Inseung Kang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Research Interests
Inseung Kang has a total of 38 co-authors. Here are the top co-authors by number of shared publications:
- Aaron J. Young
Georgia Institute of Technology
10 shared publications - Dawit Lee
Department of Bioengineering, Stanford University, Stanford, CA, USA
5 shared publications Aaron J. Young
George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
3 shared publicationsDean D. Molinaro
Robotics and AI (RAI) Institute, Cambridge, MA, USA
3 shared publications- Dean D. Molinaro
Institute for Robotics and Intelligent Machines (IRIM), Georgia Institute of Technology
3 shared publications - Pratik Kunapuli
Department of Computer and Information Science, University of Pennsylvania, Philadelphia, PA, USA
3 shared publications
Latest Publications
- Open Access
Appellian Neural Networks
The movement of articulated bodies, whether biological or robotic, involves two interrelated computational challenges. Inverse dynamics determines the forces or torques required to achieve a specified movement, while forward dynamics computes the resulting accelerations from given forces or torques. Although both computations are governed by the same equations of motion, they are typically implemented as separate inverse and forward mappings in both model-based and data-driven approaches. This division can obscure the underlying structure common to both problems. Here, we introduce the Appellian Neural Network (ANNet), a physics-informed neural network that unifies both computations through a shared learned representation by modeling a single scalar quantity from classical mechanics, specifically Appell’s acceleration energy. Inverse dynamics is obtained by differentiating the learned energy function with respect to acceleration to recover joint torques. Forward dynamics is computed without retraining by embedding the same learned energy landscape into an optimization objective, where the unregularized minimum satisfies the Gibbs–Appell equation. The resulting accelerations are then integrated forward in time. ANNet is evaluated using a double pendulum paradigm. For trajectories not encountered during training, both inverse and forward simulations demonstrate accuracy and real-time capability. Both computations propagate input noise similarly to their analytical counterparts, and adaptation of the inverse dynamics to a modified pendulum generalizes to forward simulation without additional forward-specific training. These results establish a first-principles approach for employing a single learned representation to support both inverse and forward dynamic computations.
Open AccessAppellian Neural Networks
The movement of articulated bodies, whether biological or robotic, involves two interrelated computational challenges. Inverse dynamics determines the forces or torques required to achieve a specified movement, while forward dynamics computes the resulting accelerations from given forces or torques. Although both computations are governed by the same equations of motion, they are typically implemented as separate inverse and forward mappings in both model-based and data-driven approaches. This division can obscure the underlying structure common to both problems. Here, we introduce the Appellian Neural Network (ANNet), a physics-informed neural network that unifies both computations through a shared learned representation by modeling a single scalar quantity from classical mechanics, specifically Appell’s acceleration energy. Inverse dynamics is obtained by differentiating the learned energy function with respect to acceleration to recover joint torques. Forward dynamics is computed without retraining by embedding the same learned energy landscape into an optimization objective, where the unregularized minimum satisfies the Gibbs–Appell equation. The resulting accelerations are then integrated forward in time. ANNet is evaluated using a double pendulum paradigm. For trajectories not encountered during training, both inverse and forward simulations demonstrate accuracy and real-time capability. Both computations propagate input noise similarly to their analytical counterparts, and adaptation of the inverse dynamics to a modified pendulum generalizes to forward simulation without additional forward-specific training. These results establish a first-principles approach for employing a single learned representation to support both inverse and forward dynamic computations.
