The scarcity of high-quality surface electromyography (sEMG) data, caused by ethical constraints, privacy concerns, and noise interference, poses significant challenges for developing robust deep learning models in sEMG analysis. Multi-channel sEMG signals exhibit complex inter-channel correlations reflecting neuromuscular coordination. However, existing generative methods suffer from error accumulation in sequential channel generation, insufficient inter-channel relationship modeling, and lack of physiological constraints, producing data-driven valid but physiologically implausible signals that compromise biological fidelity for clinical applications. To address these fundamental limitations, we propose a Muscle Synergy-Constrained Conditional GAN (MS-cGAN) framework that simultaneously generates multi-channel sEMG signals while preserving bio-mechanical fidelity. Firstly, A novel Graph Convolutional Network (GCN)-based generator architecture specifically tailored for sparse sEMG signals, which enables the generator to capture and model complex inter-channel relationship features through graph-based representation learning, thereby circumventing error accumulation issues by leveraging the inherent inter-channel correlations. Secondly, Integration of Muscle Synergy (MS) prior constraints as dynamic loss functions based on MS theory, which enforces generator optimization within a physiologically plausible parameter space and ensures signals maintain synergistic consistency with underlying physiological mechanisms. Lastly, experiments on IRASS datasets and public datasets (NinaPro DB1 and DB2) demonstrate that MS-cGAN significantly improves signal authenticity and enhances downstream task performance compared to traditional GANs and state-of-the-art diffusion models. The generated data effectively supplement scarce sEMG datasets and improve kinematic prediction precision for deep learning models.

Steady-state visual evoked potential (SSVEP)-based brain-computer interfaces (BCIs) can achieve high recognition accuracy with sufficient training data. Transfer learning presents a promising solution to alleviate data requirements for the target subject by leveraging data from source subjects; however, effectively addressing individual variability among both target and source subjects remains a challenge. This paper proposes a novel transfer learning framework, termed instance-based task-related component analysis (iTRCA), which leverages knowledge from source subjects while considering their individual contributions. iTRCA extracts two types of features: (1) the subject-general feature, capturing shared information between source and target subjects in a common latent space, and (2) the subject-specific feature, preserving the unique characteristics of the target subject. To mitigate negative transfer, we further design an enhanced framework, subject selection-based iTRCA (SS-iTRCA), which integrates a similarity-based subject selection strategy to identify appropriate source subjects for transfer based on their task-related components (TRCs). Comparative evaluations on the Benchmark, BETA, and a self-collected dataset demonstrate the effectiveness of the proposed iTRCA and SS-iTRCA frameworks. This study provides a potential solution for developing high-performance SSVEP-based BCIs with reduced target subject data.

The key to achieving assist-as-needed (AAN) control in rehabilitation robots lies in accurately predicting patient motion intentions. This study, for the first time, redefines motion intention prediction from the perspective of sequence-to-sequence translation by analogizing sEMG signals and joint angles to the source language and target language, respectively. The proposed 3DCNN-TF model achieves precise translation of neural control signals into kinematic representations. This model comprises three modules: an sEMG “sentence” generation module that compiles multiple sEMG sliding windows into a “sentence,” a 3DCNN module based on muscle anatomy and electrode placement to extract muscle synergy features from each “word” in the “sentence,” and a Transformer (TF) module that autoregressively generates the next joint angle as the translation result. Experimental results indicate that the 3DCNN-TF model achieves superior overall performance compared to eight baseline models and existing studies in continuously predicting wrist and knee flexion/extension angles across varying speeds. Moreover, the 3DCNN-TF achieves an optimal balance between prediction accuracy and computational efficiency while exhibiting exceptional robustness and generalizability. Specifically, the 3DCNN-TF achieves average nRMSE and R2values of (6.2%/95.5%) and (5.5%/96.2%) on wrist and knee datasets, respectively, with an average training time of less than two minutes. Additionally, the 3DCNN-TF can predict joint angles up to 300 ms in advance without compromising accuracy, which is critical for real-time AAN control in rehabilitation robots.

