Improved fast non-singular adaptive super-twisting sliding mode control based on radial basis function neural network approximation for robot joint module.

This study proposes an improved fast non-singular adaptive super-twisting control scheme based on neural network to address the precise control issues of robot joint modules. Firstly, to facilitate the application of advanced control algorithms, a second-order state-space model of the joint module considering nonlinear friction and stiffness is established using the Lagrangian energy equation method. Then, an improved fast non-singular terminal sliding surface is proposed to avoid singularity and accelerate convergence. Subsequently, an equivalent control law compensator using a radial basis function neural network is designed to counteract the impact of uncertain model factors, thereby ensuring the system states closely follow the sliding surface. Furthermore, an adaptive switching control law is designed that does not require precise disturbance information, enhancing its practicality for engineering applications. Finally, simulation and experimental results under two different reference trajectories and in the presence of external disturbances demonstrate the superior trajectory tracking capability and disturbance rejection performance of the proposed control scheme.