Output-feedback robust saturated actor-critic multi-layer neural network controller for multi-body electrically driven tractors with n-trailer guaranteeing prescribed output constraints

This paper proposes a novel robust saturated actor-critic multi-layer neural network controller for electrically-driven tractors with n-trailer with unmeasurable linear and angular velocities, uncertain complex dynamics and actuator saturation while guaranteeing a prescribed performance with em-ploying the motor dynamics. The proposed controller consists of four control loops. In the first loop, tracking errors are transformed into constraint errors via prescribed performance bounds. Then, a kinematic controller is designed. In the second loop, an output feedback robust dynamic controller is proposed via multi-layer actor-critic neural networks to approximate model uncertainties, a high-gain observer (HGO) to estimate velocities, and an adaptive robust controller to compensate external dynamic disturbances. Afterwards, a robust actuator controller is designed in third loop by employing multi-layer actor-critic neural networks to deeply diminish unknown nonlinear functions effects, and an adaptive robust controller to handle the bounded actuator disturbances. An auxiliary subsystem is considered in the final loop to reduce the danger of actuator saturation by designing an auxiliary intermediate controller. The stability under the proposed controller is studied by the Lyapunov stability synthesis, and it is proven that tracking errors remain uniformly ultimately bounded. Finally, the validity, reliability, and effectiveness of the proposed reinforcement learning-based controller is shown by means of multiple simulations and some comparisons with a quantitative study. (c) 2022 Elsevier B.V. All rights reserved.