Optimizing the feature representation and decoding efficiency of the steady-state visual evoked potentials (SSVEP) is critical to enhance the performance of neural signal decoding systems. Current deep learning models usually overlook the physical topological information of EEG channels, resulting in suboptimal feature extraction and limited recognition performance. To address these challenges, this study proposes a synergistically designed SSVEP recognition framework to alleviate data insufficiency, improve the feature representation, and enhance decoding efficiency. Specifically, a slicing-and-scaling technique is adopted to improve the model generalization under limited-sample scenarios. A graph-based spatial filter leverages the topological relationships among EEG channels to suppress redundant information and enhance spatial feature quality. A lightweight convolutional neural network (CNN) with fewer parameters is developed to efficiently extract discriminative temporal–spatial features for accurate SSVEP classification. Experimental results on two public benchmark datasets and one self-collected dataset demonstrate that the proposed framework outperforms baseline deep learning models, yielding improvements of at least 6.8 %, 8.5 %, and 0.5 % in peak average classification accuracy, respectively. The maximum average information transfer rates (ITRs) achieved on the three datasets were 221.4 ,106.7 , and 133.9 , respectively. By simultaneously reducing model complexity and improving decoding performance, the proposed framework offers an effective and promising approach for efficient neural signal decoding in SSVEP recognition.

This article focuses on the trajectory tracking problem of mobile manipulators (MMs). Firstly, we construct a position and orientation model predictive tracking control (POMPTC) scheme for mobile manipulators. The proposed POMPTC scheme can simultaneously minimize the tracking error, joint velocity, and joint acceleration. Moreover, it can achieve synchronous control for the position and orientation of the end-effector. Secondly, a finite-time convergent neural dynamics (FTCND) model is constructed to find the optimal solution of the POMPTC scheme. Then, based on the proposed POMPTC scheme, a non-singular fast terminal sliding model (NFTSM) control method is presented, which considers the disturbances caused by the floating-base on the manipulator at the dynamic level. It can achieve finite-time tracking performance and improve the anti-disturbances ability. Finally, simulation and experiments show that the proposed control method has the advantages of strong robustness, fast convergence, and high control accuracy.

Post-stroke upper limb dysfunction severely impacts patients' daily life quality. Utilizing sEMG signals to predict patients' motion intentions enables more effective rehabilitation by precisely adjusting the assistance level of rehabilitation robots. Employing the muscle synergy (MS) features can establish more accurate and robust mappings between sEMG and motion intentions. However, traditional matrix factorization algorithms based on blind source separation still exhibit certain limitations in extracting MS features. This paper proposes four deep learning models to extract MS features from four distinct perspectives: spatiotemporal convolutional kernels, compression and reconstruction of sEMG, graph topological structure, and the anatomy of target muscles. Among these models, the one based on 3DCNN predicts motion intentions from the muscle anatomy perspective for the first time. It reconstructs 1D sEMG samples collected at each time point into 2D sEMG frames based on the anatomical distribution of target muscles and sEMG electrode placement. These 2D frames are then stacked as video segments and input into 3DCNN for MS feature extraction. Experimental results on both our wrist motion dataset and public Ninapro DB2 dataset demonstrate that the proposed 3DCNN model outperforms other models in terms of prediction accuracy, robustness, training efficiency, and MS feature extraction for continuous prediction of wrist flexion/extension angles. Specifically, the average nRMSE and R2values of 3DCNN on these two datasets are (0.14/0.93) and (0.04/0.95), respectively. Furthermore, compared to existing studies, the 3DCNN outperforms musculoskeletal models based on direct collocation optimization, physics-informed GANs, and CNN-LSTM-based deep Kalman filter models when evaluated on our dataset.

