this paper is to perform a group-level neurophysiological validation of the PC system identification technique using electromyography (EMG) measurements. Methods: Ten healthy people participated in the study. Perturbations were applied to the ankle joint to elicit reflexes and allow for system identification. Participants completed 20 hold periods of 60 seconds, assumed to have constant joint impedance, with varying magnitudes of intrinsic and reflexive joint impedances across periods. Each hold period provided a paired data point between the PC-based estimates and neurophysiological measures, i.e. between intrinsic stiffness and background EMG, and between reflexive gain and reflex EMG. Results: The intrinsic paired data points, with all subjects combined, were strongly correlated, with a range of r=[0.870.91] in both ankle plantarflexors and dorsiflexors. The reflexive paired data points were moderately correlated, with r=[0.640.69] in the ankle plantarflexors only. Conclusion: An agreement with the neurophysiological basis on which PC algorithms are built is necessary to support its clinical application in people with joint hyper-resistance. Our results show this agreement for the PC system identification technique on group-level. Consequently, these results show the validity of the use of the technique for the integrated assessment and training of people with joint hyper-resistance in clinical practice.

Motorized assessment of the stretch reflex is instrumental to gain understanding of the stretch reflex, its physiological origin and to differentiate effects of neurological disorders, like spasticity. Both short-latency (M1) and medium-latency (M2) stretch reflexes have been reported to depend on the velocity and acceleration of an applied ramp-and-hold perturbation. In the upper limb, M2 has also been reported to depend on stretch duration. However, wrong conclusions might have been drawn in previous studies as the interdependence of perturbation parameters (amplitude, duration, velocity, and acceleration) possibly created uncontrolled, confounding effects. We disentangled the duration-, velocity-, and acceleration-dependence and their interactions of the M1 and M2 stretch reflex in the ankle plantarflexors. To disentangle the parameter interdependence, 49 unique ramp-and-hold joint perturbations elicited reflexes in 10 healthy volunteers during a torque control task. Linear mixed model analysis showed that M1 depended on acceleration, not velocity or duration, whereas M2 depended on acceleration, velocity, and duration. Simulations of the muscle spindle Ia afferents coupled to a motoneuron pool corroborated these experimental findings. In addition, this simulation model did show a nonlinear M1 velocity- and duration-dependence for perturbation parameters outside the experimental scope. In conclusion, motorized assessment of the stretch reflex or spasticity using ramp-and-hold perturbations should be systematically executed and reported. Our systematic motorized and simulation assessments showed that M1 and M2 depend on acceleration, velocity, and duration of the applied perturbation. The simulation model suggested that these dependencies emerge from: muscle-tendon unit and muscle cross-bridge dynamics, Ia sensitivity to force and yank, and motoneuron synchronization.

Centre of mass (CoM) motion during human balance recovery is largely influenced by the ground reaction force (GRF) and the centre of pressure (CoP). During gait, foot placement creates a region of possible CoP locations in the following double support (DS). This study aims to increase insight into how humans modulate the CoP during DS, and which CoP modulations are theoretically possible to maintain balance in the sagittal plane. Three variables sufficient to describe the shape, length and duration of the DS CoP trajectory of the total GRF, were assessed in perturbed human walking. To counteract the forward perturbations, braking was required and all CoP variables showed modulations correlated to the observed change in CoM velocity over the DS phase. These correlations were absent after backward perturbations, when only little propulsion was needed to counteract the perturbation. Using a linearized inverted pendulum model we could explore how the observed parameter modulations are effective in controlling the CoM. The distance the CoP travels forward and the instant the leading leg was loaded largely affected the CoM velocity, while the duration mainly affected the CoM position. The simulations also showed that various combinations of CoP parameters can reach a desired CoM position and velocity at the end of DS, and that even a full recovery in the sagittal plane within DS would theoretically have been possible. However, the human subjects did not exploit the therefore required CoP modulations. Overall, modulating the CoP trajectory in DS does effectively contributes to balance recovery.

