I-BaR: integrated balance rehabilitation framework

Neurological diseases are observed in approximately 1 billion people worldwide. A further increase is foreseen at the global level as a result of population growth and aging. Individuals with neurological disorders often experience cognitive, motor, sensory, and lower extremity dysfunctions. Thus, the possibility of falling and balance problems arise due to the postural control deficiencies that occur as a result of the deterioration in the integration of multi-sensory information. We propose a novel rehabilitation framework, Integrated Balance Rehabilitation (I-BaR), to improve the effectiveness of the rehabilitation with objective assessment, individualized therapy, convenience with different disability levels and adoption of assist-as-needed paradigm and, with integrated rehabilitation process as whole, that is, ankle-foot preparation, balance, and stepping phases, respectively. Integrated Balance Rehabilitation allows patients to improve their balance ability by providing multi-modal feedback: visual via utilization of virtual reality; vestibular via anteroposterior and mediolateral perturbations with the robotic platform; proprioceptive via haptic feedback.

Knee Angle Estimation from Surface EMG during Walking Using Attention-Based Deep Recurrent Neural Networks: Feasibility and Initial Demonstration in Cerebral Palsy

Accurately estimating knee joint angle during walking from surface electromyography (sEMG) signals can enable more natural control of wearable robotics like exoskeletons. However, challenges exist due to variability across individuals and sessions. This study evaluates an attention-based deep recurrent neural network combining gated recurrent units (GRUs) and an attention mechanism (AM) for knee angle estimation. Three experiments were conducted. First, the GRU-AM model was tested on four healthy adolescents, demonstrating improved estimation compared to GRU alone. A sensitivity analysis revealed that the key contributing muscles were the knee flexor and extensors, highlighting the ability of the AM to focus on the most salient inputs. Second, transfer learning was shown by pretraining the model on an open source dataset before additional training and testing on the four adolescents. Third, the model was progressively adapted over three sessions for one child with cerebral palsy (CP). The GRU-AM model demonstrated robust knee angle estimation across participants with healthy participants (mean RMSE 7 degrees) and participants with CP (RMSE 37 degrees). Further, estimation accuracy improved by 14 degrees on average across successive sessions of walking in the child with CP. These results demonstrate the feasibility of using attention-based deep networks for joint angle estimation in adolescents and clinical populations and support their further development for deployment in wearable robotics.

Patient's Healthy-Limb Motion Characteristic-Based Assist-As-Needed Control Strategy for Upper-Limb Rehabilitation Robots

The implementation of a progressive rehabilitation training model to promote patients' motivation efforts can greatly restore damaged central nervous system function in patients. Patients' active engagement can be effectively stimulated by assist-as-needed (AAN) robot rehabilitation training. However, its application in robotic therapy has been hindered by a simple determination method of robot-assisted torque which focuses on the evaluation of only the affected limb's movement ability. Moreover, the expected effect of assistance depends on the designer and deviates from the patient's expectations, and its applicability to different patients is deficient. In this study, we propose a control method with personalized treatment features based on the idea of estimating and mapping the stiffness of the patient's healthy limb. This control method comprises an interactive control module in the task-oriented space based on the quantitative evaluation of motion needs and an inner-loop position control module for the pneumatic swing cylinder in the joint space. An upper-limb endpoint stiffness estimation model was constructed, and a parameter identification algorithm was designed. The upper limb endpoint stiffness which characterizes the patient's ability to complete training movements was obtained by collecting surface electromyographic (sEMG) signals and human-robot interaction forces during patient movement. Then, the motor needs of the affected limb when completing the same movement were quantified based on the performance of the healthy limb. A stiffness-mapping algorithm was designed to dynamically adjust the rehabilitation training trajectory and auxiliary force of the robot based on the actual movement ability of the affected limb, achieving AAN control. Experimental studies were conducted on a self-developed pneumatic upper limb rehabilitation robot, and the results showed that the proposed AAN control method could effectively estimate the patient's movement needs and achieve progressive rehabilitation training. This rehabilitation training robot that simulates the movement characteristics of the patient's healthy limb drives the affected limb, making the intensity of the rehabilitation training task more in line with the patient's pre-morbid limb-use habits and also beneficial for the consistency of bilateral limb movements.

