Stabilization of Continuous-discrete Fractional-order 2D Roesser Model via Reinforcement Learning and Linear Matrix Equalities

The stabilization problem of continuous-discrete fractional-order 2D Roesser model is investigated in this paper. Firstly, the new stabilization conditions via state feedback controller are proposed based on the necessary and sufficient stability conditions in our previous results. Then, considering the proposed stabilization conditions are not linear matrix equalities, the new method based on Soft Actor-Critic reinforcement learning algorithm and linear matrix equalities solver is established for solving the proposed stabilization conditions and obtaining the stabilization controllers. This method can be used efficiently in solving such non-convex optimization problems. The illustrated example is provided to verify the effectiveness of our results.