Iterative Learning Model Predictive Control for Lateral Control of Autonomous Vehicles

The work proposes a novel iterative learning model predictive control method for lateral tracking control of au-tonomous vehicles. It endows the traditional model predictive controller with the ability of learning from previous experiences. In contrast to the existent iterative learning controllers that mainly concern the control performance in the iteration domain, the proposed control algorithm also takes the dynamical variations of the controlled system along the time axis to expedite the learning speed while ensuring the driving safety. To further improve the robustness of the control system under uncertain driving environments, a control-affine feedforward neural network is incorporated to the proposed controller to deal with the system un-certainties and external disturbances. The convergence of the ILMPC method is rigorously analyzed, and a numerical simulation is illustrated to verify the effectiveness of the proposed method.