Abstract: McKibben muscles are increasingly used in many robotic applications due to their advantages of lightweight, compliant, and skeletal muscles-like behaviours. However, there are still huge challenges in the motion control of McKibben muscles due to the system nonlinearity (e.g., hysteresis) and model uncertainties. To investigate the control issues, a soft artificial arm actuated by an antagonistic pair of McKibben muscles, mimicking the biological structure of skeleton-muscle systems, is developed. Inspired by the biological motor control capability that humans can control and coordinate a group of muscles to achieve complex motions, a cerebellum-like controller based on Spiking Neural Networks (SNNs) is employed for the motion control of the developed artificial arm. Benefit from the employment of the SNN-based cerebellar model, the proposed control scheme provides online adaptive learning capability, good computational efficiency, fast response, and strong robustness. Finally, several simulations and experiments are conducted subject to different environmental disturbances. Both simulation and experimental results verify that the proposed method can achieve good tracking performance, adaptability, and strong robustness.

Key words: McKibben muscle , · Cerebellum-like controller , · Spiking Neural Network , · Soft artificial arm , · Adaptive control