Robust LPV design with neural network for the steering control of autonomous vehicles

The paper proposes a robust Linear Parameter-Varying (LPV) control, in which the results of the approximation of neural network is built. The motivation of the paper is to replace the neural network feedback control with an LPV control architecture, in which some signal values are set by the neural network. Thus, the stability of the closed-loop system through the robust LPV design is guaranteed, while the performances of the neural network are preserved. In the method the output signal of the neural network is approximated by a formulation, which incorporates the scheduling variable and an additional input disturbance in the output of the LPV control. The selections of the scheduling variable and disturbance values lead to an iterative design of the LPV control. The proposed method is applied to the design problem of the autonomous steering control, whose efficiency is presented through simulation examples.