Stroke-induced abnormal upper extremity (UE) joint coupling limits independent joint control and impairs functional arm use. While previous studies have predominantly evaluated UE joint coupling within the plane of motion (in-plane) and its effect on functional task performance, minimizing unnecessary joint movements outside the plane of motion (out-of-plane) across all UE degrees of freedom (DOFs) is essential for coordinated movements. To address this limitation, we developed an experimental procedure leveraging 3D motion capture to evaluate in- and out-of-plane joint coupling for seven UE DOFs during isolated shoulder, elbow, and wrist joint movements, and functional task performance during an object transfer task. We introduce a method to calculate the in- and out-of-plane joint coupling ratio (JCR) for seven UE DOFs. We investigated the contribution of stroke-induced abnormal in- and out-of-plane joint coupling to explain deficits in functional task performance using hierarchical regression analysis. In 18 individuals post-stroke, joint coupling was abnormal in- and out-of-plane for seven UE DOFs, evident by significantly higher JCR values for the paretic arm compared to the non-paretic arm. The regression model using both in- and out-of-plane joint coupling explained significantly higher variance, up to 33.8%, in stroke-induced deficits in movement duration, hand trajectory smoothness, trunk displacement, hand movement extent, and peak velocity time compared to models only using either in-plane or out-of-plane joint coupling. Our work advances post-stroke abnormal joint coupling evaluation methods across all UE DOFs, required to more comprehensively understand stroke-induced impairments in independent UE joint control and their effect on functional task performance.

Shoulder motion is essential for daily activities, yet accurate quantification outside laboratory settings remains challenging. Existing wearable systems either accumulate drift over extended recordings or rely on elaborate calibration protocols that limit real-world deployment. Here we report a wearable platform for long-duration tracking of three-dimensional shoulder kinematics with minimal calibration. The system combines a sensing shirt instrumented with inertial measurement units and soft strain sensors with a lightweight learning-based fusion framework to estimate joint orientation. Using only a few minutes of unconstrained arm movements and no laboratory equipment, the approach achieves tracking errors below 5° across all degrees of freedom during more than 1 h of continuous functional activity. Sensitivity analyses indicate robust performance with shortened calibration and reduced sensor configurations. This strategy establishes a practical route towards reliable, long-duration shoulder kinematics in unconstrained environments, supporting applications in movement science, clinical monitoring and assistive technologies. Wearable inertial and soft sensors combined with a lightweight learning-based fusion framework achieve long-duration, accurate three-dimensional tracking of shoulder motion with minimal calibration.

Introduction: Walking impairments are a leading cause of long-term disability after stroke. Early, targeted interventions are critical for neuromotor recovery, yet precise treatment targets remain unclear. Deficits in central drive to the paretic plantarflexors (i.e., the percentage of plantarflexor strength capacity accessed during maximum voluntary contraction) are linked to impaired gait propulsion and slow walking speeds in chronic stroke, but are poorly understood in subacute stroke, largely due to lack of clinically accessible assessment tools. To address this, our team developed a mobile system that validly measures plantarflexor central drive in clinical settings. This study evaluates associations between central drive deficits, gait quality, and real-world physical activity in individuals with sub-acute stroke. Hypothesis: We hypothesized that plantarflexor central drive deficits would be associated with impaired gait quality and reduced real-world physical activity. Methods: Six participants with sub-acute stroke (age: 63 ± 12.25 years; stroke onset: 2.41 ± 0.93 months) were enrolled at discharge from inpatient rehabilitation. Physical therapists conducted all assessments. Gait was measured using wearable sensors during the 6-Minute Walk Test to capture stride length, cadence, speed. Central drive was evaluated using a custom-built, portable, neurostimulation-integrated dynamometer. Physical activity was tracked for 7 days using Actigraph monitors to quantify time spent in light (≤3.0 METs) and moderate-to-vigorous (>3.0 METs) activity. Spearman correlations assessed relationships between central drive, gait metrics, and activity levels. Results: Participants had an average central drive of 87.65 ± 14.60% (100%= no impairment). Greater central drive deficits were significantly associated with shorter stride length ( r s =0.943 ; p = 0.005 ) and more time in light activity ( r s = -0.886 ; p = 0.019), but not with moderate-to-vigorous daily physical activity (p > 0.05). No significant associations were found with gait speed or cadence. Conclusions: Preliminary findings support the clinical value of central drive in subacute stroke, revealing strong associations with impaired gait, particularly shorter stride length, and limited engagement in health-promoting moderate-to-vigorous physical activity. Early interventions targeting plantarflexor central drive deficits may improve gait, daily physical activity, and ultimately long-term health after stroke.

