Offline and Real-Time Implementation of a Terrain Classification Pipeline for Pushrim-Activated Power-Assisted Wheelchairs.

Pushrim-activated power-assisted wheelchairs (PAPAWs) are assistive technologies that provide propulsion assist to wheelchair users and enable access to various indoor and outdoor terrains. Therefore, it is beneficial to use PAPAW controllers that adapt to different terrain conditions. To achieve this objective, terrain classification techniques can be used as an integral part of the control architecture. Previously, the feasibility of using learning-based terrain classification models was investigated for offline applications. In this paper, we examine the effects of three model parameters (i.e., feature characteristics, terrain types, and the length of data segments) on offline and real-time classification accuracy. Our findings revealed that Random Forest classifiers are computationally efficient and can be used effectively for real-time terrain classification. These classifiers have the highest performance accuracy when used with a combination of time- and frequency-domain features. Additionally, we found that increasing the number of data points used for terrain estimation improves the prediction accuracy. Finally, our results revealed that classification accuracy can be improved by considering terrains with similar characteristics under one umbrella group. These findings can contribute to the development of real-time adaptive controllers that enhance PAPAW usability on different terrains.

A Comparison Between Conventional and Terrain-Specific Adaptive Pushrim-Activated Power-Assisted Wheelchairs.

Pushrim-activated power-assisted wheels (PAPAWs) are assistive technologies that provide on-demand propulsion assistance to wheelchair users. In this study, we aimed to develop an adaptive PAPAW controller that responds effectively to changes in environmental conditions (e.g., type of surface or terrain). Experiments were conducted to collect kinematics of wheelchair motion using a frame-mounted inertial measurement unit (IMU) while performing a variety of wheelchair activities on different indoor/outdoor terrains. Statistical characteristics of velocity and acceleration measurements were extracted and used to develop a terrain classification framework to identify certain indoor and outdoor terrains. The terrain classification framework, based on random forest classification algorithms and kinematic features, was implemented and tested in our laboratory-developed PAPAW. This computationally efficient terrain classification framework was successfully implemented and tested in real-time. The power-assist ratio of each wheel was adjusted based on the type of terrain (e.g., more assistance was provided on outdoor terrains). Our findings revealed that propulsion effort (e.g., peak input torque) on asphalt was significantly reduced when using adaptive controllers compared to conventional PAPAW controllers. In addition, subjective views of participants regarding the workload of wheelchair propulsion (e.g., physical/cognitive effort) supported the positive effects of adaptive PAPAW controllers. We believe that the adoption of terrain-specific adaptive controllers has the potential to improve the accessibility of outdoor terrains and to prevent or delay upper extremity joint degeneration or pain.

A Comparison Between Conventional and User-Intention-Based Adaptive Pushrim-Activated Power-Assisted Wheelchairs.

Pushrim-activated power-assisted wheel (PAPAW) users ideally require different levels of assistance depending on activity and preference. Therefore, it is important to design and develop adaptive PAPAW controllers to account for these differences. The main objective of this work was to integrate a user intention estimation framework into a PAPAW and develop personalized adaptive controllers. We performed experiments to gather kinetics of wheelchair propulsion for a variety of daily life wheelchair activities. The propulsion characteristics (i.e., pushrim forces) were used to train intention estimation models and characterize implicit user intentions when performing daily life wheelchair maneuvers. These intentions included moving straight forward, performing a right/left turn, and braking. The intention estimation framework, based on random forest classification algorithms and kinetic features, was implemented and tested in our laboratory-developed PAPAW. This computationally efficient framework was successfully implemented and tested for each participant in real-time. Our results revealed that the real-time user intention predictions were similar to the offline models. The power-assist ratio of each wheel was adjusted based on which user intention was identified. Data collected from four participants provided evidence regarding the effectiveness of using adaptive intention-based controllers. For instance, the propulsion effort was significantly reduced when using an adaptive PAPAW controller. Subjective views of participants regarding the workload of wheelchair propulsion (e.g., physical/cognitive effort) were also gathered. Our findings suggest that rankings of different controllers varied among different participants and across different wheelchair maneuvers, indicating the need for customized adaptive controllers to fit different users' activities and preferences.

Development of A Learning-Based Terrain Classification Framework for Pushrim-Activated Power-Assisted Wheelchairs.

