Design of a Bio-Inspired Adaptive Central Pattern Generator Control for Pneumatic Muscle Antagonistic Joint

Abstract

This brief proposes a control scheme for an antagonistic joint driven by pneumatic artificial muscles (PAMs). Departing from the common practice of treating central pattern generators (CPGs) as trajectory planners that feed a secondary controller, our architecture elevates the CPGs network to the primary controller, issuing control inputs directly. Inspired by biological motor circuits, this yields a tightly integrated and efficient control loop. The method also removes reliance on an explicit internal dynamic model, thereby avoiding singular behavior, numerical fragility in inverse-dynamics calculations, and the heavy computation of inertia and Coriolis terms. Consequently, the architecture is simpler, achieves better real-time responsiveness, and generalizes well across diverse trajectories and operating conditions. The robustness of the proposed approach is validated through simulations and hardware experiments, which demonstrate smooth and precise tracking performance in the presence of perturbations and measurement noise. Moreover, in a highly dynamic scenario, the chattering amplitude is reduced by 75.7% compared with sliding mode control (SMC) and 68.4% compared with proportional–integral–derivative control. These enhancements improve system stability and alleviate actuator stress. In addition, this scheme pushes forward bio-inspired robotic systems, providing autonomous, flexible, and high-efficiency approaches for real-time control applications.