A novel policy iteration technique for the optimal control of partially unknown continuous bilinear systems is presented in this work. We aim to find optimal controllers for two different kinds of problems: Optimal state feedback and optimal output feedback controller design. It is assumed that the matrices related to system dynamics are unknown (matrices A and B) for both cases. In the first scenario, a new optimal policy iteration is suggested based on observed states and inputs to find an optimal state feedback controller. Furthermore, based on the development of case one\\\\\\\'s results, a new optimal policy iteration is provided for case two, where the system\\\\\\\'s outputs are only ones accessible, to obtain the optimal output feedback controller. In this case, the states that only the outputs are used are also assumed to be known. It has been proved that the suggested algorithms converge to the locally optimal controllers.

Keywords