Online Policy Learning-Based Output-Feedback Optimal Control of Continuous-Time Systems
Jun Zhao, Yongfeng Lv, Qingliang Zeng, Lirong Wan
Abstract
Although state-feedback optimal control of the continuous-time (CT) systems has been extensively studied, resolving optimal control online via output-feedback is still challenging, especially only input-output information can be used. In this brief, we develop an innovative technique to online design the output-feedback optimal control (OFOC) of the CT systems. Firstly, to synthesis the OFOC, an output-feedback algebraic Riccati equation (OARE) is constructed, which can be solved using input-output information. Then, an online policy learning (PL) algorithm is developed to compute the solution of the OARE, where only the input-output information is required and the conventional offline learning procedure is avoided. Simulations based on an aircraft model are provided to test the developed control method and online learning algorithm.