Optimal design of low computational burden model predictive control based on SSDA towards autonomous vehicle under vision dynamics

Abstract

The adjusting of autonomous vehicles (AVs) steering angle represents a challenging issue in the intelligent vehicular applications. Among different control strategies, the model predictive control (MPC) is devoted to providing efficacious steering control for the AVs. However, the formulation of MPC with a few parameters is the main challenge to overcome the complicated computation that happens due to the long prediction and control horizons. Hybridization between discrete-time Laguerre function (DTLF) and MPC is suggested to formulate the MPC with few parameters. While the hybrid DTLF–MPC requires an adequate tuning technique for their gains to improve the system performance. A new low computational burden design is proposed in this paper to adjust the gains of the hybrid DTLF–MPC based on the social ski driver algorithm (SSDA). This algorithm is developed as a recently artificial intelligence technique. The tuned hybrid DTLF–MPC by SSDA is devoted to control AV's steering angle. Furthermore, the proposed system takes into account the vision dynamics. The system performance based on the tuned DTLF–MPC by SSDA is evaluated with the MPC based on the neural network algorithm for comparison. Different scenarios are carried out to confirm the superiority of the suggested control mechanism to handle small as well as large-scale variations of road curvatures.