Upper limb rehabilitation robot can assist stroke patients to complete daily activities to promote the recovery of upper-limb motor functions. However, the robot uncertainty and the patient’s unconscious disturbance impose great difficulties on the high-performance trajectory tracking of the rehabilitation robot. In this paper, an uncertainty compensated high-order adaptive iterative learning controller (UCHAILC) is proposed to reduce the impact of uncertainty from inside and outside of the robot during the rehabilitation process. The nonlinear system is converted into a dynamic linearization model with uncertainty compensation, and the optimization criterion method is adopted to estimate the pseudo-partial derivative (PPD) parameters and the uncertainty respectively, then the previous iterations are used to update the current parameters through a high-order learning scheme. The convergence of UCHAILC is theoretically proved. Simulation and control experiments on a rehabilitation robot are given to validate the effectiveness of the proposed method, which is significant to improve the training security and physiotherapy effect of robot-assisted rehabilitation.Note to Practitioners—This paper was motivated by the need to assist stroke patients to restore motor function for executing daily activities. The inherent difficulties lie in reducing the tracking errors of rehabilitation robots caused by uncertainty and involuntary disturbance from patients to avoid secondary injury. The proposed UCHAILC can transform the complex nonlinear system into a dynamic linear model with uncertainty compensation, then the PPD parameters and uncertainty are estimated through high-order learning law. Theoretical analysis, simulation, and experiments verified the feasibility of the method. Furthermore, the proposed controller is not limited to the dynamic model and hardware driving mode of the robot system, which can be easily transplanted to other nonlinear control systems with uncertainties.

The assessment of implant status and complications of Total Hip Replacement (THR) relies mainly on the clinical evaluation of the X-ray images to analyse the implant and the surrounding rigid structures. Current clinical practise depends on the manual identification of important landmarks to define the implant boundary and to analyse many features in arthroplasty X-ray images, which is time-consuming and could be prone to human error. Semantic segmentation based on the Convolutional Neural Network (CNN) has demonstrated successful results in many medical segmentation tasks. However, these networks cannot define explicit properties that lead to inaccurate segmentation, especially with the limited size of image datasets. Our work integrates clinical knowledge with CNN to segment the implant and detect important features simultaneously. This is instrumental in the diagnosis of complications of arthroplasty, particularly for loose implant and implant-closed bone fractures, where the location of the fracture in relation to the implant must be accurately determined. In this work, we define the points of interest using Gruen zones that represent the interface of the implant with the surrounding bone to build a Statistical Shape Model (SSM). We propose a multitask CNN that combines regression of pose and shape parameters constructed from the SSM and semantic segmentation of the implant. This integrated approach has improved the estimation of implant shape, from 74% to 80% dice score, making segmentation realistic and allowing automatic detection of Gruen zones. To train and evaluate our method, we generated a dataset of annotated hip arthroplasty X-ray images that will be made available.

Brain-computer interface (BCI) can provide a direct communication path between the human brain and an external device. The steady-state visual evoked potential (SSVEP)-based BCI has been widely explored in the past decades due to its high signal-to-noise ratio and fast communication rate. Several spatial filtering methods have been developed for frequency detection. However the temporal knowledge contained in the SSVEP signal is not effectively utilized. In this study, we propose a canonical correlation analysis (CCA)-based spatio-temporal filtering method to improve target classification. The training signal and two types of template signals (i.e. individual template and artificial sine-cosine reference) are first augmented via temporal information. Three sets of augmented data are then concatenated by trials. The CCA is performed twice, between the newly obtained training data and each template. The trained four spatial filters can be applied in the following test process. A public benchmark dataset was used to evaluate the performance of the proposed method and the other three comparing methods, such as CCA, MsetCCA, and TRCA. The experimental results indicate that the proposed method yields significantly higher performance. This paper also explored the effects of the number of electrodes and training blocks on classification accuracy. The results further demonstrated the effectiveness of the proposed method in SSVEP detection.

This article considers the bearing-only formation control problem, where the control of each agent only relies on relative bearings of their neighbors. A new control law is proposed to achieve target formations in finite time. Different from the existing results, the control law is based on a time-varying scaling gain. Hence, the convergence time can be arbitrarily chosen by users, and the derivative of the control input is continuous. Furthermore, sufficient conditions are given to guarantee almost global convergence and interagent collision avoidance. Then, a leader-follower control structure is proposed to achieve global convergence. By exploring the properties of the bearing Laplacian matrix, the collision avoidance and smooth control input are preserved. A multirobot hardware platform is designed to validate the theoretical results. Both simulation and experimental results demonstrate the effectiveness of our design.