- 4 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Robotics
- Vol. 41 (15523098) , 4941-4959
Online Adaptation Framework Enables Personalization of Exoskeleton Assistance During Locomotion in Patients Affected by Stroke
Robotic exoskeletons can transform mobility for individuals with lower limb disabilities. However, their widespread adoption is limited by controller degradation caused by varying gait dynamics across different users and environments. Here, we propose an online adaptation framework that leverages real-time data streams to continuously update the user state estimator model. This approach allows the exoskeleton to learn the user-specific gait patterns, effectively customizing the model for each new user. In addition, we demonstrate a sensor signal transformation technique that enables model transfer across different exoskeleton hardware (from a research-grade exoskeleton to a commercial device). With less than one minute of adaptation, our framework improved gait phase estimation, which directly affects assistance timing, by 40.9% for able-bodied subjects and 65.9% for stroke survivors ($p$ $< $ 0.05), and reduced torque profile error by 32.7% compared to the baseline model ($p$ $< $ 0.05). Furthermore, in a pilot test, we applied our adaptation framework with human-in-the-loop optimization for control tuning. In a single stroke survivor, this approach led to a 21.8% increase in walking speed and a 6.5% reduction in metabolic cost compared to walking without exoskeleton. While preliminary, these results suggest the potential for personalized exoskeleton assistance in clinical populations.
- 4 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Robotics
- Vol. 41 (15523098) , 4941-4959
Online Adaptation Framework Enables Personalization of Exoskeleton Assistance During Locomotion in Patients Affected by Stroke
Robotic exoskeletons can transform mobility for individuals with lower limb disabilities. However, their widespread adoption is limited by controller degradation caused by varying gait dynamics across different users and environments. Here, we propose an online adaptation framework that leverages real-time data streams to continuously update the user state estimator model. This approach allows the exoskeleton to learn the user-specific gait patterns, effectively customizing the model for each new user. In addition, we demonstrate a sensor signal transformation technique that enables model transfer across different exoskeleton hardware (from a research-grade exoskeleton to a commercial device). With less than one minute of adaptation, our framework improved gait phase estimation, which directly affects assistance timing, by 40.9% for able-bodied subjects and 65.9% for stroke survivors ($p$ $< $ 0.05), and reduced torque profile error by 32.7% compared to the baseline model ($p$ $< $ 0.05). Furthermore, in a pilot test, we applied our adaptation framework with human-in-the-loop optimization for control tuning. In a single stroke survivor, this approach led to a 21.8% increase in walking speed and a 6.5% reduction in metabolic cost compared to walking without exoskeleton. While preliminary, these results suggest the potential for personalized exoskeleton assistance in clinical populations.
- 14 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Robotics and Bionics
- Vol. 7 (2) , 666-677
Machine Learning Enables Rapid Detection of Slips Using a Robotic Hip Exoskeleton
Fall incidents due to slips are some of the most common causes of injuries for industry workers and older adults, motivating research to assist balance recovery following slips. To assist balance recovery during a slip, a detection algorithm that can work with an assistive device, such as an exoskeleton, needs to be able to detect slips rapidly after onset, which remains a critical gap in the field. Here, we compared the ability of linear discriminant analysis (LDA), extreme gradient boosting (XGBoost), and convolutional neural networks (CNN) to detect slip using only native sensors on a hip exoskeleton. We trained and evaluated user-independent models on early-stance (ES) and late-stance (LS) slips of various magnitudes collected through treadmill-based slips. All models, except LDA with LS slips, detected slips with ¿90% accuracy. Overall, the best model was XGBoost, with its fastest results achieving average detection times and median accuracies of 155.06 ms at 96.25% for ES slips and 228.88 ms at 93.75% for LS slips, while also achieving 100% sensitivity at 195.64 ms (ES) and 266.24 ms (LS). Our results indicate a promising direction for further research into designing a generalizable model for balance recovery during slip perturbations using robotic hip exoskeletons.
- 14 April 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Medical Robotics and Bionics
- Vol. 7 (2) , 666-677
Machine Learning Enables Rapid Detection of Slips Using a Robotic Hip Exoskeleton
Fall incidents due to slips are some of the most common causes of injuries for industry workers and older adults, motivating research to assist balance recovery following slips. To assist balance recovery during a slip, a detection algorithm that can work with an assistive device, such as an exoskeleton, needs to be able to detect slips rapidly after onset, which remains a critical gap in the field. Here, we compared the ability of linear discriminant analysis (LDA), extreme gradient boosting (XGBoost), and convolutional neural networks (CNN) to detect slip using only native sensors on a hip exoskeleton. We trained and evaluated user-independent models on early-stance (ES) and late-stance (LS) slips of various magnitudes collected through treadmill-based slips. All models, except LDA with LS slips, detected slips with ¿90% accuracy. Overall, the best model was XGBoost, with its fastest results achieving average detection times and median accuracies of 155.06 ms at 96.25% for ES slips and 228.88 ms at 93.75% for LS slips, while also achieving 100% sensitivity at 195.64 ms (ES) and 266.24 ms (LS). Our results indicate a promising direction for further research into designing a generalizable model for balance recovery during slip perturbations using robotic hip exoskeletons.