Actuators for wearable robots are difficult to design due to the competing requirements of minimizing weight and form factor while achieving desired performance requirements. Current solutions often rely on a "one-size-fits-all" approach utilizing classical motors and gears. Here, we present a thin, high torque density wearable actuator consisting of a cycloid gear and PCB axial flux motor, along with a modeling and design optimization framework to minimize actuator and battery mass while maintaining target performance. The model predicts motor mass with 95.9% accuracy and battery power with 75.9-87.9% accuracy. We then apply this design and optimization framework to quantify the benefits of creating optimized sizes of wearable actuators for gait assistance for stroke survivors, which was found to reduce average actuator and battery mass by 384 g with just 5 sizes. Furthermore, sizing the actuator and battery together resulted in an average of 28.9% to 48.3% more weight savings than sizing only the motor or only the battery, respectively, for individually optimized actuators. This paradigm of sizing actuators like clothing, along with new actuator architectures, may yield weight savings that could improve adoption of daily-wear assistive devices.

Back exoskeletons are gaining attention for preventing occupational back injuries, but they can disrupt movement, a burden that risks abandonment. Enhanced adaptability is proposed to mitigate burdens, but perceptual benefits are less known. This study investigates the perceptual and biomechanical impacts of a SLACK suit (non-assistive) controller versus three controllers with varying adaptability: a Weight-Direction-Angle adaptive (WDA-ADPT) that scales assistance based on the weight of the boxes using a chest-mounted camera and machine learning algorithm, movement direction, and trunk flexion angle, and standard Direction-Angle adaptive (DA-ADPT) and Angle adaptive (A-ADPT) controllers. Fifteen participants performed a variable weight (2, 8, 14 kg) box-transfer task. WDA-ADPT achieved the highest perceptual score (88%) across survey categories and reduced peak back extensor (BE) muscle amplitudes by 10.1%. DA-ADPT had slightly lower perceptual (76%) and peak BE reduction (8.5%). A-ADPT induced hip restriction, which could explain the lowest perceptual score (55%) despite providing the largest reductions in peak BE muscle activity (17.3%). Reduced perceptual scores achieved by DA and A-ADPT were explained by controllers providing too much or little assistance versus actual task demands. These findings underscore that scaling assistance to task demands improves biomechanical benefits and the perception of the device’s suitability.

Portable wearable robots offer promise for assisting people with upper limb disabilities. However, movement variability between individuals and trade-offs between supportiveness and transparency complicate robot control during real-world tasks. We address these challenges by first developing a personalized ML intention detection model to decode user's motion intention from IMU and compression sensors. Second, we leverage a physics-based hysteresis model to enhance control transparency and adapt it for practical use in real-world tasks. Third, we combine and integrate these two models into a real-time controller to modulate the assistance level based on the user's intention and kinematic state. Fourth, we evaluate the effectiveness of our control strategy in improving arm function in a multi-day evaluation. For 5 individuals post-stroke and 4 living with ALS wearing a soft shoulder robot, we demonstrate that the controller identifies shoulder movement with 94.2% accuracy from minimal change in the shoulder angles (elevation: 3.4°, depression: 1.7°) and reduces arm-lowering force by 31.9% compared to a baseline controller. Furthermore, the robot improves movement quality by increasing their shoulder elevation/depression (17.5°), elbow (10.6°) and wrist flexion/extension (7.6°) ROMs; reducing trunk compensation (up to 25.4%); and improving hand-path efficiency (up to 53.8%).