Pushrim-activated power-assisted wheels (PAPAWs) are assistive technologies that provide on-demand torque assistance to wheelchair users. Although the available power can reduce the physical load of wheelchair propulsion, it may also cause maneuverability and controllability issues. Commercially-available PAPAW controllers are insensitive to environmental changes, leading to inefficient and/or unsafe wheelchair movements. In this regard, adaptive velocity/torque control strategies could be employed to improve safety and stability. To investigate this objective, we propose a context-aware sensory framework to recognize terrain conditions. In this paper, we present a learning-based terrain classification framework for PAPAWs. Study participants performed various maneuvers consisting of common daily-life wheelchair propulsion routines on different indoor and outdoor terrains. Relevant features from wheelchair frame-mounted gyroscope and accelerometer measurements were extracted and used to train and test the proposed classifiers. Our findings revealed that a one-stage multi-label classification framework has a higher accuracy performance compared to a two-stage classification pipeline with an indoor-outdoor classification in the first stage. We also found that, on average, outdoor terrains can be classified with higher accuracy (90%) compared to indoor terrains (65%). This framework can be used for real-time terrain classification applications and provide the required information for an adaptive velocity/torque controller design.

Error Augmentation in Immersive Virtual Reality for Bimanual Upper-Limb Rehabilitation in Individuals With and Without Hemiplegic Cerebral Palsy.

With more readily available commercial immersive virtual reality (VR) technologies, the potential of new feedback strategies as tools to facilitate motor rehabilitation should be investigated. Augmented feedback or error augmentation (EA) can easily be shown in a virtual environment. Here, visual EA provided via immersive VR was tested for its effectiveness to improve bimanual symmetry in a reaching task. A single-session crossover design was used to test two training cases, with or without EA. With EA, the distance between hands in the forward direction was augmented. Participants were recruited from typically developing (TD) populations (n = 12, ages 13-21) and performed in an adapted environment with an initial asymmetry between limbs. Also, five participants with hemiplegic cerebral palsy (CP) (ages 14-21, MACS I-III) completed the study. Among TD participants, a significantly larger change in symmetry in the adapted environment was shown after EA than training without EA (F (1, 10) = 9.64, p = 0.01). Each participant in the CP group also improved more after EA training (8.8-103.7)%, such that they achieved lower symmetry error after training with EA. As participants in both groups adapted more symmetrically with EA, beneficial changes from this training method could be evaluated in future studies for longer-term functional changes.

Towards the Development of a Learning-Based Intention Classification Framework for Pushrim-Activated Power-Assisted Wheelchairs.

There has been a growth in the design and use of power assist devices for manual wheelchairs (MWCs) to alleviate the physical load of MWC use. A pushrim-activated power-assisted wheel (PAPAW) is an example of a power assist device that replaces the conventional wheel of a MWC. Although the use of PAPAWs provides some benefits to MWC users, it can also cause difficulties in maneuvering the wheelchair. In this research, we examined the characteristics of wheelchair propulsion when using manual and powered wheels. We used the left and right wheels' angular velocity to calculate the linear and angular velocity of the wheelchair. Results of this analysis revealed that the powered wheel's controller is not optimally designed to reflect the intentions of a wheelchair user. To address some of the challenges with coordinating the pushes on PAPAWs, we proposed the design of a user-intention detection framework. We used the kinematic data of MWC experiments and tested six supervised learning algorithms to classify one of four movements: "not moving", "moving straight forward", "turning left", and "turning right". We found that all the classification algorithms determined the type of movement with high accuracy and low computation time. The proposed intention detection framework can be used in the design of learning-based controllers for PAPAWs that take into account the individualized characteristics of wheelchair users. Such a system may improve the experience of PAPAW users.

Studies on Practical Applications of Safe-Fall Control Strategies for Lower Limb Exoskeletons.

Lower limb exoskeletons (LLEs) are susceptible to falls, and users are at risk of head and/or hip injuries. To address concerns regarding the safety of LLE users, optimization techniques were used to study safe-fall control strategies. Simulation results of these studies showed promising performance that leads to head impact avoidance and mitigation of hip impact velocity. The motivation for the current research was to extend the application of previously developed optimization techniques to study more realistic human-LLE fall conditions. We examined a range of feasible fall durations for the human-LLE model and found the optimal fall duration for which the user's safety is maximized. Next, we used a range of coefficients of friction to examine fall strategies on different ground surface conditions. We found that the effectiveness of a safe-fall strategy is higher when falling on less slippery surfaces compared to more slippery ones. The simulation results were implemented in a half-scale physical model of a three-link inverted pendulum, which represented a human-LLE model. Results of our experiments verified that the optimal safe-fall strategy could be implemented in a mechanical test setup. The hip linear velocity at impact was found to have similar values in both the experimental (2.04 m/s) and simulation results (2.09 m/s). Further studies should be conducted with appropriate software and hardware platforms to successfully implement safe-fall strategies in an actual LLE.