Monitoring Active Patient Participation During Robotic Rehabilitation: Comparison Between a Robot-Based Metric and an EMG-Based Metric

While rehabilitation robots present a much-needed solution to improving early mobilization therapy in demanding clinical settings, they also present new challenges and opportunities in patient monitoring. Aside from the fundamental challenge of quantifying a patient's voluntary contribution during robot-led therapy motion, many sensors cannot be used in clinical settings due to time and space limitations. In this paper, we present and compare two metrics for monitoring a patient's active participation in the motion. The two metrics, each derived from first principles, have the same biomechanical interpretability, i.e., active work by the patient during the robotic mobilization therapy, but are calculated in two different spaces (Cartesian vs. muscle space). Furthermore, the sensors used to quantify these two metrics are fully independent from each other and the associated measurements are unrelated. Specifically, the robot-based work metric utilizes robot-integrated force sensors, while the EMG-based work metric requires electrophysiological sensors. We then apply the two metrics to therapy performed using a clinically certified, commercially available robotic system and compare them against the specific instructions given to the healthy subjects as well as against each other. Both metric outputs qualitatively match the expected behavior of the healthy subjects. Additionally, strong correlations (median [Formula: see text]) are shown between the two metrics, not only for healthy subjects (n = 12) but also for patients (n = 2), providing solid evidence for their validity and translatability. Importantly, the robot-based work metric does not rely on any sensors outside of those integrated into the robot, thus making it ideal for application in clinical settings.

Using Deep Learning Models to Predict Prosthetic Ankle Torque

Inverse dynamics from motion capture is the most common technique for acquiring biomechanical kinetic data. However, this method is time-intensive, limited to a gait laboratory setting, and requires a large array of reflective markers to be attached to the body. A practical alternative must be developed to provide biomechanical information to high-bandwidth prosthesis control systems to enable predictive controllers. In this study, we applied deep learning to build dynamical system models capable of accurately estimating and predicting prosthetic ankle torque from inverse dynamics using only six input signals. We performed a hyperparameter optimization protocol that automatically selected the model architectures and learning parameters that resulted in the most accurate predictions. We show that the trained deep neural networks predict ankle torques one sample into the future with an average RMSE of 0.04 ± 0.02 Nm/kg, corresponding to 2.9 ± 1.6% of the ankle torque's dynamic range. Comparatively, a manually derived analytical regression model predicted ankle torques with a RMSE of 0.35 ± 0.53 Nm/kg, corresponding to 26.6 ± 40.9% of the ankle torque's dynamic range. In addition, the deep neural networks predicted ankle torque values half a gait cycle into the future with an average decrease in performance of 1.7% of the ankle torque's dynamic range when compared to the one-sample-ahead prediction. This application of deep learning provides an avenue towards the development of predictive control systems for powered limbs aimed at optimizing prosthetic ankle torque.

Attention-Based Deep Recurrent Neural Network to Estimate Knee Angle During Walking from Lower-Limb EMG

Accurate prediction of joint angle during walking from surface electromyography (sEMG) offers the potential to infer movement intention and therefore represents a potentially useful approach for adaptive control of wearable robotics. Here, we present the use of a recurrent neural network (RNN) with gated recurrent units (GRUs) and an attention mechanism to estimate knee angle during overground walking from sEMG and its initial offline validation in healthy adolescents. Our results show that the attention mechanism improved estimation accuracy by focusing on the most relevant parts of the input dataset within each time window, particularly muscles active during knee excursion. Sensitivity analysis revealed knee extensor and flexor muscles to be most salient in accurately estimating joint angle. Additionally, we demonstrate the ability of the GRU-RNN approach to accurately estimate knee angle during overground walking in a child with cerebral palsy (CP) in the presence of exoskeleton knee extension assistance. Collectively, our findings establish the initial feasibility of using this approach to estimate user movement from sEMG, which is particularly important for developing robotic exoskeletons for children with neuromuscular disorders such as CP.

