Robot-assisted rehabilitation, which can provide repetitive, intensive and high-precision physics training, has a positive influence on motor function recovery of stroke patients. Current robots need to be more intelligent and more reliable in clinical practice. Machine learning algorithms (MLAs) are able to learn from data and predict future unknown conditions, which is of benefit to improve the effectiveness of robot-assisted rehabilitation. In this paper, we conduct a focused review on machine learning-based methods for robot-assisted upper limb rehabilitation. Firstly, the current status of upper rehabilitation robots is presented. Then, we outline and analyze the designs and applications of MLAs for upper limb movement intention recognition, human-robot interaction control and quantitative assessment of motor function. Meanwhile, we discuss the future directions of MLAs-based robotic rehabilitation. This review article provides a summary of MLAs for robotic upper limb rehabilitation and contributes to the design and development of future advanced intelligent medical devices.

Motor Imagery (MI) electroencephalography (EEG) is widely studied for its non-invasiveness, easy availability, portability, and high temporal resolution. As for MI EEG signal processing, the high dimensions of features represent a research challenge. It is necessary to eliminate redundant features, which not only create an additional overhead of managing the space complexity, but also might include outliers, thereby reducing classification accuracy. The firefly algorithm (FA) can adaptively select the best subset of features, and improve classification accuracy. However, the FA is easily entrapped in a local optimum. To solve this problem, this paper proposes a method of combining the firefly algorithm and learning automata (LA) to optimize feature selection for motor imagery EEG. We employed a method of combining common spatial pattern (CSP) and local characteristic-scale decomposition (LCD) algorithms to obtain a high dimensional feature set, and classified it by using the spectral regression discriminant analysis (SRDA) classifier. Both the fourth brain–computer interface competition data and real-time data acquired in our designed experiments were used to verify the validation of the proposed method. Compared with genetic and adaptive weight particle swarm optimization algorithms, the experimental results show that our proposed method effectively eliminates redundant features, and improves the classification accuracy of MI EEG signals. In addition, a real-time brain–computer interface system was implemented to verify the feasibility of our proposed methods being applied in practical brain–computer interface systems.

To realize fully functioning inland waterway communications, portable and affordable unmanned surface vehicles (USVs) have become increasingly significant in wireless inland ship networks. However, USVs are struggling in handling resource-intensive services due to the limited battery and CPU resources. In addition, the power of signal transmitted between USVs and the shore base stations suffers from considerable attenuation. In this paper, an intelligent reflecting surface (IRS)-aided wireless inland ship mobile edge computing (MEC) network architecture is proposed to support USVs’ communication and computation. The minimization of network energy consumption is formulated under a number of QoS requirements. To achieve an efficient solution to the formulated challenging problem, a low-complexity enhanced block coordinate descent algorithm is proposed and thus the original optimization problem can be divided into a list of sub-problems. The extensive results verify the effectiveness of jointly considering USVs’ offloading decisions, USVs’ transmission power, unmanned aerial vehicle (UAV) trajectory and IRS phase shift-vector design for energy efficiency in a spare wireless inland ship MEC network and show that the proposed algorithm can decrease energy consumption by approximately 75% compared to a baseline algorithm without optimization design. Also, the network performance under different IRS elements and accuracy parameters are investigated.

Since surface electromyograghic (sEMG) signals are non-invasive and capable of reflecting humans’ motion intention, they have been widely used for the motion recognition of upper limbs. However, limited research has been conducted for lower limbs, because the sEMGs of lower limbs are easily affected by body gravity and muscle jitter. In this paper, sEMG signals and accelerometer signals are acquired and fused to recognize the motion patterns of lower limbs. A curve fitting method based on median filtering is proposed to remove accelerometer noise. As for movement onset detection, an sEMG power spectral correlation coefficient method is used to detect the start and end points of active signals. Then, the time-domain features and wavelet coefficients of sEMG signals are extracted, and a dynamic time warping (DTW) distance is used for feature extraction of acceleration signals. At last, five lower limbs’ motions are classified and recognized by using Gaussian kernel-based linear discriminant analysis (LDA) and support vector machine (SVM) respectively. The results prove that the fused feature-based classification outperforms the classification with only sEMG signals or accelerometer signals, and the fused feature can achieve 95% or higher recognition accuracy, demonstrating the validity of the proposed method.

Fault diagnosis is one of the important applications of edge computing in the Industrial Internet of Things (IIoT). To address the issue that traditional fault diagnosis methods often struggle to effectively extract fault features, this paper proposes a novel rolling bearing fault diagnosis method that integrates Gramian Angular Field (GAF), Convolutional Neural Network (CNN), and Vision Transformer (ViT). First, GAF is used to convert one-dimensional vibration signals from sensors into two-dimensional images, effectively retaining the fault features of the vibration signal. Then, the CNN branch is used to extract the local features of the image, which are combined with the global features extracted by the ViT branch to diagnose the bearing fault. The effectiveness of this method is validated with two datasets. Experimental results show that the proposed method achieves average accuracies of 99.79% and 99.63% on the CWRU and XJTU-SY rolling bearing fault datasets, respectively. Compared with several widely used fault diagnosis methods, the proposed method achieves higher accuracy for different fault classifications, providing reliable technical support for performing complex fault diagnosis on edge devices.