Robust control for affine nonlinear system with unknown time‐varying uncertainty under reinforcement learning framework

This paper investigates the adaptive robust control problem based on reinforcement learning for an affine nonlinear system with unknown time-varying uncertainty. Inspired by ability to estimate uncertainty of neural network, a novel policy iteration algorithm is proposed which alternates between value evaluation, estimation, and update steps until law obtained. Especially during step approximated radial basis function network introduce it into framework. By designing appropriate utility function, improves both convergence rate final approximate error comparing existing algorithm. The Lyapunov stability theorem provides theoretical demonstrations convergence. Furthermore, uniformly ultimately bounded demonstrated Finally, performance through torsion pendulum system.