An Attention-based Bidirectional LSTM Model for Continuous Cross-Subject Estimation of Knee Joint Angle during Running from sEMG Signals

Accurate and robust estimation of joint kinematics via surface electromyogram (sEMG) signals provides a human-machine interaction (HMI)-based method that can be used to adequately control rehabilitation robots while performing complex movements, such as running, for motor function restoration in affected individuals. To this end, this paper proposes a deep learning-based model (AM-BiLSTM) that integrates a bidirectional long short-term memory (BiLSTM) network and an attention mechanism (AM) for robust estimation of joint kinematics. The proposed model was appraised using knee joint kinematic and sEMG signals collected from fourteen subjects who performed running at the speed of 2 m/s. The proposed model's generalizability was tested for both within- and cross-subject scenarios and compared with long short-term memory (LSTM) and multi-layer perceptron (MLP) networks in terms of normalized root-mean-square error and correlation coefficient metrics. Based on the statistical tests, the proposed AM-BiLSTM model significantly outperformed the LSTM and MLP methods in both within- and cross-subject scenarios (p<0.05) and achieved state-of-the-art performance.Clinical Relevance- The promising results of this study suggest that the AM-BiLSTM model has the potential for continuous cross-subject estimation of lower limb kinematics during running, which can be used to control sEMG-driven exoskeleton robots oriented towards rehabilitation training.

Assist-As-Needed Control Strategy of Bilateral Upper Limb Rehabilitation Robot Based on GMM

Robotic-assisted rehabilitation therapy has been shown to be effective in improving upper limb motor function and the daily behavior of patients with motor dysfunction. At present, the majority of upper limb rehabilitation robots can only move in the two-dimensional plane, and cannot adjust the assistance mode in real-time according to the patient’s rehabilitation needs. In this paper, according to the shortcomings of the current rehabilitation robot only moving in the two-dimensional plane, a type of bilateral mirror upper limb rehabilitation robot structure with the healthy side assisting the affected side is proposed. This can move in three-dimensional space. Additionally, an assist-as-needed (AAN) control strategy for upper limb rehabilitation training is proposed based on the bilateral upper limb rehabilitation robot. The control strategy adopts Gaussian Mixture Model (GMM) and impedance controller to maximize the patient’s rehabilitation effect. In the task’s design, there is no need to rely on the assistance of the therapist, only the patients who completed the task independently. GMM guides the rehabilitation robot to provide different assistance for the patients at different task stages and induces the patients to complete the rehabilitation training independently by judging the extent to which the patients can complete the task. Furthermore, in this paper, the effectiveness of the proposed control strategy was verified by three volunteers participating in a two-dimensional task. The experimental results show that the proposed AAN control strategy can effectively provide appropriate assistance according to the classification stage of the interaction between the patients and the rehabilitation robot, and thus, patients can better achieve the rehabilitation effect during the rehabilitation task as much as possible.

A Mirror Bilateral Neuro-Rehabilitation Robot System with the sEMG-Based Real-Time Patient Active Participant Assessment

In this paper, a novel mirror visual feedback-based (MVF) bilateral neurorehabilitation system with surface electromyography (sEMG)-based patient active force assessment was proposed for upper limb motor recovery and improvement of limb inter-coordination. A mirror visual feedback-based human–robot interface was designed to facilitate the bilateral isometric force output training task. To achieve patient active participant assessment, an sEMG signals-based elbow joint isometric force estimation method was implemented into the proposed system for real-time affected side force assessment and participation evaluation. To assist the affected side limb efficiently and precisely, a mirror bilateral control framework was presented for bilateral limb coordination. Preliminary experiments were conducted to evaluate the estimation accuracy of force estimation method and force tracking accuracy of system performance. The experimental results show the proposed force estimation method can efficiently calculate the elbow joint force in real-time, and the affected side limb of patients can be assisted to track output force of the non-paretic side limb for better limb coordination by the proposed bilateral rehabilitation system.

Estimation of Lower Limb Kinematics during Squat Task in Different Loading Using sEMG Activity and Deep Recurrent Neural Networks

The aim of the present study was to predict the kinematics of the knee and the ankle joints during a squat training task of different intensities. Lower limb surface electromyographic (sEMG) signals and the 3-D kinematics of lower extremity joints were recorded from 19 body builders during squat training at four loading conditions. A long-short term memory (LSTM) was used to estimate the kinematics of the knee and the ankle joints. The accuracy, in terms root-mean-square error (RMSE) metric, of the LSTM network for the knee and ankle joints were 6.774 ± 1.197 and 6.961 ± 1.200, respectively. The LSTM network with inputs processed by cross-correlation (CC) method showed 3.8% and 4.7% better performance in the knee and ankle joints, respectively, compared to when the CC method was not used. Our results showed that in the prediction, regardless of the intensity of movement and inter-subject variability, an off-the-shelf LSTM decoder outperforms conventional fully connected neural networks.