Application of Commercial Games for Home-Based Rehabilitation for People with Hemiparesis: Challenges and Lessons Learned.

OBJECTIVE: To identify the factors that influence the use of an at-home virtual rehabilitation gaming system from the perspective of therapists, engineers, and adults and adolescents with hemiparesis secondary to stroke, brain injury, and cerebral palsy. MATERIALS AND METHODS: This study reports on qualitative findings from a study, involving seven adults (two female; mean age: 65 ± 8 years) and three adolescents (one female; mean age: 15 ± 2 years) with hemiparesis, evaluating the feasibility and clinical effectiveness of a home-based custom-designed virtual rehabilitation system over 2 months. Thematic analysis was used to analyze qualitative data from therapists' weekly telephone interview notes, research team documentation regarding issues raised during technical support interactions, and the transcript of a poststudy debriefing session involving research team members and collaborators. RESULTS: Qualitative themes that emerged suggested that system use was associated with three key factors as follows: (1) the technology itself (e.g., characteristics of the games and their clinical implications, system accessibility, and hardware and software design); (2) communication processes (e.g., preferences and effectiveness of methods used during the study); and (3) knowledge and training of participants and therapists on the technology's use (e.g., familiarity with Facebook, time required to gain competence with the system, and need for clinical observations during remote therapy). Strategies to address these factors are proposed. CONCLUSION: Lessons learned from this study can inform future clinical and implementation research using commercial videogames and social media platforms. The capacity to track compensatory movements, clinical considerations in game selection, the provision of kinematic and treatment progress reports to participants, and effective communication and training for therapists and participants may enhance research success, system usability, and adoption.

Biofeedback vs. game scores for reducing trunk compensation after stroke: a randomized crossover trial.

Background Compensatory movements are commonly employed by stroke survivors, and their use can have negative effects on motor recovery. Current practices to reduce them rely on strapping a person to a chair. The use of technology to substitute or supplement this methodology has not being thoroughly investigated. Objective To compare the use of Scores + Visual + Force and Visual + Force feedback for reducing trunk compensation. Methods Fourteen hemiparetic stroke survivors performed bimanual reaching movements while receiving feedback on trunk compensation. Participants held onto two robotic arms and performed movements in the anterior/posterior direction toward a target displayed on a monitor. A motion-tracking camera tracked trunk compensation; the robots provided force feedback; the monitor displayed the visual feedback and scores. Kinematic variables, a post-test questionnaire, and system usability were analyzed. Results Both conditions reduced trunk compensation from baseline: Scores + Visual + Force: 51.7% (40.8), p = 0.000; Visual + Force: 55.2% (40.9), p = 0.000. No statistically significant difference was found between modalities. Secondary outcome measures were not improved. Most participants would like to receive game scores to reduce trunk compensation, and the usability of the system was rated "Good." Conclusions Multimodal feedback about stroke survivors' trunk compensation levels resulted in reduced trunk displacement. No difference between feedback modalities was obtained. The positive effects of including game scores might not have been observed in a short-term intervention. Longer studies should investigate if the use of game scores could result in trunk compensation improvements when compared to trunk restraint strategies. Clinical Trial Registration Clinicaltrials.gov, NCT02912923, https://clinicaltrials.gov/ct2/show/NCT02912923?term=reaching+in+stroke&rank=2 .

Developing safe fall strategies for lower limb exoskeletons.

One of the main challenges in the use of a powered lower limb exoskeleton (LLE) is to ensure that balance is maintained throughout the operation of the device. Since no control strategy has yet been implemented that prevents falls in the case of a loss of balance, head or other serious injuries may occur during independent use of LLEs in the event of a fall. These safety concerns limit LLEs in the community to supervised use only. Using the backward fall as a model, we used optimization techniques to develop safe fall control strategies in order to avoid head impact and mitigate the impact velocity of the hips. From available human biomechanics data, we first developed an optimization methodology to study falls of healthy people. The results showed similar kinematic and dynamic characteristics to findings of previous studies on real-life human falls. Second, we extended the optimization methodology to include characteristics of a hypothetical LLE and to generate optimal joint trajectories and optimal torque profiles for the fall duration. The results revealed that by applying the optimal fall strategy, the severity of a simulated fall was minimized compared to when the device fell with locked joints (i.e., how currently used exoskeletons fall): head impact was avoided and hip impact velocity was reduced by more than 50%.

