An adaptive neural network-based controller for car driving simulators

The paper proposes an approach to deal with problems stemming from inaccurate elements of car driving simulators and the disadvantages of nonlinear model-based controllers, including the Backstepping and Sliding Mode Control. For this kind of model, it is challenging to design nonlinear model-based controllers to guarantee smooth and accurate movements due to unmodelled parts in the system dynamics and the appearance of external forces. Furthermore, there are well-known demerits in the aforementioned controls, such as ‘explosion of terms’ and chattering phenomena. The main contribution focuses on constructing an adaptive and robust neural network-based controller with online learning laws, which not only ensures the high accuracy of the robot’s motion under uncertain components and external factors but also mitigate the disadvantages of conventional controllers. The stability and convergence of the proposed method are proven by Lyapunov’s stability. The simulation results show the validity and efficiency of the proposed control algorithm.