Challenges and solutions for application and wider adoption of wearable robots

The science and technology of wearable robots are steadily advancing, and the use of such robots in our everyday life appears to be within reach. Nevertheless, widespread adoption of wearable robots should not be taken for granted, especially since many recent attempts to bring them to real-life applications resulted in mixed outcomes. The aim of this article is to address the current challenges that are limiting the application and wider adoption of wearable robots that are typically worn over the human body. We categorized the challenges into mechanical layout, actuation, sensing, body interface, control, human-robot interfacing and coadaptation, and benchmarking. For each category, we discuss specific challenges and the rationale for why solving them is important, followed by an overview of relevant recent works. We conclude with an opinion that summarizes possible solutions that could contribute to the wider adoption of wearable robots.

Robust Torque Predictions From Electromyography Across Multiple Levels of Active Exoskeleton Assistance Despite Non-linear Reorganization of Locomotor Output

Robotic exoskeletons can assist humans with walking by providing supplemental torque in proportion to the user's joint torque. Electromyographic (EMG) control algorithms can estimate a user's joint torque directly using real-time EMG recordings from the muscles that generate the torque. However, EMG signals change as a result of supplemental torque from an exoskeleton, resulting in unreliable estimates of the user's joint torque during active exoskeleton assistance. Here, we present an EMG control framework for robotic exoskeletons that provides consistent joint torque predictions across varying levels of assistance. Experiments with three healthy human participants showed that using diverse training data (from different levels of assistance) enables robust torque predictions, and that a convolutional neural network (CNN), but not a Kalman filter (KF), can capture the non-linear transformations in EMG due to exoskeleton assistance. With diverse training, the CNN could reliably predict joint torque from EMG during zero, low, medium, and high levels of exoskeleton assistance [root mean squared error (RMSE) below 0.096 N-m/kg]. In contrast, without diverse training, RMSE of the CNN ranged from 0.106 to 0.144 N-m/kg. RMSE of the KF ranged from 0.137 to 0.182 N-m/kg without diverse training, and did not improve with diverse training. When participant time is limited, training data should emphasize the highest levels of assistance first and utilize at least 35 full gait cycles for the CNN. The results presented here constitute an important step toward adaptive and robust human augmentation via robotic exoskeletons. This work also highlights the non-linear reorganization of locomotor output when using assistive exoskeletons; significant reductions in EMG activity were observed for the soleus and gastrocnemius, and a significant increase in EMG activity was observed for the erector spinae. Control algorithms that can accommodate spatiotemporal changes in muscle activity have broad implications for exoskeleton-based assistance and rehabilitation following neuromuscular injury.

Simultaneous Control of Tonic Vibration Reflex and Kinesthetic Illusion for Elbow Joint Motion Toward Novel Robotic Rehabilitation

Robotic rehabilitation is one of the most promising applications of robotic technologies. It is known that patients' active participation in rehabilitation is important for their recovery. On the other hand, mechanical vibration stimulation to muscles induces tonic vibration reflex (TVR) and kinesthetic illusion (KI) in the joint motion. In this paper, the possibility of a novel robotic rehabilitation method, in which the TVR is applied to an agonist muscle to enhance the intended motion of patients and the KI is simultaneously applied to an antagonist muscle to enhance the kinesthetic movement sensation of the generating intended motion by changing the frequency of vibration stimulation, is investigated. As the first step toward novel robotic rehabilitation, the proposed method is evaluated in elbow joint motion. The experimental results show the possibility of the proposed novel rehabilitation method.Clinical Relevance- This study shows the possibility of novel robotic rehabilitation.

Design and Implementation of a Robotic Hip Exoskeleton for Gait Rehabilitation

In this paper, a four degrees-of-freedom robotic hip exoskeleton was proposed for gait rehabilitation. The robotic hip exoskeleton was designed with active flexion/extension and passive abduction/adduction at each hip joint to comply with the movement of the thigh. Due to each user’s different lower limbs characteristics and unknown torques at hip joints, model-free linear extended state observer (LESO)-based controllers were proposed for rehabilitation gait control. The prototypes of the robotic hip exoskeleton and controller designs were validated and compared through walking and ascending rehabilitation experiments. Additionally, a motion captured system and EMG signals were used to investigate the walking assistance of the robotic hip exoskeleton.

