This article proposes a variable-damping prosthetic knee (ReKnee) with a novel hydraulic damper, and the intelligent control system is proposed. The LSTM network is proposed to identify the locomotion pattern under five road conditions. Predictive control is first applied to the hydraulic damping control for prosthetic knee. Prototype experiments are carried out to verify the performance of the control system. The experiment results show that LSTM possesses better accuracy and robustness than the traditional RNN network and other shallow machine learning algorithms, and NNPC damping control improves the gait symmetry compared to fuzzy logic control with decreasing in the range of 5.75–19.27% of the SI values when walking speed varies. It is demonstrated that the ReKnee can make a good performance on enhancing the accuracy of angle tracking and the approximation of normal gait characteristics.
Read full abstract