The Fugl Meyer Assessment (FMA) is a widely-used assessment for tracking motor function recovery post-stroke. Due to the limited access to rehabilitation, there exists a need for remote and automated assessment solutions. Wearable sensors and data-driven methods have shown promise for enabling automatic upper extremity FMA (FMA-UE) estimation, but minimizing user input motion and aligning with current clinical activities will aid the adoption of sensor-based assessments. In this work, we present an FMA-UE estimator which can make score predictions for a key subset of the assessment (70% of all items) using data from inertial measurement units (IMUs) placed on the arms and the trunk from three volitional reaching motions representative of functional daily activities. We collected a dataset of eleven stroke participants performing a subset of FMA-UE, and three reaching motions. The FMA-UE of each participant was assessed by an occupational therapist providing the labeled score for the training data. The estimator was trained on windowed data during FMA-UE motions and was able to make score estimates from reaching motions. Through leave-one-subjectout cross validation, the estimator achieved a normalized RMSE of 7%, which is comparable to or below the established minimal clinically important difference and minimal detectable change of FMA-UE of post-stroke individuals. Comparison experiments of various model designs also revealed the importance of trunk-based features inspired by compensation strategies common post stroke and features extracted from the hand sensor. The proposed estimator has the potential to broaden the possibility of automatic assessment via wearable sensors.

Agricultural systems are both emitters of greenhouse gases and have the potential to sequester carbon, especially agroforestry systems. Coffee agroforestry systems offer a wide range of intensities of use of agricultural inputs and densities and management of shade trees. We assessed the agronomic carbon footprint (up to farm gate) and modelled the carbon sequestration of a range of coffee agroforestry systems across 180 farms in Costa Rica and Guatemala. The agronomic carbon footprint included upstream, direct and indirect processes associated with chemical and organic fertiliser use and energy consumption (excluding processing of cherries). Carbon sequestration was modelled using the CAF2021 model a processed based model of the C, N and water dynamics specifically designed for coffee agroforestry systems. The carbon footprint per kg of coffee cherries was significantly and positively related to the level of nitrogen inputs. Modelled changes in C stocks i.e. carbon sequestration was significantly and positively related to the Leaf Area Index (LAI) of the trees, and the levels of nitrogen inputs. Increasing nitrogen inputs per hectare was positively associated with emission per kg as nitrogen efficiency varied significantly across the sample. The net carbon balance, defined as sequestration minus CO2e emissions was also positively related to shade tree LAI but negatively with yield and N application. Carbon positive farms were characterized by shade cover over 60 %, but low yields and low net income. However, farms that were close to carbon neutral had higher yields and higher net income, with shade levels of about 50 % cover, while carbon negative farms which had shade cover averaging 40 %. Nevertheless, farms showed a large variation in performance with all combinations of positive and negative for carbon balance and net income. However, among the farms with a positive net income, those with a positive carbon balance had a significantly lower net income than those that were carbon negative (i.e. net emitters). This confirms the economic trade-off for farmers seeking to maximise these two goals. If farmers are expected to generate positive carbon balances and potentially to offset emissions higher in the supply chain, then they should receive economic support to compensate continued on-farm carbon accumulation.