Assist-As-Needed Exoskeleton for Hand Joint Rehabilitation Based on Muscle Effort Detection

Robotic-assisted systems have gained significant traction in post-stroke therapies to support rehabilitation, since these systems can provide high-intensity and high-frequency treatment while allowing accurate motion-control over the patient's progress. In this paper, we tackle how to provide active support through a robotic-assisted exoskeleton by developing a novel closed-loop architecture that continually measures electromyographic signals (EMG), in order to adjust the assistance given by the exoskeleton. We used EMG signals acquired from four patients with post-stroke hand impairments for training machine learning models used to characterize muscle effort by classifying three muscular condition levels based on contraction strength, co-activation, and muscular activation measurements. The proposed closed-loop system takes into account the EMG muscle effort to modulate the exoskeleton velocity during the rehabilitation therapy. Experimental results indicate the maximum variation on velocity was 0.7 mm/s, while the proposed control system effectively modulated the movements of the exoskeleton based on the EMG readings, keeping a reference tracking error <5%.

Improvement of EMG Pattern Recognition Model Performance in Repeated Uses by Combining Feature Selection and Incremental Transfer Learning

Electromyography (EMG) pattern recognition is one of the widely used methods to control the rehabilitation robots and prostheses. However, the changes in the distribution of EMG data due to electrodes shifting results in classification decline, which hinders its clinical application in repeated uses. Adaptive learning can solve this problem but takes additional time. To address this, an efficient scheme is developed by comparing the performance of 12 combinations of three feature selection methods [no feature selection (NFS), sequential forward search (SFS), and particle swarm optimization (PSO)] and four classification methods [non-adaptive support vector machine (N-SVM), incremental SVM (I-SVM), SVM based on TrAdaBoost (T-SVM), and I-SVM based on TrAdaBoost (TI-SVM)] in the classification of EMG data of 12 subjects for 5 consecutive days. Our results showed that TI-SVM achieved the highest classification accuracy among the classification methods (p < 0.05). The SFS method achieved the same classification accuracy as that of the scheme trained with the feature vectors selected by the NFS method (p = 0.999) while achieving a lower training time than that of TI-SVM combined with the NFS method (p = 0.043). Although the PSO method outperformed the NFS and SFS methods by achieving reduced training and response times (p < 0.05), the PSO method achieved a considerably lower classification accuracy than that of the scheme trained with the feature vectors selected by the NFS (p = 0.001) or SFS (p = 0.001) method. Furthermore, TI-SVM combined with the SFS method outperformed the CNN method with fine-tuning in classification accuracy on a small data set (p = 0.001). The results indicate that TI-SVM combined with the SFS method is suitable for improving the performance of EMG pattern recognition in repeated uses.

An Assist-as-Needed Controller for Passive, Assistant, Active, and Resistive Robot-Aided Rehabilitation Training of the Upper Extremity

Clinical studies have demonstrated that robot-involved therapy can effectively improve the rehabilitation training effect of motor ability and daily behavior ability of subjects with an upper limb motor dysfunction. This paper presents an impedance-based assist-as-needed controller that can be used in robot-aided rehabilitation training for subjects with an upper extremity dysfunction. Then, the controller is implemented on an end-effector upper extremity rehabilitation robot which could assist subjects in performing training with a spatial trajectory. The proposed controller enables subjects’ arms to have motion freedom by building a fault-tolerant region around the rehabilitation trajectory. Subjects could move their upper limb without any assistance within the fault-tolerant region while the robot would provide assistance according to the subjects’ functional ability when deviating from the fault-tolerant region. Besides, we also put forward the stiffness field around the fault-tolerant region to increase the robot’s assistance when subjects’ hand is moving outside the fault-tolerant region. A series of columnar rigid walls would be constructed in the controller according to the subjects’ functional ability, and the stiffness of the wall increases as the motion performance deteriorates. Furthermore, the controller contains five adjustable parameters. The controller would show different performances by adjusting these parameters and satisfy the requirement of robot-aided rehabilitation training at different rehabilitation stages such as passive, assistant, active, and resistant training. Finally, the controller was tested with an elderly female participant with different controller parameters, and experimental results verified the correctness of the controller and its potential ability to satisfy the training requirements at different rehabilitation stages. In the close future, the proposed controller in this work is planned to be applied on more subjects and also patients who have upper limb motor dysfunctions to demonstrate performance of the controller with different parameters.

