A New Model-Free Shared Control for Lane-Keeping Assist System: Theory and Experimental Validation
Hossam Eddine Glida, Mohamed Radjeb Oudainia, Chouki Sentouh, Jean‐Christophe Popieul
Abstract
This work presents a novel model-free approach for lateral lane-keeping assist (LKA) shared control design for steer-by-wire road vehicles. By integrating the Time Delay Estimation (TDE) approach combined with Radial Basis Function Neural Networks (RBFNN), the proposed method addresses limitations inherent in traditional model-based control strategies. The model-free approach allows the control system to handle unknown dynamic functions, estimated using the TDE approach, without relying on a predefined model of the system. The RBFNN is then used to compensate for estimation errors. The optimal selection of the designed parameters has been addressed using metaheuristic optimization method, based on the Gray Wolf Algorithm (GWA). This proposed approach is developed to enhance control performance, particularly in the case of vehicle lateral shared control where accurate modeling of the driver-vehicle dynamics is challenging. Experimental validation using simulations and real-world tests conducted on the SHERPA simulator demonstrates the effectiveness and robustness of the proposed approach. The results indicate improved stability and reliability in LKA systems, showcasing the potential for practical implementation in real-world in other similar systems.