Enhancing Lateral Control of Autonomous Vehicles through Adaptive Model Predictive Control

The use of Model Predictive Control (MPC) for autonomous vehicle control and path tracking is a widely adopted technique. However, the implementation of MPC involves approximating the nonlinear kinematics model of the vehicle into a linear form and transforming the path tracking problem into a suitable format for convex quadratic programming solvers. This study focuses on investigating the application of Adaptive Model Predictive Control (AMPC) specifically for lane tracking in autonomous driving scenarios. A 3 Degree-of-Freedom bicycle model is employed, which incorporates dynamic adjustments through sensor fusion. A key aspect of this research involves systematically tuning control weights and making real-time adjustments to demonstrate the effectiveness of AMPC in various driving situations. The obtained results indicate that the implemented Adaptive Model Predictive Control (AMPC) significantly improves tracking accuracy, particularly in dealing with changes in longitudinal velocities. Furthermore, the AMPC demonstrates its capability to effectively handle varying working conditions and successfully reject external disturbances. To enhance the practicality of this approach, the study utilizes the Unreal Engine 3D simulation platform, highlighting the potential for real-time implementation of AMPC in autonomous driving and lane following. The continuous advancements in control algorithms and sensor technologies are expected to contribute to the ongoing enhancement of AMPC performance.