Data-driven policy learning strategy for nonlinear robust control with unknown perturbation

摘要

In this paper, we propose a robust optimal control policy for nonlinear systems with bounded unknown perturbation by using data-driven policy learning strategy. The robust control problem is transformed into a corresponding optimal control design with specific cost function. Neural-network-based data-driven policy learning strategy is presented to solve the problem without system dynamics. The solution of the optimal control problem can asymptotically stabilize the unknown system. An example is given to illustrate the established method.