Accurately estimating kinetic metrics, such as braking and propulsion forces, in real-world running environments enhances our understanding of performance, fatigue, and injury. Wearable inertial measurement units (IMUs) offer a potential solution to estimate kinetic metrics outside the lab when combined with machine learning. However, current IMU-based kinetic estimation models are trained and evaluated within a single environment, often on lab treadmills. The transferability of these treadmill-trained models during overground running in and out of the lab is underexplored, and the individualization and validation of such models remain a challenge. Toward bridging this gap, we trained a generalized model on treadmill data of 15 recreational runners and evaluated braking and propulsion force estimates during overground running in and out of the lab. We explored fine-tuning with individual data from lab-based overground running to quantify model performance improvements with individualization. The generalized and fine-tuned models were extrapolated to outdoor running for a subset of five participants, and estimates were compared to lab-based overground measurements. Evaluating the generalized model with a leave-one-out cross validation yielded overground braking and propulsion force root mean squared error of 4.3 ± 1.1 % bodyweight (%BW). Fine-tuning this model with eight strides reduced error to 2.6 ± 0.5 %BW. Outdoor force predictions from the fine-tuned model better aligned with expected linear trends between braking/propulsion impulses and speed than the generalized model. These results provide insights into the accuracy and applicability of IMU data-driven models for braking and propulsion estimation during overground running, facilitating the development of practical, individualized biomechanical analysis tools for real-world use.

Bayesian optimization (BO) has become indispensable for black box optimization. However, BO is often considered a black box itself, lacking transparency in the rationale behind proposed parameters. This is particularly relevant in human-in-the-loop applications like personalization of wearable robotic devices. We address BO's opacity by proposing ShapleyBO, a framework for interpreting BO proposals by game-theoretic Shapley values. Our approach quantifies the contribution of each parameter to BO's acquisition function (AF). By leveraging the linearity of Shapley values, ShapleyBO can identify the influence of each parameter on BO's exploration and exploitation behaviors. Our method gives rise to a ShapleyBO-assisted human-machine interface (HMI), allowing users to interfere with BO in case proposals do not align with human reasoning. We demonstrate these HMI's benefits for the use case of personalizing wearable robotic devices (assistive back exosuits) by human-in-the-loop BO. Results suggest that human-BO teams with access to ShapleyBO outperform teams without access to ShapleyBO. (Open Science: ShapleyBO as well as code and data to reproduce findings available at https://github.com/rodemann/ShapleyBO. This work builds upon the master's thesis of the second author supervised by the last author [18], and substantially extends and formalizes the results presented therein.).

ABSTRACT:Amyotrophic lateral sclerosis is a progressive neurodegenerative disease affecting upper and lower motor neurons that control voluntary muscles. With no known cure, clinical care is focused on symptom management to maximize function and quality of life. Assistive technology plays a crucial role and enables some restoration of movement and function despite disease progression. This scoping review assesses the effectiveness of assistive technologies tested in people living with amyotrophic lateral sclerosis, specifically those designed to compensate for upper and lower extremity, trunk, and cervical muscle weakness. A comprehensive search was conducted across PubMed, CINAHL, ERIC, and Google Scholar and through citation chasing. We included 26 articles that tested an assistive device on at least one person living with amyotrophic lateral sclerosis and evaluated the device's effectiveness in restoring movement or providing stabilization to support functional mobility or activities of daily living. Most studies were pilot feasibility or usability trials, with small numbers of amyotrophic lateral sclerosis participants. The devices showed various benefits, including improved range of motion, function, and participation in daily activities. This review highlights the potential for assistive devices to enhance function in people living with amyotrophic lateral sclerosis and underscores the need for comprehensive studies involving larger cohorts of individuals at different stages of amyotrophic lateral sclerosis.

Post-stroke motor impairments limit the ability to functionally move the arm against gravity which leads to a reduction in upper limb movement quality. Hence, providing anti-gravity arm support is widely adopted as an effective strategy in motor rehabilitation to enable functional upper limb movement. Recently, we developed a soft wearable shoulder robot that lifts the arm against gravity using a pneumatic actuator attached underneath the arm on a custom-made shirt. This study aims to evaluate the effectiveness of the soft wearable robot in improving upper limb movement quality for individuals post-stroke. We recruited four individuals post-stroke (age = 60.5±18.3 years, UE-FMA scores = 45.5±11) and asked them to lift their arm against gravity to 90° flexion and abduction, three times each with the robot turned on and off. We used an optical motion capture system to record and evaluate upper limb movement by calculating 1) shoulder elevation/depression range of motion (ROM), 2) end-effector movement quality (hand-path-ratio of the three-dimensional (3D) hand center of mass (COM) trajectory), and 3) trunk compensation (displacement of 3D trunk COM). We found that the soft wearable robot increased shoulder elevation/depression ROM, on average, by 6.1±1.7°, for the four participants. Additionally, the robot improved end-effector movement quality and reduced trunk compensation by decreasing hand-path-ratio and trunk displacement, on average, by 50.9±7.7% and 15.3±6.7%, respectively. The demonstrated improvement in movement quality highlights the potential utility of our device for upper limb motor rehabilitation.