- 30 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 72 (7) , 2147-2156
Human-in-the-Loop Optimization of Hip Exoskeleton Assistance During Stair Climbing
Objective: This study applies human-in-the-loop optimization to identify optimal hip exoskeleton assistance patterns for stair climbing. Methods: Ten participants underwent optimization to individualize hip flexion and extension assistance, followed by a validation comparing optimized assistance (OPT) to biological hip moment-based assistance (BIO), no assistance (No-Assist), and no exoskeleton (No-Exo) conditions. Results: OPT reduced metabolic cost by 4.5% compared to No-Exo, 11.44% compared to No-Assist, and 5.02% compared to BIO, demonstrating the effectiveness of the optimization approach. Statistical analysis revealed distinct characteristics in optimal assistance timing and magnitude that deviated systematically from biological hip moment patterns. Compared to BIO, OPT exhibited later peak flexion timing (76.4 $\pm$ 3.7% vs 65.0%), shorter flexion duration (29.2 $\pm$ 3.6% vs 40.0%), later peak extension timing (26.7 $\pm$ 3.8% vs 20.0% of gait cycle), and higher peak flexion magnitude (11.1 $\pm$ 1.5 Nm vs 10.0 Nm). While individual optimal assistance profiles varied across participants, comparison between individually optimized parameters and the best subject-independent parameters identified through post-hoc analysis showed consistency. On average, metabolic rate convergence was achieved after 18 iterations, while most exoskeleton control parameters did not reach our convergence criteria within 20 iterations. Conclusion: These findings demonstrate that human-in-the-loop optimization can effectively identify task-specific assistance patterns for stair climbing, while the consistency between individual and subject-independent parameters suggests the potential for developing generalized assistance strategies. The systematic differences between optimized and biological moment-based assistance underscore the fundamental distinctions between biological torque-based control and optimal control for exoskeleton assistance.
- 30 January 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 72 (7) , 2147-2156
Human-in-the-Loop Optimization of Hip Exoskeleton Assistance During Stair Climbing
Objective: This study applies human-in-the-loop optimization to identify optimal hip exoskeleton assistance patterns for stair climbing. Methods: Ten participants underwent optimization to individualize hip flexion and extension assistance, followed by a validation comparing optimized assistance (OPT) to biological hip moment-based assistance (BIO), no assistance (No-Assist), and no exoskeleton (No-Exo) conditions. Results: OPT reduced metabolic cost by 4.5% compared to No-Exo, 11.44% compared to No-Assist, and 5.02% compared to BIO, demonstrating the effectiveness of the optimization approach. Statistical analysis revealed distinct characteristics in optimal assistance timing and magnitude that deviated systematically from biological hip moment patterns. Compared to BIO, OPT exhibited later peak flexion timing (76.4 $\pm$ 3.7% vs 65.0%), shorter flexion duration (29.2 $\pm$ 3.6% vs 40.0%), later peak extension timing (26.7 $\pm$ 3.8% vs 20.0% of gait cycle), and higher peak flexion magnitude (11.1 $\pm$ 1.5 Nm vs 10.0 Nm). While individual optimal assistance profiles varied across participants, comparison between individually optimized parameters and the best subject-independent parameters identified through post-hoc analysis showed consistency. On average, metabolic rate convergence was achieved after 18 iterations, while most exoskeleton control parameters did not reach our convergence criteria within 20 iterations. Conclusion: These findings demonstrate that human-in-the-loop optimization can effectively identify task-specific assistance patterns for stair climbing, while the consistency between individual and subject-independent parameters suggests the potential for developing generalized assistance strategies. The systematic differences between optimized and biological moment-based assistance underscore the fundamental distinctions between biological torque-based control and optimal control for exoskeleton assistance.
