Neural networks adaptive predefined-time control for pure-feedback nonlinear systems: a case study on robotic exoskeleton systems

A predefined-time (PT) tracking adaptive control method is studied for non-affine pure-feedback nonlinear systems, with an emphasis on its practical application in robotic exoskeleton technology. A novel PT neural networks control algorithm is implemented, by leveraging the approximation capabilities of neural networks, backstepping technique, barrier functions and Mean Value Theorem. The neural networks are used to approximate the unknown nonlinearities inherent in the system's control dynamics, while the adaptive law is meticulously designed based on the PT Lyapunov stability criterion. By Lyapunov PT theory, the developed methodology guarantees the system's convergence within a pre-established time, therefore offering enhanced performance over conventional fixed-time control methodologies. Simulation results validate the efficacy of this proposed control approach, demonstrating its practical implications for controlling robotic exoskeletons under state constraints, thus validating its potential for real-world applications.