Background: Recently developed controllers for robot-assisted gait training allow for the adjustment of assistance for specific subtasks (i.e. specific joints and intervals of the gait cycle that are related to common impairments after stroke). However, not much is known about possible interactions between subtasks and a better understanding of this can help to optimize (manual or automatic) assistance tuning in the future. In this study, we assessed the effect of separately assisting three commonly impaired subtasks after stroke: foot clearance (FC, knee flexion/extension during swing), stability during stance (SS, knee flexion/extension during stance) and weight shift (WS, lateral pelvis movement). For each of the assisted subtasks, we determined the influence on the performance of the respective subtask, and possible effects on other subtasks of walking and spatiotemporal gait parameters. Methods: The robotic assistance for the FC, SS and WS subtasks was assessed in nine mildly impaired chronic stroke survivors while walking in the LOPES II gait trainer. Seven trials were performed for each participant in a randomized order: six trials in which either 20% or 80% of assistance was provided for each of the selected subtasks, and one baseline trial where the participant did not receive subtask-specific assistance. The influence of the assistance on performances (errors compared to reference trajectories) for the assisted subtasks and other subtasks of walking as well as spatiotemporal parameters (step length, width and height, swing and stance time) was analyzed. Results: Performances for the impaired subtasks (FC, SS and WS) improved significantly when assistance was applied for the respective subtask. Although WS performance improved when assisting this subtask, participants were not shifting their weight well towards the paretic leg. On a group level, not many effects on other subtasks and spatiotemporal parameters were found. Still, performance for the leading limb angle subtask improved significantly resulting in a larger step length when applying FC assistance. Conclusion: FC and SS assistance leads to clear improvements in performance for the respective subtask, while our WS assistance needs further improvement. As effects of the assistance were mainly confined to the assisted subtasks, tuning of FC, SS and WS can be done simultaneously. Our findings suggest that there may be no need for specific, time-intensive tuning protocols (e.g. tuning subtasks after each other) in mildly impaired stroke survivors.

Haptic interaction between two humans, for example, a physiotherapist assisting a patient regaining the ability to grasp a cup, likely facilitates motor skill acquisition. Haptic human–human interaction has been shown to enhance individual performance improvement in a tracking task with a visuomotor rotation perturbation. These results are remarkable given that haptically assisting or guiding an individual rarely benefits their individual improvement when the assistance is removed. We, therefore, replicated a study that reported that haptic interaction between humans was beneficial for individual improvement for tracking a target in a visuomotor rotation perturbation. In addition, we tested the effect of more interaction time and a stronger haptic coupling between the partners on individual improvement in the same task. We found no benefits of haptic interaction on individual improvement compared to individuals who practised the task alone, independent of interaction time or interaction strength.

Human walking speeds can be influenced by multiple factors, from energetic considerations to the time to reach a destination. Neurological deficits or lower-limb injuries can lead to slower walking speeds, and the recovery of able-bodied gait speed and behavior from impaired gait is considered an important rehabilitation goal. Because gait studies are typically performed at faster speeds, little normative data exists for very slow speeds (less than 0.6 ms^{− 1}). The purpose of our study was to investigate healthy gait mechanics at extremely slow walking speeds. We recorded kinematic and kinetic data from eight adult subjects walking at four slow speeds from 0.1 ms^{− 1} to 0.6 ms^{− 1} and at their self-selected speed. We found that known relations for spatiotemporal and work measures are still valid at very slow speeds. Trends derived from slow speeds largely provided reasonable estimates of gait measures at self-selected speeds. Our study helps enable valuable comparisons between able-bodied and impaired gait, including which pathological behaviors can be attributed to slow speeds and which to gait deficits. We also provide a slow walking dataset, which may serve as normative data for clinical evaluations and gait rehabilitative devices.