Mitigating nonlinear torque oscillations in pneumatic systems with adaptive RBFNN compensated gain-adaptive ADRC

Pneumatic rotary actuators are widely employed in industrial and aerospace applications due to their high power density and cleanliness. However, the strong nonlinearities arising from air compressibility and model uncertainties cause overshoot and oscillations, severely impacting control accuracy. Despite the proven effectiveness of active disturbance rejection control (ADRC) in handling nonlinear systems, two critical limitations hinder its application in pneumatic servos: first, the bandwidth constraints of the extended state observer (ESO) result in inaccurate estimation of time-varying nonlinearities; second, the numerous parameters of ADRC pose significant challenges for tuning. To address these issues, a gain-adaptive active disturbance rejection control strategy based on an adaptive radial basis function neural network compensation (RBFGAADRC) is presented, which improves upon ADRC in two key aspects. First, to overcome ESO limitations, an adaptive radial basis function neural network is employed to replace the ESO, leveraging its superior approximation capability to achieve more accurate estimation and compensation of nonlinearities. Second, to simplify parameter tuning, a gain-adaptive iterative strategy is introduced to dynamically adjust the parameters of the nonlinear error feedback controller, significantly simplifying the tuning process. Experimental results demonstrate that RBFGAADRC achieves a steady-state error of 0.062 Nm without overshoot and a maximum tracking error of 0.35 Nm without oscillations in sinusoidal torque trajectory tracking at 0.5 Hz. These results validate the effectiveness of the proposed method in improving the control performance of pneumatic servo systems.