Performance-Based Hybrid Control of a Cable-Driven Upper-Limb Rehabilitation Robot

Patients after stroke may have different rehabilitation needs due to various levels of disability. To satisfy such needs, a performance-based hybrid control is proposed for a cable-driven upper-limb rehabilitation robot (CDULRR). The controller includes three working modes, i.e., resistance mode, assistance mode and restriction mode, which are switched by the tracking error since it is a common index to represent motor performance. In resistance mode, the proper damping force would be provided for subjects, which is in the opposite direction to the actual velocity. In assistance mode, a method of adjusting stiffness coefficient by fuzzy logic is adopted to provide suitable assistance to help subjects. In restriction mode, the damping force is applied again to limit the movement and ensure the safety. To verify the effectiveness of the controller, the task-oriented experiments with different disturbance were conducted by ten healthy subjects. The experiments results demonstrated that the controller can adjust working modes by the subjects' motor performance. It was found that, as the increasing disturbance led to a decrease in the motor performance, the robot provided more assistance in the trainings. Adaptive adjustment of damping force and stiffness coefficient allowed the controller to induce more active effort.

Continuous Estimation of Knee Joint Angle Based on Surface Electromyography Using a Long Short-Term Memory Neural Network and Time-Advanced Feature

Continuous joint angle estimation based on a surface electromyography (sEMG) signal can be used to improve the man-machine coordination performance of the exoskeleton. In this study, we proposed a time-advanced feature and utilized long short-term memory (LSTM) with a root mean square (RMS) feature and its time-advanced feature (RMSTAF; collectively referred to as RRTAF) of sEMG to estimate the knee joint angle. To evaluate the effect of joint angle estimation, we used root mean square error (RMSE) and cross-correlation coefficient ρ between the estimated angle and actual angle. We also compared three methods (i.e., LSTM using RMS, BPNN (back propagation neural network) using RRTAF, and BPNN using RMS) with LSTM using RRTAF to highlight its good performance. Five healthy subjects participated in the experiment and their eight muscle (i.e., rectus femoris (RF), biceps femoris (BF), semitendinosus (ST), gracilis (GC), semimembranosus (SM), sartorius (SR), medial gastrocnemius (MG), and tibialis anterior (TA)) sEMG signals were taken as algorithm inputs. Moreover, the knee joint angles were used as target values. The experimental results showed that, compared with LSTM using RMS, BPNN using RRTAF, and BPNN using RMS, the average RMSE values of LSTM using RRTAF were respectively reduced by 8.57%, 46.62%, and 68.69%, whereas the average ρ values were respectively increased by 0.31%, 4.15%, and 18.35%. The results demonstrated that LSTM using RRTAF, which contained the time-advanced feature, had better performance for estimating the knee joint motion.

Voluntary Control of an Ankle Joint Exoskeleton by Able-Bodied Individuals and Stroke Survivors Using EMG-Based Admittance Control Scheme

Control schemes based on electromyography (EMG) have demonstrated their superiority in human-robot cooperation due to the fact that motion intention can be well estimated by EMG signals. However, there are several limitations due to the noisy nature of EMG signals and the inaccuracy of EMG-force/torque estimation, which might deteriorate the stability of human-robot cooperation movement. To improve the movement stability, an EMG-based admittance control scheme (EACS) was proposed, comprised of an EMG-driven musculoskeletal model (EDMM), an admittance filter and an inner position controller. To investigate the performance of EACS, a series of sinusoidal tracking tasks were conducted with 12 healthy participants and 4 stroke survivors in an ankle exoskeleton in comparison with the EMG-based open-loop control scheme (EOCS). The experimental results indicated that both EACS and EOCS could improve stroke survivors' ankle range of motion (ROM). The experimental results of both healthy participants and stroke survivors showed that the assistance torque, tracking error and jerk values of EACS were lower than those of EOCS. The interaction torque of EACS decreased towards the increasing assistance ratio while that of EOCS increased. Moreover, the EMG levels of tibialis anterior (TA) decreased towards the increasing assistance ratio but were higher than those of EOCS. EACS was effective in improving movements stability, and had the potential to be applied in robot-assisted rehabilitation training to address the foot-drop problem.