- Top Cited Papers
- 20 March 2024
- Published by American Association for the Advancement of Science (AAAS) in Science Robotics
- Vol. 9 (88) , eadi8852
Estimating human joint moments unifies exoskeleton control, reducing user effort
Robotic lower-limb exoskeletons can augment human mobility, but current systems require extensive, context-specific considerations, limiting their real-world viability. Here, we present a unified exoskeleton control framework that autonomously adapts assistance on the basis of instantaneous user joint moment estimates from a temporal convolutional network (TCN). When deployed on our hip exoskeleton, the TCN achieved an average root mean square error of 0.142 newton-meters per kilogram across 35 ambulatory conditions without any user-specific calibration. Further, the unified controller significantly reduced user metabolic cost and lower-limb positive work during level-ground and incline walking compared with walking without wearing the exoskeleton. This advancement bridges the gap between in-lab exoskeleton technology and real-world human ambulation, making exoskeleton control technology viable for a broad community.
- 20 March 2024
- Published by American Association for the Advancement of Science (AAAS) in Science Robotics
- Vol. 9 (88) , eadi8852
Top Cited PapersEstimating human joint moments unifies exoskeleton control, reducing user effort
Robotic lower-limb exoskeletons can augment human mobility, but current systems require extensive, context-specific considerations, limiting their real-world viability. Here, we present a unified exoskeleton control framework that autonomously adapts assistance on the basis of instantaneous user joint moment estimates from a temporal convolutional network (TCN). When deployed on our hip exoskeleton, the TCN achieved an average root mean square error of 0.142 newton-meters per kilogram across 35 ambulatory conditions without any user-specific calibration. Further, the unified controller significantly reduced user metabolic cost and lower-limb positive work during level-ground and incline walking compared with walking without wearing the exoskeleton. This advancement bridges the gap between in-lab exoskeleton technology and real-world human ambulation, making exoskeleton control technology viable for a broad community.
- Open Access
- 13 August 2022
- Published by Springer Nature in Annals of Biomedical Engineering
- Vol. 51 (2) , 410-421
Effects of Bilateral Assistance for Hemiparetic Gait Post-Stroke Using a Powered Hip Exoskeleton
Hemiparetic gait due to stroke is characterized by an asymmetric gait due to weakness in the paretic lower limb. These inter-limb asymmetries increase the biomechanical demand and reduce walking speed, leading to reduced community mobility and quality of life. With recent progress in the field of wearable technologies, powered exoskeletons have shown great promise as a potential solution for improving gait post-stroke. While previous studies have adopted different exoskeleton control methodologies for restoring gait post-stroke, the results are highly variable due to limited understanding of the biomechanical effect of exoskeletons on hemiparetic gait. In this study, we investigated the effect of different hip exoskeleton assistance strategies on gait function and gait biomechanics of individuals post-stroke. We found that, compared to walking without a device, powered assistance from hip exoskeletons improved stroke participants’ self-selected overground walking speed by 17.6 ± 2.5% and 11.1 ± 2.7% with a bilateral and unilateral assistance strategy, respectively (p < 0.05). Furthermore, both bilateral and unilateral assistance strategies significantly increased the paretic and non-paretic step length (p < 0.05). Our findings suggest that powered assistance from hip exoskeletons is an effective means to increase walking speed post-stroke and tuning the balance of assistance between non-paretic and paretic limbs (i.e., a bilateral strategy) may be most effective to maximize performance gains.
