A new control for the pneumatic muscle bionic legged robot based on neural network

The bionic joints composed of pneumatic muscles (PMs) can simulate the motion biological joints. However, PMs themselves have non-linear characteristics such as hysteresis and creep, which make it difficult to achieve high-precision trajectory tracking control PM-driven robots. In order effectively suppress adverse effects non-linearity on performance improve dynamic legged robot, this study designs a double closed-loop structure based neural network. First, according model joint, mapping between PM contraction force joint torque is proposed. Second, strategy designed for inner loop outer angle. loop, feedforward neuron Proportional-Integral-Derivative controller three-element model. sliding mode robust with local approximation by using radial basis function network capability. Finally, verified simulation physical experiments that suitable humanoid antagonistic joints, satisfy requirements reliability, flexibility, bionics during human–robot collaboration.