Dielectric elastomer actuators (DEAs) are a promising technology for soft robotics. The use of DEAs has many advantages, including light weight, resilience, and fast response for its applications, such as grippers, artificial muscles, and heel strike generators. Grippers are commonly used as grasping devices. In this study, we focus on DEA applications and propose a technology to expand the applicability of a soft gripper. The advantages of gripper-based DEAs include light weight, fast response, and low cost. We fabricated soft grippers using multiple DEA layers. The grippers successfully held or gripped an object, and we investigated the response time of the grippers and their angle characteristics. We studied the relationship between the number of DEA layers and the performance of our grippers. Our experimental results show that the multi-layered DEAs have the potential to be strong grippers.

Objective To evaluate the existing evidence of a machine learning-based classification system that stratifies patients with stroke. Methods The authors carried out a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) recommendations for a review article. PubMed, MEDLINE, Web of Science, and CINAHL Plus Full Text were searched from January 2015 to February 2021. Results There are twelve studies included in this systematic review. Fifteen algorithms were used in the included studies. The most common forms of machine learning (ML) used to classify stroke patients were the support vector machine (SVM) (n = 8 studies), followed by random forest (RF) (n = 7 studies), decision tree (DT) (n = 4 studies), gradient boosting (GB) (n = 4 studies), neural networks (NNs) (n = 3 studies), deep learning (n = 2 studies), and k-nearest neighbor (k-NN) (n = 2 studies), respectively. Forty-four features of inputs were used in the included studies, and age and gender are the most common features in the ML model. Discussion There is no single algorithm that performed better or worse than all others at classifying patients with stroke, in part because different input data require different algorithms to achieve optimal outcomes.

Human-movement recognition is a novel challenge in soft robotics. In recent years, there have been several attempts to develop soft wearable devices for supporting human-robot interfaces. Many algorithms and programming languages are available to integrate a wearable device with a soft robot. One such promising algorithm is reservoir computing (RC), which includes of a group of recurrently and randomly connected nodes. The RC model can easily process multidimensional signal data and can handle nonlineardata and has been extensively used in robotic control. It has been reported that the RC algorithm can speed up network training and solve complex data sets. However, the main existing limitations in handlocomotion classification are the considerable run-time and the delayed response. In this study, we figure out the best machine learning algorithms to handle three-dimensional hand-gesture data. We employ a two-part strategy: a loopback filter is included in the preprocessing of the initial dataset to support the 3dimensional (3D) signs of each hand posture; subsequently, the training dataset is applied to the machine learning algorithm which includes an artificial neural network (ANN), convolutional neural network (CNN), long short-term memory(LSTM), and reservoir computing(RC). Each training network is optimized with various hyperparameters. Furthermore, we compare the performance of several machine-learning algorithms in classifying the three-dimensional hand-signal posture data. The results show that the classification of nonlinear hand-locomotion signals by RC requires a comparatively shorter training duration (12 minutes for training times), and that optimal accuracy 94.17, precision 94.10, and recall 93.99 is realized for time series data.INDEX TERMS Human-machine interface, human hand-locomotion signal, reservoir computing, time series, multi dimension, and nonlinear data.

This research applied the convolutional neural network (CNN algorithm) to determine the misalignment of vertebral column from the processed image. The raw data was the 3D-computerized tomography (CT) provided by the Suranaree University of Technology Hospital, Nakhon Ratchasima, Thailand. There were 93 data sets that comprised 40 data of misalignment vertebral columns. These studies first extracted front, rear, left, and right images of the vertebral column from 3D CT images by RadiAnt Program (Version 2020.2). In the second step, the images were processed by the Ridge detection algorithm with various parameters. The combinations processed were of sigma 1, 4, 7, and 10 with the two low-high thresholds, 10-30 and 20-20. The last step was about the Python code development (with Tensorflow, Numpy, and Sklearn libraries) for creating the model to classify the normal and abnormal vertebral column image sets by the CNN algorithm. The best model could perform very well. The model with Ridge detection preprocessing of parameters sigma=7, low threshold=20, and high threshold=20 performed faultlessly. The performance was accuracy 100 percent, precision 100 percent, and recall 100 percent.