- 13 August 2022
- Published by Springer Nature in Annals of Biomedical Engineering
- Vol. 51 (2) , 410-421
Open AccessEffects of Bilateral Assistance for Hemiparetic Gait Post-Stroke Using a Powered Hip Exoskeleton
Hemiparetic gait due to stroke is characterized by an asymmetric gait due to weakness in the paretic lower limb. These inter-limb asymmetries increase the biomechanical demand and reduce walking speed, leading to reduced community mobility and quality of life. With recent progress in the field of wearable technologies, powered exoskeletons have shown great promise as a potential solution for improving gait post-stroke. While previous studies have adopted different exoskeleton control methodologies for restoring gait post-stroke, the results are highly variable due to limited understanding of the biomechanical effect of exoskeletons on hemiparetic gait. In this study, we investigated the effect of different hip exoskeleton assistance strategies on gait function and gait biomechanics of individuals post-stroke. We found that, compared to walking without a device, powered assistance from hip exoskeletons improved stroke participants’ self-selected overground walking speed by 17.6 ± 2.5% and 11.1 ± 2.7% with a bilateral and unilateral assistance strategy, respectively (p < 0.05). Furthermore, both bilateral and unilateral assistance strategies significantly increased the paretic and non-paretic step length (p < 0.05). Our findings suggest that powered assistance from hip exoskeletons is an effective means to increase walking speed post-stroke and tuning the balance of assistance between non-paretic and paretic limbs (i.e., a bilateral strategy) may be most effective to maximize performance gains.
- 21 June 2022
- Published by ASME International in Journal of Mechanisms and Robotics
- Vol. 15 (2)
Design and Validation of a Torque-Controllable Series Elastic Actuator-Based Hip Exoskeleton for Dynamic Locomotion
Series elastic actuators (SEAs) are increasingly popular in wearable robotics due to their high fidelity closed-loop torque control capability. Therefore, it has become increasingly important to characterize its performance when used in dynamic environments. However, the conventional design approach does not fully capture the complexity of the entire exoskeleton system. These limitations stem from identifying design criteria with inadequate biomechanics data, utilizing an off-the-shelf user interface, and applying a benchtop-based proportional-integral-derivative control for actual low-level torque tracking. While this approach shows decent actuator performance, it does not consider human factors such as the dynamic back-driving nature of human-exoskeleton systems as well as soft human tissue dampening during the load transfer. Using holistic design guidelines to improve the SEA-based exoskeleton performance during dynamic locomotion, our final system has an overall mass of 4.8 kg (SEA mass of 1.1 kg) and can provide a peak joint torque of 108 Nm with a maximum velocity of 5.2 rad/s. Additionally, we present a user state-based feedforward controller to further improve the low-level torque tracking for diverse walking conditions. Our study results provide future exoskeleton designers with a foundation to further improve SEA-based exoskeleton’s torque tracking response for maximizing human-exoskeleton performance during dynamic locomotion.
- 21 June 2022
- Published by ASME International in Journal of Mechanisms and Robotics
- Vol. 15 (2)
Design and Validation of a Torque-Controllable Series Elastic Actuator-Based Hip Exoskeleton for Dynamic Locomotion
Series elastic actuators (SEAs) are increasingly popular in wearable robotics due to their high fidelity closed-loop torque control capability. Therefore, it has become increasingly important to characterize its performance when used in dynamic environments. However, the conventional design approach does not fully capture the complexity of the entire exoskeleton system. These limitations stem from identifying design criteria with inadequate biomechanics data, utilizing an off-the-shelf user interface, and applying a benchtop-based proportional-integral-derivative control for actual low-level torque tracking. While this approach shows decent actuator performance, it does not consider human factors such as the dynamic back-driving nature of human-exoskeleton systems as well as soft human tissue dampening during the load transfer. Using holistic design guidelines to improve the SEA-based exoskeleton performance during dynamic locomotion, our final system has an overall mass of 4.8 kg (SEA mass of 1.1 kg) and can provide a peak joint torque of 108 Nm with a maximum velocity of 5.2 rad/s. Additionally, we present a user state-based feedforward controller to further improve the low-level torque tracking for diverse walking conditions. Our study results provide future exoskeleton designers with a foundation to further improve SEA-based exoskeleton’s torque tracking response for maximizing human-exoskeleton performance during dynamic locomotion.
