Model Predictive Control of Pneumatic Artificial Muscle Antagonist Joint Angle Tracking Based on BP Neural Network Modelling

A back propagation neural network prediction method is designed for the predictive control of joint angles in pneumatic artificial muscles in this paper, aimed at improving trajectory tracking accuracy for antagonistic joints. Firstly, a back propagation neural network model of pneumatic muscle antagonistic joints (PMAJ) is constructed based on the joint error angle and the error angular velocity. Secondly, the BPNN model is discretized to derive the back propagation neural network prediction model for antagonistic joints. Subsequently, a back propagation neural network-based prediction controller is designed to control the antagonistic joints, and the system’s stability is demonstrated. Finally, the efficacy of the proposed algorithm is demonstrated through simulations and physical experiments in comparison with other algorithms. The results show that the backpropagation neural network model predictive control (BP NNMPC) can significantly improve trajectory tracking accuracy. The key innovation of this work lies in integrating a data-driven back propagation neural network predictor into the model predictive control framework, thereby enhancing the accuracy of the nonlinear muscle dynamics model. Compared to conventional model predictive control, the overall error rate of the BP NNMPC designed in this paper is reduced by 40.5%.