Improved grey wolf optimizer based on opposition and quasi learning approaches for optimization: case study autonomous vehicle including vision system

摘要

The adapting of lateral deviation during the change of road curvature with less error, system settling time, and overshoot is the main challenge against the steering angle control of autonomous vehicles (AVs). In this regard, this paper introduces new learning techniques defined opposition-based learning (OBL) and quasi OBL (QOBL) to improve the exploration as well as exploitation manner of the grey wolf optimizer (GWO). The involved approach can enhance the searching behavior of the original GWO against the trapping in local optima. The proposed modified GWO (MGWO) is applied to detect the optimal factors of the adaptive model predictive control (AMPC) for AVs. The suggested MGWO-based AMPC is evaluated with the classical MPC and the adaptive fuzzy logic controller. Furthermore, the inspired MGWO is compared with the original GWO, neural network algorithm (NNA), heap-based optimizer, and equilibrium optimizer in literature. The performance of the introduced technique is tested to follow different road curvatures. Moreover, the presented method is approved against the time delay of the vision system and the produced uncertainty of system variables from the change of vehicle speed and look–ahead distance. Furthermore, the introduced MGWO-based AMPC can tackle the system settling time and overshoot to be less than 1 s and 1.696% respectively for the response of lateral deviation compared to other techniques. The attained results emphasize that the proposed MGWO-based AMPC controller has high damped and effective performance evaluated with other controllers.