Adaptive position control of pneumatic artificial muscles using an RBF-tuned single-neuron PID controller

Purpose This paper aims to improve the position regulation performance of pneumatic artificial muscle (PAM) systems by developing an adaptive outer-loop control strategy capable of handling nonlinear and time-varying dynamics. Design/methodology/approach An adaptive single-neuron proportional–integral–derivative (SN-PID) controller is proposed, which tunes its gains in real time using a radial basis function (RBF) neural network. The controller incorporates Hebbian learning and Jacobian-based feedback to enhance adaptability. Three experimental scenarios were implemented: position tracking across various load and reference conditions, dynamic transitions under sequential step inputs and disturbance rejection under sudden external load changes. Findings The SN-PID controller consistently outperformed a conventional proportional–integral–derivative controller in all test scenarios. It achieved up to 86% overshoot reduction, nearly 90% faster settling during step transitions and halved the recovery time under load disturbances ranging from 5 to 10 kg, while effectively suppressing oscillations. Originality/value The study introduces a biologically inspired, adaptive control approach for real-time outer-loop regulation in PAM systems. The proposed SN-PID controller enhances the robustness and responsiveness of PAM actuators, making it a promising solution for real-time soft robotics and assistive device applications.