Nearly all soft wearable robots rely on textiles to distribute actuation forces to the human body; however, the mechanical hysteresis of these materials significantly complicates device control. If not properly accounted for, this history-dependent behavior can result in substantial over-/under-support for which the human user must actively compensate. While a number of hysteresis modeling approaches have been proposed, these techniques are either (a) heuristic-driven and do not accurately reflect the observed physical behavior or (b) rely on complex benchtop calibration procedures that are not amenable to wearable applications where the complete human-robot system must be holistically considered. In this work, we present a new strategy to predict the complex hysteretic response of the combined human-robot system given its full state history using a mathematical technique known as a Preisach model. Our approach is directly personalized to each individual with data collected on the body in ∼ 90 seconds. We demonstrate the technique with a previously proposed soft wearable robot for shoulder assistance, though the concept is applicable to any joint. To benchmark the efficacy of our approach against previously proposed strategies, we performed an open-loop trajectory tracking procedure with 12 human participants and an articulated mannequin. Our strategy achieved an average shoulder elevation angle tracking accuracy of 5.3° across human participants, representing a significant improvement compared to prior techniques. We anticipate that this new approach will facilitate significantly improved soft wearable robot control by providing reliable estimates of the full hysteretic system response, enabling more robust physical human-robot interaction and coordination.

Gait rehabilitation is critical for regaining locomotor independence after neuromotor injuries like stroke. Rehabilitation literature indicates the need for such therapy to continue beyond the clinic in order to maintain motor function and support recovery. However, implementing community-based rehabilitation requires the ability to monitor gait in the real-world with clinically relevant accuracies. Despite advances in machine learning, achieving this performance with single sensing modalities has been challenging using wearable sensors like inertial measurement units (IMUs) and pressure insoles. Here, we investigate the benefits of multi-modal sensing by integrating IMU and insole data to develop individualized machine learning models in people post-stroke that estimate propulsion, a key biomechanical variable. We show that in the lab, IMU + Insole models improve performance relative to IMU only and Insole only models, with an average root-mean-squared-error (RMSE) of 0.80 %bodyweight (%BW) across the stance phase. We obtain RMSEs of 0.71%BW for peak paretic propulsion and 0.19%BW s for paretic propulsion impulse, which are within corresponding clinical thresholds. We then explore the application of this algorithm to track propulsion changes in the real-world for two participants during variable-speed walking and two participants during active gait interventions, either functional electrical stimulation or exosuit-applied resistance. For these participants, we observe similar changes in measured propulsion in the lab and estimated propulsion out of the lab across speeds and interventions. Overall, this work aims to address the challenges in applying machine learning methods for individuals post-stroke and presents an investigation into the feasibility of developing estimation methods for real-world propulsion estimation during gait rehabilitation.

Wearable robotic devices have become increasingly prevalent in both occupational and rehabilitative settings, yet their widespread adoption remains inhibited by usability barriers related to comfort, restriction, and noticeable functional benefits. Acknowledging the importance of user perception in this context, this study explores preference-based controller optimization for a back exosuit that assists lifting. Considering the high mental and metabolic effort discrete motor tasks impose, we used a forced-choice Bayesian Optimization approach that promotes sampling efficiency by leveraging domain knowledge about just noticeable differences between assistance settings. Optimizing over two control parameters, preferred settings were consistent within and uniquely different between participants. We discovered that overall, participants preferred asymmetric parameter configurations with more lifting than lowering assistance, and that preferences were sensitive to user anthropometrics. These findings highlight the potential of perceptually guided assistance optimization for wearable robotic devices, marking a step toward more pervasive adoption of these systems in the real world.

