sEMG and IMU Data-Based Angle Prediction-Based Model-Free Control Strategy for Exoskeleton-Assisted Rehabilitation

Abstract

Exoskeleton-assisted rehabilitation necessitates specific methodologies for the accurate prediction of motorized limb joint angles to achieve targeted rehabilitation training. In this article, surface electromyographic (sEMG) and inertial measurement unit (IMU) data-based angle prediction-based model-free control strategy (SAPMFCS) is proposed. First, a hybrid model integrating convolutional neural network (CNN) with bidirectional long short-term memory (LSTM), named CNN-BiLSTM, is employed for real-time prediction of elbow joint angle. Second, time delay estimation-variable gain sliding model controller (TDE-VGSMC) is developed to employ the predicted joint angle as the desired trajectory to facilitate the completion of corresponding rehabilitation exercises. Semiphysical and real-time experiments show that the enhanced efficacy demonstrated by the SAPMFCS introduced in this article suggests a potential enhancement in the versatility and applicability of exoskeleton-assisted rehabilitation.