Developing a parameter adaptive non-linear model predictive control for motion control of semi-trailer trucks.

The widespread use of semi-trailer trucks as a critical mode of freight transportation, owing to their high payload capacity and operational flexibility, necessitates advancements in motion control systems to enhance their safety, efficiency, and adaptability to dynamic road conditions. Model Predictive Control (MPC) is an advanced control strategy extensively employed in the field of motion control for autonomous vehicle guidance. The performance of the MPC method directly depends on how accurately the prediction model within the MPC can replicate real vehicle dynamics. Using a precise parameter set is crucial to achieving a reliable and robust controller, as any parameter mismatch between the controller and the real vehicle could lead to MPC failure. Therefore, for systems that must operate under a wide range of conditions and environments, it is essential to incorporate the ability to adapt working parameters in real time. According to the literature, one widely used method for parameter estimation in complex nonlinear systems is Moving Horizon Estimation (MHE). However, implementing this method requires precision and careful consideration of various factors. This thesis proposes an adaptive Nonlinear Model Predictive Control (NMPC) approach for semi-trailer trucks to enhance the controller’s performance in the presence of parameter mismatches. First, a single-track dynamic model was developed based on the equations of motion for the semi-trailer truck. Subsequently, an adaptive controller integrating Nonlinear Model Predictive Control (NMPC) and Moving Horizon Estimation (MHE) was designed and tested. The results indicate that the proposed controller significantly enhances performance. In open-loop prediction, the Average Relative Distance Error (ARDE) metric improved by 80% to 93%. For closed-loop behavior, the average lateral deviation was reduced by at least 99% and up to 102% compared to a standard MPC controller without adaptivity. Furthermore, MHE proved valuable for state estimation, particularly when certain states were not directly measurable.