Convolutional neural networks (CNNs) have been widely utilized to identify hand gestures from surface electromyography (sEMG) signals. However, due to the nonstationary characteristics of sEMG, the classification accuracy usually degrades significantly in the daily living environment involving complex hand movements. To further improve the reliability of a classifier, unconfident classifications are expected to be identified and rejected. In this study, we propose a novel approach to estimate the probability of correctness for each classification. Specifically, a confidence estimation model is established to generate confidence scores (ConfScore) based on posterior probabilities of CNN, and an objective function is designed to train the parameters of this model. In addition, a comprehensive metric that combines the true acceptance rate (TAR) and the true rejection rate (TRR) is proposed to evaluate the rejection performance of ConfScore, so that the tradeoff between system security and control lag could be fully considered. The effectiveness of ConfScore is verified using data from public databases and our online platform. The experimental results illustrate that ConfScore can better reflect the correctness of CNN classifications than traditional confidence features, i.e., maximum posterior probability and entropy of the probability vector. Moreover, the rejection performance is observed to be less sensitive to variations in rejection thresholds.

To develop multi-functionalhuman-machine interfaces that can help disabled people reconstruct lost functions of upper-limbs, machine learning (ML) and deep learning (DL) techniques have been widely implemented to decode human movement intentions from surface electromyography (sEMG) signals. However, due to the high complexity of upper-limb movements and the inherent non-stable characteristics of sEMG, the usability of ML/DL based control schemes is still greatly limited in practical scenarios. To this end, tremendous efforts have been made to improve model robustness, adaptation, and reliability. In this article, we provide a systematic review on recent achievements, mainly from three categories: multi-modal sensing fusion to gain additional information of the user, transfer learning (TL) methods to eliminate domain shift impacts on estimation models, and post-processing approaches to obtain more reliable outcomes. Special attention is given to fusion strategies, deep TL frameworks, and confidence estimation. Research challenges and emerging opportunities, with respect to hardware development, public resources, and decoding strategies, are also analysed to provide perspectives for future developments.

Data-driven spatial filtering approaches have been widely used for steady-state visual evoked potentials (SSVEPs) detection toward the brain-computer interface (BCI). The existing methods tend to learn the spatial filter parameters for a certain stimulation frequency only using the training trials from the same stimulus, which may ignore the information from the other stimuli. In this paper, we propose a novel multi-objective optimisation-based spatial filtering method for enhancing SSVEP recognition. Spatial filters are defined via maximising the correlation among the training data from the same stimulus whilst minimising the correlation from different stimuli. We collected SSVEP signals using 16 electrodes from six healthy subjects at 4 different stimulation frequencies: 14Hz, 15Hz, 16Hz, and 17Hz. The experimental study was implemented, and our method can achieve an average recognition accuracy of 94.17%, which illustrates its effectiveness.

In this paper, we propose a novel iterative learning control (ILC) scheme for precise state tracking of pneumatic muscle (PM) actuators. Two critical issues are considered in our scheme: 1) state constraints on PM position and velocity; 2) uncertainties of the PM model. Based on the three-element form, a PM model is constructed that takes both parametric and nonparametric uncertainties into consideration. By introducing the composite energy function (CEF) approach incorporated with a barrier Lyapunov function (BLF), full state constraints of PM will not be violated and uncertainties are effectively compensated. Through rigorous analysis, we show that under proposed ILC scheme, uniform convergence of PM state tracking errors are guaranteed. Simulation results validate the performance of the proposed scheme.

EMG-driven musculoskeletal model has been broadly used to detect human intention in rehabilitation robots. This approach computes muscle-tendon force and translates it to the joint kinematics. However, the muscle-tendon parameters of the musculoskeletal model are difficult to measure in vivo and varied across subjects. In this study, a direct collocation (DC) method is proposed to optimize the subject-specific parameters in a wrist musculoskeletal model. The resultant optimized parameters are used to estimate the wrist flexion/extension motion. The estimation performance is compared with the parameters optimized by the genetic algorithm. Experiment results show that the DC methods have a similar performance compared with GA, in which the mean correlation are 0.96 and 0.93 for the genetic algorithm and DC method respectively. But the direction collocation method requires less optimization time.

A previously developed pneumatic muscles' (PMs) actuated gait exoskeleton (with only knee joint) has been demonstrated in achieving appropriate actuation torque, range of motion (ROM), and control bandwidth for task-specific gait training. While the adopted multi-input-multi-output (MIMO) sliding mode (SM) strategy has preliminarily implemented simultaneous control of the exoskeleton's angular trajectory and compliance, its efficacy with human users during gait cycles has not been investigated. This article presents an improved bi-joint gait rehabilitation exoskeleton (BiGREX) with integrated human hip and knee joints. The results with 12 healthy subjects demonstrated that the system's compliance can be effectively adjusted while guiding the subjects walking in predefined trajectories.

