Neural Network Adaptive Control for Pneumatic Muscle Joint Systems with Unknown Nonsymmetric Actuator Dead-Zone

In this paper, the adaptive control problem of pneumatic muscle (PM) joint systems with external disturbance, reconstruction error, and actuator dead-zone is researched. Compared with the existing results, the tracking performance of the PM joint system has been enhanced, and the system now guarantees the boundedness of the actuator output. First, an adaptive neural network is used to estimate the unknown dynamical behavior and the external disturbance of the system online, enabling real-time estimation of systematic errors. A static neural network is constructed to compensate for the unknown asymmetric dead-zone nonlinearity of the actuator. Second, an online robust update term is introduced to counteract reconstruction errors of the neural network and the external disturbance. Third, the Lyapunov theory is used to derive a smooth control law, ensuring the stability of the system and rigorously proving the uniform ultimate boundedness of the weight parameters of each neural network. Finally, the feasibility and effectiveness of the proposed scheme are demonstrated through simulations.