Data sample size needed for analysis of kinematic and muscle synergies in healthy and stroke populations.

Multiple studies have suggested the central nervous system (CNS) generates motions by using modular control of muscles and joints (synergies). However, the synergies reported by these studies are task dependent and might not reflect the true control strategies adopted by the CNS. Studying exploratory motions (EMs) can reveal biomechanical constraints and motor control strategies in healthy and clinical populations. The first logical step to consider EMs in study of motor synergies is to determine how much data is required to reliably and fully profile the motion patterns of an individual. Here we present how the quality of motor synergies analysis depends on the amount of EM data included in the analysis. We recruited 10 healthy and 10 post-stroke participants and collected electromyography (EMG) and joint motion data of their arms as they completed a motor exploration task. We compared the effects of clinical status and limb strength/dominance on the amount of data required to identify synergies. Clinical status had a significant elïect on the required amount of data for both datasets. Limb strength had a significant effect only for kinematic data. We determined the upper bound 95% confidence interval to set the amount of data required for synergy analysis in both populations: 235 sec for EMG data and 265 sec for kinematic data. Our results provide an important step toward using motor exploration in the study of healthy motor synergies and how stroke alters them.

Reducing Trunk Compensation in Stroke Survivors: A Randomized Crossover Trial Comparing Visual and Force Feedback Modalities.

OBJECTIVE: To investigate whether the compensatory trunk movements of stroke survivors observed during reaching tasks can be decreased by force and visual feedback, and to examine whether one of these feedback modalities is more efficacious than the other in reducing this compensatory tendency. DESIGN: Randomized crossover trial. SETTING: University research laboratory. PARTICIPANTS: Community-dwelling older adults (N=15; 5 women; mean age, 64±11y) with hemiplegia from nontraumatic hemorrhagic or ischemic stroke (>3mo poststroke), recruited from stroke recovery groups, the research group's website, and the community. INTERVENTIONS: In a single session, participants received augmented feedback about their trunk compensation during a bimanual reaching task. Visual feedback (60 trials) was delivered through a computer monitor, and force feedback (60 trials) was delivered through 2 robotic devices. MAIN OUTCOME MEASURES: Primary outcome measure included change in anterior trunk displacement measured by motion tracking camera. Secondary outcomes included trunk rotation, index of curvature (measure of straightness of hands' path toward target), root mean square error of hands' movement (differences between hand position on every iteration of the program), completion time for each trial, and posttest questionnaire to evaluate users' experience and system's usability. RESULTS: Both visual (-45.6% [45.8 SD] change from baseline, P=.004) and force (-41.1% [46.1 SD], P=.004) feedback were effective in reducing trunk compensation. Scores on secondary outcome measures did not improve with either feedback modality. Neither feedback condition was superior. CONCLUSIONS: Visual and force feedback show promise as 2 modalities that could be used to decrease trunk compensation in stroke survivors during reaching tasks. It remains to be established which one of these 2 feedback modalities is more efficacious than the other as a cue to reduce compensatory trunk movement.

On identifying kinematic and muscle synergies: a comparison of matrix factorization methods using experimental data from the healthy population.

Human motor behavior is highly goal directed, requiring the central nervous system to coordinate different aspects of motion generation to achieve the motion goals. The concept of motor synergies provides an approach to quantify the covariation of joint motions and of muscle activations, i.e., elemental variables, during a task. To analyze goal-directed movements, factorization methods can be used to reduce the high dimensionality of these variables while accounting for much of the variance in large data sets. Three factorization methods considered in this paper are principal component analysis (PCA), nonnegative matrix factorization (NNMF), and independent component analysis (ICA). Bilateral human reaching data sets are used to compare the methods, and advantages of each are presented and discussed. PCA and NNMF had a comparable performance on both EMG and joint motion data and both outperformed ICA. However, NNMF's nonnegativity condition for activation of basis vectors is a useful attribute in identifying physiologically meaningful synergies, making it a more appealing method for future studies. A simulated data set is introduced to clarify the approaches and interpretation of the synergy structures returned by the three factorization methods. NEW & NOTEWORTHY: Literature on comparing factorization methods in identifying motor synergies using numerically generated, simulation, and muscle activation data from animal studies already exists. We present an empirical evaluation of the performance of three of these methods on muscle activation and joint angles data from human reaching motion: principal component analysis, nonnegative matrix factorization, and independent component analysis. Using numerical simulation, we also studied the meaning and differences in the synergy structures returned by each method. The results can be used to unify approaches in identifying and interpreting motor synergies.