While newer designs and control approaches are being proposed for rehabilitation robots, vital information from the human musculoskeletal system should also be considered. Incorporating knowledge about joint biomechanics during the development of robot controllers can enhance the safety and performance of robot-aided treatments. In this article, the optimal path or trajectories of a parallel ankle rehabilitation robot were generated by minimizing joint reaction moments and the tension along ligaments and muscle-tendon units. The simulations showed that using optimized robot paths, user efforts could be reduced to 80%, thereby ensuring less strain on weaker or stiffer ligaments, etc. Additionally, to limit the moments applied by the robot in stiff or constrained directions, the intended robot path was modified to move the commanded position in the direction opposite to that of the position error. Such online modification of the robot path can lead to a reduction in forces applied by a robot to the subject. Simulation results and experimental findings with healthy subjects using an ankle rehabilitation robot prototype and subsequent statistical analysis further validated that path modification based on ankle joint biomechanics results in a reduction in undesired forces experienced by human users during treatment.

In the past decades, classical machine learning (ML) methods have been widely investigated in wrist kinematics estimation for the control of prosthetic hands. Currently deeper structures have shown great potential to further improve prediction accuracy. In this paper we present a single stream convolutional neural network (CNN) for mapping surface electromyography (sEMG) to wrist angles within three degrees-of-freedom (DOFs). Two types of two dimensional (2D) sEMG images are constructed in time domain and spectrum as CNN inputs, respectively. Six typical linear and nonlinear ML models are implemented for comparison, where four efficient time-spatial hand-crafted features are extracted to represent feature engineering. Experiment results with four able-bodied participants illustrate that CNN with 2D spectrum sEMG images can achieve highest accuracy in most testing sessions. In other sessions, it is still competitive to the most promising ML techniques. The core strength of deep learning (DL), i.e. feature learning via deep structures and efficient algorithms, is verified to be more powerful than classical feature engineering, particularly in smaller datasets.

To solve the poor compliance and safety problems in current rehabilitation robots, a novel two-degrees-of-freedom (2-DOF) soft ankle rehabilitation robot driven by pneumatic muscles (PMs) is presented, taking advantages of the PM's inherent compliance and the parallel structure's high stiffness and payload capacity. However, the PM's nonlinear, time-varying and hysteresis characteristics, and the coupling interference from parallel structure, as well as the unpredicted disturbance caused by arbitrary human behavior all raise difficulties in achieving high-precision control of the robot. In this paper, a multi-input-multi-output disturbance compensated sliding mode controller (MIMO-DCSMC) is proposed to tackle these problems. The proposed control method can tackle the un-modeled uncertainties and the coupling interference existed in multiple PMs' synchronous movement, even with the subject's participation. Experiment results on a healthy subject confirmed that the PMs-actuated ankle rehabilitation robot controlled by the proposed MIMO-DCSMC is able to assist patients to perform high-accuracy rehabilitation tasks by tracking the desired trajectory in a compliant manner.

This paper describes the design analysis and optimization of a novel 3-degrees of freedom (DOF) wearable parallel robot developed for ankle rehabilitation treatments. To address the challenges arising from the use of a parallel mechanism, flexible actuators, and the constraints imposed by the ankle rehabilitation treatment, a complete robot design analysis is performed. Three design stages of the robot, namely, kinematic design, actuation design, and structural design are identified and investigated, and, in the process, six important performance objectives are identified which are vital to achieve design goals. Initially, the optimization is performed by considering only a single objective. Further analysis revealed that some of these objectives are conflicting, and hence these are required to be simultaneously optimized. To investigate a further improvement in the optimal values of design objectives, a preference-based approach and evolutionary-algorithm-based nondominated sorting algorithm (NSGA II) are adapted to the present design optimization problem. Results from NSGA II are compared with the results obtained from the single objective optimization and preference-based optimization approaches. It is found that NSGA II is able to provide better design solutions and is adequate to optimize all of the objective functions concurrently. Finally, a fuzzy-based ranking method has been devised and implemented in order to select the final design solution from the set of nondominated solutions obtained through NSGA II. The proposed design analysis of parallel robots together with the multiobjective optimization and subsequent fuzzy-based ranking can be generalized with modest efforts for the development of all of the classes of parallel robots.