Reference Trajectory Reshaping Optimization and Control of Robotic Exoskeletons for Human-Robot Co-Manipulation

For human-robot co-manipulation by robotic exoskeletons, the interaction forces provide a communication channel through which the human and the robot can coordinate their actions. In this article, an optimization approach for reshaping the physical interactive trajectory is presented in the co-manipulation tasks, which combines impedance control to enable the human to adjust both the desired and the actual trajectories of the robot. Different from previous studies, the proposed method significantly reshapes the desired trajectory during physical human-robot interaction (pHRI) based on force feedback, without requiring constant human guidance. The proposed scheme first formulates a quadratically constrained programming problem, which is then solved by neural dynamics optimization to obtain a smooth and minimal-energy trajectory similar to the natural human movement. Then, we propose an adaptive neural-network controller based on the barrier Lyapunov function (BLF), which enables the robot to handle the uncertain dynamics and the joint space constraints directly. To validate the proposed method, we perform experiments on the exoskeleton robot with human operators for co-manipulation tasks. The experimental results demonstrate that the proposed controller could complete the co-manipulation tasks effectively.

Processing Surface EMG Signals for Exoskeleton Motion Control

The surface electromyography (sEMG) signal has been used for volitional control of robotic assistive devices. There are still challenges in improving system performance accuracy and signal processing to remove systematic noise. This study presents procedures and a pilot validation of the EMG-driven speed-control of exoskeleton and integrated treadmill with a goal to provide better interaction between a user and the system. The gait cycle duration (GCD) was extracted from sEMG signals using the autocorrelation algorithm and Bayesian fusion algorithm. GCDs of various walking speeds were then programmed to control the motion speed of exoskeleton robotic system. The performance and efficiency of this sEMG-controlled robotic assistive ambulation system was tested and validated among 6 healthy volunteers. The results demonstrated that the autocorrelation algorithm extracted the GCD from individual muscle contraction. The GCDs of individual muscles had variability between different walking steps under a designated walking speed. Bayesian fusion algorithms processed the GCDs of multiple muscles yielding a final GCD with the least variance. The fused GCD effectively controlled the motion speeds of exoskeleton and treadmill. The higher amplitude of EMG signals with shorter GCD was found during a faster walking speed. The algorithms using fused GCDs and gait stride length yielded trajectory joint motion tracks in a shape of sine curve waveform. The joint angles of the exoskeleton measured by a decoder mounted on the hip turned out to be in sine waveforms. The hip joint motion track of the exoskeleton matched the angles projected by trajectory curve generated by computer algorithms based on the fused GCDs with high agreement. The EMG-driven speed-control provided the human-machine inter-limb coordination mechanisms for an intuitive speed control of the exoskeleton-treadmill system at the user's intents. Potentially the whole system can be used for gait rehabilitation of incomplete spinal cord hemispheric stroke patients as goal-directed and task-oriented training tool.

Bi-Level Adaptive Computed-Current Impedance Controller for Electrically Driven Robots

This paper presents a bi-level adaptive computed-current impedance controller for electrically driven robots. This study aims to reduce calculation complexities by utilizing the electrical equations of actuators, instead of the entire model of the electromechanical system. Moreover, taking the dynamical effects of mechanical parts into account through the current’s feedback, external disturbances are compensated. In order to handle uncertainties, a bi-level optimization problem is formulated to obtain guaranteed stability besides the estimation convergence. An adaptation rule and its optimal tuning gain are achieved. The proposed method is applied to control of a rehabilitation robot to evaluate its performance.

Adjusting Assistance Commensurates with Patient Effort During Robot-Assisted Upper Limb Training for a Patient with Spasticity After Cervical Spinal Cord Injury: A Case Report

Limited evidence is available on optimal patient effort and degree of assistance to achieve preferable changes during robot-assisted training (RAT) for spinal cord injury (SCI) patients with spasticity. To investigate the relationship between patient effort and robotic assistance, we performed training using an electromyography-based robotic assistance device (HAL-SJ) in an SCI patient at multiple settings adjusted to patient effort. In this exploratory study, we report immediate change in muscle contraction patterns, patient effort, and spasticity in a 64-year-old man, diagnosed with cervical SCI and with American Spinal Injury Association Impairment Scale C level and C4 neurological level, who underwent RAT using HAL-SJ from post-injury day 403. Three patient effort conditions (comfortable, somewhat hard, and no-effort) by adjusting HAL-SJ’s assists were set for each training session. Degree of effort during flexion and extension exercise was assessed by visual analog scale, muscle contraction pattern by electromyography, modified Ashworth scale, and maximum elbow extension and flexion torques, immediately before and after each training session, without HAL-SJ. The amount of effort during training with the HAL-SJ at each session was evaluated. The degree of effort during training can be set to three effort conditions as we intended by adjusting HAL-SJ. In sessions other than the no-effort setting, spasticity improved, and the level of effort was reduced immediately after training. Spasticity did not decrease in the training session using HAL-SJ with the no-effort setting, but co-contraction further increased during extension after training. Extension torque was unchanged in all sessions, and flexion torque decreased in all sessions. When performing upper-limb training with HAL-SJ in this SCI patient, the level of assistance with some effort may reduce spasticity and too strong assistance may increase co-contraction. Sometimes, a patient’s effort may be seemingly unmeasurable; hence, the degree of patient effort should be further measured.

