Neural Network Structure-Based Adaptive SMC for Pneumatic Artificial Muscle Systems with State Constraints and Input Dead Zones

Pneumatic artificial muscle (PAM) systems demonstrate superior compliance and practicality when driving robotic exoskeletons. However, filling them with highly compressed gas makes PAMs susceptible to external disturbances and system uncertainties, which may degrade state response and increase control difficulty. Furthermore, most existing controllers require linearization operations. To address this, this paper proposes a state-constrained adaptive neural network controller that overcomes the deadband obstacle within a finite time interval, achieving satisfactory motion control that precisely tracks the operational trajectory of pneumatic artificial muscles. Simultaneously, it ensures the output state converges within a finite time without relying on parameter design. Rigorous stability analysis is provided, and the tracking performance of this method is validated through simulation.