Adaptive super-twisting sliding mode steering control of autonomous ground vehicles under critical driving conditions: an integrated model-based design validation approach

Abstract

Autonomous driving systems convert environmental perception into vehicle motion. This process relies on a layered architecture composed of perception, planning and control. Among these layers, the control is considered the primordial part because of its commands that are responsible for path tracking and ensuring stability. The complex behavior and unmodeled dynamics of the lateral vehicle lead to mismatches between the mathematical model and the real plant, which affect the control strategy especially at critical conditions such as high speed and physical limits. Accordingly, this research presents a lateral control method that minimizes lateral path tracking error while maintaining the yaw stability of the autonomous ground vehicle under changing road conditions. Primarily, the lateral deviation calculated based on the single-track model and the reference road model is suppressed by robust path-following controller that developed via super-twisting hierarchical sliding mode control to maintain stability of the vehicle. Next, an adaptive mechanism based on a radial basis function neural network is designed to approximate the whole dynamic modeling and unknown disturbances. For the purpose of reducing the uncertainties and tolerate external disturbances using online learning. However, the parameters of the proposed controller are optimized through a neural network optimization algorithm to enhance the effectiveness of the controller. With the aim of validating the proposed method, the lateral tracking controller simulations were performed according to the Model-Based Design approach, including Model in the Loop, Software in the Loop, and Processor in the Loop tests. The results of these simulations show effective path tracking, confirming the relevance and feasibility of implementation in the real environment of the developed control strategy.