Recent work has shown that exoskeletons controlled through data-driven methods can dynamically adapt assistance to various tasks for healthy young adults. However, applying these methods to populations with neuromotor gait deficits, such as post-stroke hemiparesis, is challenging. This is due not only to high population heterogeneity and gait variability but also to a lack of post-stroke gait datasets to train accurate models. Despite these challenges, data-driven methods offer a promising avenue for control, potentially allowing exoskeletons to function safely and effectively in unstructured community settings. This work presents a first step towards enabling adaptive plantarflexion and dorsiflexion assistance from data-driven torque estimation during post-stroke walking. We trained a multi-task Temporal Convolutional Network (TCN) using collected data from four post-stroke participants walking on a treadmill (R^2 of 0.74 ± 0.13). The model uses data from three inertial measurement units (IMU) and was pretrained on healthy walking data from 6 participants. We implemented a wearable prototype for our ankle torque estimation approach for exoskeleton control and demonstrated the viability of real-time sensing, estimation, and actuation with one post-stroke participant.

Static aerodynamic surfaces are inherently limited in their ability to adapt to dynamic velocity profiles or environmental changes, restricting their performance under variable operating conditions. This challenge is particularly pronounced in high-speed competitive sports, such as cycling and downhill skiing, where the properties of a static textile surface are mismatched with highly dynamic wind-speed profiles. Here, an textile metamaterial is introduced that is capable of variable aerodynamic profiles through a stretch-induced dimpling mechanism, even when tightly conformed to a body or object. Wind-tunnel experiments are used to characterize the variable aerodynamic performance of the dimpling mechanism, while Finite Element (FE) simulations efficiently characterize the design space to identify optimal textile metamaterial architectures. By controlling dimple size, the aerodynamic performance of the textile can be tailored for specific wind-speed ranges, resulting in an ability to modulate drag force at target wind-speeds by up to 20%. Furthermore, the potential for active control of a textiles' aerodynamic properties is demonstrated, in which controlled stretching allows the textile to sustain optimal performance across a dynamic wind-speed profile. These findings establish a new approach to aerodynamic metamaterials, with surface dimpling and thus variable fluid-dynamic properties offering transformative applications for wearables, as well as broader opportunities for aerospace, maritime, and civil engineering systems.

Spinal cord injuries (SCI) often lead to upper limb impairment, necessitating innovative solutions for daily assistance beyond traditional rigid robotics due to their impractical weight and size. Despite still preliminary, soft wearables are arising as a possible solution to fill this gap. Here, we demonstrated an enhanced version of a soft inflatable robot that assists the shoulder against gravity, previously tested with different neurological conditions. Noteworthy improvements include a single-layer actuator, simplifying manufacturing, a built-in bending angle and a nylon hammock, for better armpit conformity. We characterized the actuator (approximately 8Nm at 90∘ at 70kPa) and demonstrated its good transparency, both from a kinematic and a muscular standpoint. Then, on 11 healthy individuals, we showed reductions in shoulder muscle activity (both at the anterior and middle deltoid) while performing a lift and hold task, ranging from 16% to almost 60% of the maximum voluntary contraction. More importantly, we confirmed these effects on two SCI individuals SCI, at two different stages of recovery. While preliminary, considering the limited exploration of soft wearable robots for the shoulder in SCI cases, this is a significant advancement playing an important role in the development of future soft technology for SCI assistance.

In one person with Parkinson's disease, freezing of gait was averted through the use of a soft robotic apparel that provided a moderate level of hip-flexion assistance during the swing phase of walking. This approach delivered instantaneous effects and consistently improved walking quality and function across a range of conditions.