Design of physical user–robot interactions for model identification of soft actuators on exoskeleton robots

Recent breakthroughs in wearable robots, such as exoskeleton robots with soft actuators and soft exosuits, have enabled the use of safe and comfortable movement assistance. However, modeling and identification methods for soft actuators used in wearable robots have yet to be sufficiently explored. In this study, we propose a novel approach for obtaining accurate soft actuator models through the design of physical user–robot interactions for wearable robots, in which the user applies external forces to the robot. To obtain an accurate soft actuator model from the limited amount of data acquired through an interaction, we leverage an active learning framework based on Gaussian process regression. We conducted experiments using a two-degree-of-freedom upper-limb exoskeleton robot with four pneumatic artificial muscles (PAMs). Experimental results showed that physical interactions between the exoskeleton robot and the user were successfully designed to allow PAM models to be identified. Furthermore, we found that data acquired through an interaction could result in more accurate soft actuator models for the exoskeleton robots than data acquired without a physical interaction between the exoskeleton robot and the user.

Development, Dynamic Modeling, and Multi-Modal Control of a Therapeutic Exoskeleton for Upper Limb Rehabilitation Training

Robot-assisted training is a promising technology in clinical rehabilitation providing effective treatment to the patients with motor disability. In this paper, a multi-modal control strategy for a therapeutic upper limb exoskeleton is proposed to assist the disabled persons perform patient-passive training and patient-cooperative training. A comprehensive overview of the exoskeleton with seven actuated degrees of freedom is introduced. The dynamic modeling and parameters identification strategies of the human-robot interaction system are analyzed. Moreover, an adaptive sliding mode controller with disturbance observer (ASMCDO) is developed to ensure the position control accuracy in patient-passive training. A cascade-proportional-integral-derivative (CPID)-based impedance controller with graphical game-like interface is designed to improve interaction compliance and motivate the active participation of patients in patient-cooperative training. Three typical experiments are conducted to verify the feasibility of the proposed control strategy, including the trajectory tracking experiments, the trajectory tracking experiments with impedance adjustment, and the intention-based training experiments. The experimental results suggest that the tracking error of ASMCDO controller is smaller than that of terminal sliding mode controller. By optimally changing the impedance parameters of CPID-based impedance controller, the training intensity can be adjusted to meet the requirement of different patients.

Nondestructive Estimation of Muscle Contributions to STS Training with Different Loadings Based on Wearable Sensor System

Partial body weight support or loading sit-to-stand (STS) rehabilitation can be useful for persons with lower limb dysfunction to achieve movement again based on the internal residual muscle force and external assistance. To explicate how the muscles contribute to the kinetics and kinematics of STS performance by non-invasive in vitro detection and to nondestructively estimate the muscle contributions to STS training with different loadings, a wearable sensor system was developed with ground reaction force (GRF) platforms, motion capture inertial sensors and electromyography (EMG) sensors. To estimate the internal moments of hip, knee and ankle joints and quantify the contributions of individual muscle and gravity to STS movement, the inverse dynamics analysis on a simplified STS biomechanical model with external loading is proposed. The functional roles of the lower limb individual muscles (rectus femoris (RF), gluteus maximus (GM), vastus lateralis (VL), tibialis anterior (TA) and gastrocnemius (GAST)) during STS motion and the mechanism of the muscles' synergies to perform STS-specific subtasks were analyzed. The muscle contributions to the biomechanical STS subtasks of vertical propulsion, anteroposterior (AP) braking and propulsion for body balance in the sagittal plane were quantified by experimental studies with EMG, kinematic and kinetic data.