Trunk Compensation During Bimanual Reaching at Different Heights by Healthy and Hemiparetic Adults.

The authors explored how trunk compensation and hand symmetry in stroke survivors and healthy controls were affected by the distance and height of virtual targets during a bimanual reaching task. Participants were asked to reach to 4 different virtual targets set at: 90% of their arm length at shoulder, xiphoid process, and knee height, and 50% of their arm length at xiphoid process height. For the stroke group, for all targets, the hands' movements were more asymmetrical than those of the healthy group, with more asymmetry observed in the direction of gravity, and trunk forward displacement values were larger and more variable. The knee targets had the largest trunk displacement values; index of curvature and trunk displacement were strongly correlated with participants' impairment scores. A strong correlation was found between the hands' asymmetry in the anterior or posterior direction for the shoulder targets, and the impairment scores. The results suggest that target height influences the degree of trunk compensation and hand symmetry during bimanual reaching by hemiparetic participants.

Transformation of Vestibular Signals for the Control of Standing in Humans.

During standing balance, vestibular signals encode head movement and are transformed into coordinates that are relevant to maintaining upright posture of the whole body. This transformation must account for head-on-body orientation as well as the muscle actions generating the postural response. Here, we investigate whether this transformation is dependent upon a muscle's ability to stabilize the body along the direction of a vestibular disturbance. Subjects were braced on top of a robotic balance system that simulated the mechanics of standing while being exposed to an electrical vestibular stimulus that evoked a craniocentric vestibular error of head roll. The balance system was limited to move in a single plane while the vestibular error direction was manipulated by having subjects rotate their head in yaw. Vestibular-evoked muscle responses were greatest when the vestibular error was aligned with the balance direction and decreased to zero as the two directions became orthogonal. This demonstrates that muscles respond only to the component of the error that is aligned with the balance direction and thus relevant to the balance task, not to the cumulative afferent activity, as expected for vestibulospinal reflex loops. When we reversed the relationship between balancing motor commands and associated vestibular sensory feedback, the direction of vestibular-evoked ankle compensatory responses was also reversed. This implies that the nervous system quickly reassociates new relationships between vestibular sensory signals and motor commands related to maintaining balance. These results indicate that vestibular-evoked muscle activity is a highly flexible balance response organized to compensate for vestibular disturbances. SIGNIFICANCE STATEMENT: The postural corrections critical to standing balance and navigation rely on transformation of sensory information into reference frames that are relevant for the required motor actions. Here, we demonstrate that the nervous system transforms vestibular sensory signals of head motion according to a muscle's ability to stabilize the body along the direction of a vestibular-evoked disturbance. By manipulating the direction of the imposed vestibular signal relative to a muscle's action, we show that the vestibular contribution to muscle activity is a highly flexible and organized balance response. This study provides insight into the neural integration and central processing associated with transformed vestibulomotor relationships that are essential to standing upright.

Evaluating the User Experience of Exercising Reaching Motions With a Robot That Predicts Desired Movement Difficulty.

The notion of an optimal difficulty during practice has been articulated in many areas of cognitive psychology: flow theory, the challenge point framework, and desirable difficulties. Delivering exercises at a participant's desired difficulty has the potential to improve both motor learning and users' engagement in therapy. Motivation and engagement are among the contributing factors to the success of exercise programs. The authors previously demonstrated that error amplification can be used to introduce levels of challenge into a robotic reaching task, and that machine-learning algorithms can dynamically adjust difficulty to the desired level with 85% accuracy. Building on these findings, we present the results of a proof-of-concept study investigating the impacts of practicing under desirable difficulty conditions. A control condition with a predefined random order for difficulty levels was deemed more suitable for this study (compared to constant or continuously increasing difficulty). By practicing the task at their desirable difficulties, participants in the experimental group perceived their performance at a significantly higher level and reported lower required effort to complete the task, in comparison to a control group. Moreover, based on self-reports, participants in the experimental group were willing, on average, to continue the training session for 4.6 more training blocks (∼45 min) compared to the control group's average. This study demonstrates the efficiency of delivering the exercises at the user's desired difficulty level to improve the user's engagement in exercise tasks. Future work will focus on clinical feasibility of this approach in increasing stroke survivors' engagement in their therapy programs.