Intervention of robotic devices in the field of physical gait therapy can help in providing repetitive, systematic, and economically viable training sessions. Interactive or assist-as-needed (AAN) gait training encourages patient voluntary participation in the robotic gait training process which may aid in rapid motor function recovery. In this paper, a lightweight robotic gait training orthosis with two actuated and four passive degrees of freedom (DOFs) is proposed. The actuated DOFs were powered by pneumatic muscle actuators. An AAN gait training paradigm based on adaptive impedance control was developed to provide interactive robotic gait training. The proposed adaptive impedance control scheme adapts the robotic assistance according to the disability level and voluntary participation of human subjects. The robotic orthosis was operated in two gait training modes, namely, inactive mode and active mode, to evaluate the performance of the proposed control scheme. The adaptive impedance control scheme was evaluated on ten neurologically intact subjects. The experimental results demonstrate that an increase in voluntary participation of human subjects resulted in a decrease of the robotic assistance and vice versa. Further clinical evaluations with neurologically impaired subjects are required to establish the therapeutic efficacy of the adaptive-impedance-control-based AAN gait training strategy.

Cadence or stride frequency is an important parameter being controlled in gait training of neurologically impaired subjects. The aim of this study was to examine the effects of cadence variation on muscle activation patterns during robot assisted unimpaired gait using dynamic simulations. A twodimensional (2-D) musculoskeletal model of human gait was developed considering eight major muscle groups along with existing ground contact force (GCF) model. A 2-D model of a robotic orthosis was also developed which provides actuation to the hip, knee and ankle joints in the sagittal plane to guide subjects limbs on reference trajectories. A custom inverse dynamics algorithm was used along with a quadratic minimization algorithm to obtain a feasible set of muscle activation patterns. Predicted patterns of muscle activations during slow, natural and fast cadence were compared and the mean muscle activations were found to be increasing with an increase in cadence. The proposed dynamic simulation provide important insight into the muscle activation variations with change in cadence during robot assisted gait and provide the basis for investigating the influence of cadence regulation on neuromuscular parameters of interest during robot assisted gait.

Robot-assisted gait therapy is an emerging rehabilitation practice. This paper presents new experimental results with an intrinsically compliant robotic gait training orthosis and a trajectory tracking controller. The intrinsically compliant robotic orthosis has six degrees of freedom. Sagittal plane hip and knee joints were powered by the actuation of pneumatic muscle actuators in opposing pair configuration. The orthosis has passive hip abduction/adduction joint and passive mechanisms to allow vertical and lateral translations of the trunk. A passive foot lifter having a spring mechanism was used to ensure sufficient dorsiflexion during swing phase. A trajectory tracking controller based on a chattering-free robust variable structure control law was implemented in joint space to guide the subject's limbs on physiological gait trajectories. The performance of the robotic orthosis was evaluated during two gait training modes, namely, “trajectory tracking mode with maximum compliance” and “trajectory tracking mode with minimum compliance.” The experimental evaluations were carried out with ten neurologically intact subjects. The results show that the robotic orthosis is able to perform the gait training task during the two gait training modes. All the subjects tend to deviate from the reference joint angle trajectories with an increase in robotic compliance as the subjects have more freedom to voluntarily drive the robotic orthosis.

The principals and procedure of long bone surgery are presented and the need for robotic assistance is established. Existing problems include radiation exposure from fluoroscopy, mental strain from reconstructing 3-dimensional images and physical fatigue from overcoming fracture deforming forces. These problems are addressed by a proposed fracture reduction robot which aids in treatment planning and reduction of the fracture. A flexible parallel robot (FleP) with an active force/position controller is designed to perform the operation. A computer-aided planning treatment tool (CAPTT) provides image analysis, path planning and simulation. An advanced human machine interface (AHMI) attempts to provide a companion feeling between robot and surgeon by using human forms of communication. The result is a reduction in radiation exposure, removal of the need to reconstruct images mentally and there is no longer any physical strain