- 7 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 69 (10) , 3234-3242
Subject-Independent Continuous Locomotion Mode Classification for Robotic Hip Exoskeleton Applications
Autonomous lower-limb exoskeletons must modulate assistance based on locomotion mode (e.g., ramp or stair ascent) to adapt to the corresponding changes in human biological joint dynamics. However, current mode classification strategies for exoskeletons often require user-specific tuning, have a slow update rate, and rely on additional sensors outside of the exoskeleton sensor suite. In this study, we introduce a deep convolutional neural network-based locomotion mode classifier for hip exoskeleton applications using an open-source gait biomechanics dataset with various wearable sensors. Our approach removed the limitations of previous systems as it is 1) subject-independent (i.e., no user-specific data), 2) capable of continuously classifying for smooth and seamless mode transitions, and 3) only utilizes minimal wearable sensors native to a conventional hip exoskeleton. We optimized our model, based on several important factors contributing to overall performance, such as transition label timing, model architecture, and sensor placement, which provides a holistic understanding of mode classifier design. Our optimized DL model showed a 3.13% classification error (steady-state: 0.80 $\pm$ 0.38% and transitional: 6.49 $\pm$ 1.42%), outperforming other machine learning-based benchmarks commonly practiced in the field (p<0.05). Furthermore, our multi-modal analysis indicated that our model can maintain high performance in different settings such as unseen slopes on stairs or ramps. Thus, our study presents a novel locomotion mode framework, capable of advancing robotic exoskeleton applications toward assisting community ambulation.
- 7 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 69 (10) , 3234-3242
Subject-Independent Continuous Locomotion Mode Classification for Robotic Hip Exoskeleton Applications
Autonomous lower-limb exoskeletons must modulate assistance based on locomotion mode (e.g., ramp or stair ascent) to adapt to the corresponding changes in human biological joint dynamics. However, current mode classification strategies for exoskeletons often require user-specific tuning, have a slow update rate, and rely on additional sensors outside of the exoskeleton sensor suite. In this study, we introduce a deep convolutional neural network-based locomotion mode classifier for hip exoskeleton applications using an open-source gait biomechanics dataset with various wearable sensors. Our approach removed the limitations of previous systems as it is 1) subject-independent (i.e., no user-specific data), 2) capable of continuously classifying for smooth and seamless mode transitions, and 3) only utilizes minimal wearable sensors native to a conventional hip exoskeleton. We optimized our model, based on several important factors contributing to overall performance, such as transition label timing, model architecture, and sensor placement, which provides a holistic understanding of mode classifier design. Our optimized DL model showed a 3.13% classification error (steady-state: 0.80 $\pm$ 0.38% and transitional: 6.49 $\pm$ 1.42%), outperforming other machine learning-based benchmarks commonly practiced in the field (p<0.05). Furthermore, our multi-modal analysis indicated that our model can maintain high performance in different settings such as unseen slopes on stairs or ramps. Thus, our study presents a novel locomotion mode framework, capable of advancing robotic exoskeleton applications toward assisting community ambulation.
- 25 May 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 68 (9) , 2870-2879
Biomechanical Comparison of Assistance Strategies Using a Bilateral Robotic Knee Exoskeleton
Despite there being studies that have investigated the effects of human augmentation using a knee exoskeleton, comparing different assistance schemes on a single knee exoskeleton has not been studied. Using a light-weight, low-profile bilateral knee exoskeleton system, this study examined and compared the biomechanical effects of three common assistance strategies (biological torque, impedance, and proportional myoelectric controllers) exhibiting different levels of flexibility for the user to control the assistance. Nine subjects walked on a 15% gradient incline surface at 1.1 m/s in the three powered conditions and with the exoskeleton unpowered. All the assistance strategies significantly reduced the metabolic cost of the users compared to the unpowered condition by 3.0% on average across strategies (p <; 0.05), led by the significant reduction in the biological knee kinetic effort and knee extensor muscle activation (p <; 0.05). Between assistance strategies, the metabolic cost and biomechanics displayed no statistically significant differences. The metabolic and biomechanical results indicate that powered extension assistance during early stance can improve performance compared to the unpowered condition. However, the user's ability to control the assistance may not be significant for human augmentation when walking on an inclined surface with a knee exoskeleton.
