Gait Prediction for Rehabilitation Robots Based on Deep Learning

Abstract

Lower limb rehabilitation exoskeleton robots are used for rehabilitation training of patients with lower limb disorders, which can help patients improve their motor ability. This paper proposes a gait prediction model to predict the gait trajectory of the wearer to improve the human-robot cooperation ability of lower limb rehabilitation exoskeleton robots and the wearing experience of patients. The gait trajectory prediction model is constructed based on convolutional neural network (CNN), long short-term memory network (LSTM), and attention mechanism. For training the model, human gait data is collected utilizing the motion capture equipment. In detail, the spatial coordinates of each lower limb joint are used as the input vector, and the angles of hip and knee joints are used as the output and feedback input vectors. Results show that the model achieves excellent performance on predicting the movements of human lower limbs. The model has theoretical significance for improving the motion control and operation performance of exoskeleton robot.