Use of activity theory-based need finding for biomedical device development.

Identifying the appropriate needs for biomedical device design is challenging, especially for less structured environments. The paper proposes an alternate need-finding method based on Cultural Historical Activity Theory and expanded to explicitly examine the role of devices within a socioeconomic system. This is compared to a conventional need-finding technique in a preliminary study with engineering student teams. The initial results show that the Activity Theory-based technique allows teams to gain deeper insights into their needs space.

Therapists' perceptions of social media and video game technologies in upper limb rehabilitation.

BACKGROUND: The application of technologies, such as video gaming and social media for rehabilitation, is garnering interest in the medical field. However, little research has examined clinicians' perspectives regarding technology adoption by their clients. OBJECTIVE: The objective of our study was to explore therapists' perceptions of how young people and adults with hemiplegia use gaming and social media technologies in daily life and in rehabilitation, and to identify barriers to using these technologies in rehabilitation. METHODS: We conducted two focus groups comprised of ten occupational therapists/physiotherapists who provide neurorehabilitation to individuals with hemiplegia secondary to stroke or cerebral palsy. Data was analyzed using inductive thematic analysis. The diffusion of innovations theory provided a framework to interpret emerging themes. RESULTS: Therapists were using technology in a limited capacity. They identified barriers to using social media and gaming technology with their clients, including a lack of age appropriateness, privacy issues with social media, limited transfer of training, and a lack of accessibility of current systems. Therapists also questioned their role in the context of technology-based interventions. The opportunity for social interaction was perceived as a major benefit of integrated gaming and social media. CONCLUSIONS: This study reveals the complexities associated with adopting new technologies in clinical practice, including the need to consider both client and clinician factors. Despite reporting several challenges with applying gaming and social media technology with clinical populations, therapists identified opportunities for increased social interactions and were willing to help shape the development of an upper limb training system that could more readily meet the needs of clients with hemiplegia. By considering the needs of both therapists and clients, technology developers may increase the likelihood that clinicians will adopt innovative technologies.

A wearable vibrotactile device for upper-limb bilateral motion training in stroke rehabilitation: A case study.

Real-time feedback is essential for motor learning. Automated feedback is especially valuable for at-home stroke rehabilitation in the absence of therapist supervision. This study examined the effect of real-time corrective vibrotactile feedback for training bilateral reaching motions. A bilateral upper-limb motor learning system, comprising a wireless wearable sleeve-armband device for providing vibrotactile feedback, a computer target game, and a customized motion tracking technology, was developed and evaluated on both hemiparetic stroke survivors and able-bodied people. This paper introduces the system and presents preliminary data for one hemiparetic stroke subject and one healthy subject performing bimanual reaching motions in the transverse plane. Vibrotactile training was found to successfully alter both subjects' original trajectories and to improve the motion symmetry of the stroke subject. These preliminary findings indicated the potential efficacy of vibrotactile cues for unsupervised motor learning in both the healthy and the stroke populations.

Shifting the balance of human standing: Inter-limb coordination for the control of a robotic balance simulation.

Learning to maintain standing balance in the presence of a paretic limb is an important recovery process for many stroke survivors. In this study, we used a robotic balance simulator to investigate whether manipulating medial-lateral or anterior-posterior torque contributions (i.e. input gains) could shift the control of balance toward a targeted lower limb in healthy controls. Manipulation of medial-lateral (ML) torque gains shifted the vertical load distribution toward the virtually weakened limb, but did not result in a significant shift in anterior-posterior (AP) torque control. Instead individual participants were observed to shift AP torque control in either direction, although participants more often shifted control toward the virtually weakened limb at larger ML asymmetries. In contrast, manipulation of AP torque gains did not produce any observable changes in measured torque signals. The shift in torque contributions during ML manipulations shows promise as an implicit training method for reducing weight-bearing asymmetry. However, further work is required to ensure both vertical load and AP torque control shift in the desired direction as well as to determine the applicability of the protocol in a patient population.