- 25 May 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Biomedical Engineering
- Vol. 68 (9) , 2870-2879
Biomechanical Comparison of Assistance Strategies Using a Bilateral Robotic Knee Exoskeleton
Despite there being studies that have investigated the effects of human augmentation using a knee exoskeleton, comparing different assistance schemes on a single knee exoskeleton has not been studied. Using a light-weight, low-profile bilateral knee exoskeleton system, this study examined and compared the biomechanical effects of three common assistance strategies (biological torque, impedance, and proportional myoelectric controllers) exhibiting different levels of flexibility for the user to control the assistance. Nine subjects walked on a 15% gradient incline surface at 1.1 m/s in the three powered conditions and with the exoskeleton unpowered. All the assistance strategies significantly reduced the metabolic cost of the users compared to the unpowered condition by 3.0% on average across strategies (p <; 0.05), led by the significant reduction in the biological knee kinetic effort and knee extensor muscle activation (p <; 0.05). Between assistance strategies, the metabolic cost and biomechanics displayed no statistically significant differences. The metabolic and biomechanical results indicate that powered extension assistance during early stance can improve performance compared to the unpowered condition. However, the user's ability to control the assistance may not be significant for human augmentation when walking on an inclined surface with a knee exoskeleton.
- 17 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 6 (2) , 3995-4000
Real-Time User-Independent Slope Prediction Using Deep Learning for Modulation of Robotic Knee Exoskeleton Assistance
Ground slope incline is a critical environmental variable that influences exoskeleton control parameters since human biological joint demand is correlated to changes in slope incline. Current literature methods take a heuristic approach by numerically calculating the slope incline from on-board mechanical sensors. However, these methods often require a user-specific tuning procedure and are prone to noise and sensor drift when tested in a dynamic setting, such as overground locomotion. In this study, we propose the use of a deep learning slope prediction model capable of generalizing across users and terrain. To evaluate this approach, we collected training data (N = 10) and utilized a convolutional neural network to predict the inclination angle and actively modulate the peak assistance magnitude of a bilateral robotic knee exoskeleton in real-time. From online validation results (N = 3), our model predicted the slope incline with an average RMSE of 1.5 ° during treadmill and overground walking. Furthermore, our model accurately predicted the slope incline in the extrapolated region outside of the training data with an average RMSE of 1.7 ° during treadmill and overground walking. Our study's findings showcase the feasibility of using deep learning models to actively modulate exoskeleton assistance, translating this technology to more realistic locomotion environments.
- 17 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 6 (2) , 3995-4000
Real-Time User-Independent Slope Prediction Using Deep Learning for Modulation of Robotic Knee Exoskeleton Assistance
Ground slope incline is a critical environmental variable that influences exoskeleton control parameters since human biological joint demand is correlated to changes in slope incline. Current literature methods take a heuristic approach by numerically calculating the slope incline from on-board mechanical sensors. However, these methods often require a user-specific tuning procedure and are prone to noise and sensor drift when tested in a dynamic setting, such as overground locomotion. In this study, we propose the use of a deep learning slope prediction model capable of generalizing across users and terrain. To evaluate this approach, we collected training data (N = 10) and utilized a convolutional neural network to predict the inclination angle and actively modulate the peak assistance magnitude of a bilateral robotic knee exoskeleton in real-time. From online validation results (N = 3), our model predicted the slope incline with an average RMSE of 1.5 ° during treadmill and overground walking. Furthermore, our model accurately predicted the slope incline in the extrapolated region outside of the training data with an average RMSE of 1.7 ° during treadmill and overground walking. Our study's findings showcase the feasibility of using deep learning models to actively modulate exoskeleton assistance, translating this technology to more realistic locomotion environments.