Probabilistic Learning-Network-Based Robust Control Scheme for Nonlinear Systems
Jinglu Hu*, Kotaro Hirasawa*, Junichi Murata*, Chunzhi Jin* and Takuya Matsuoka**
*Department of Electrical and Electronic Systems Engineering Graduate School of Information Science and Electrical Engineering Kyushu University
**Engine Engineering Division II, Toyota Motor Corporation
We present a control design scheme for nonlinear systems based on a probability learning network (ProNet). ProNet is a learning network equipped with the capability to deal with stochastic signals. A plant and its controllers are described by using a set of related equations and form a unified learning network-ProNet where disturbances are considered as external inputs. In this way, controller design is transferred to ProNet learning. By including an effort to reduce variances of ProNet output in the criterion function for training, the trained ProNet has different sensitivities to signals of different frequencies. A ProNet control system is designed by taking this advantage to increase its robustness against disturbances. Computer simulations confirm the effectiveness of the ProNet control scheme.