A novel robust model-free adaptive control (R-DMFAC) algorithm is proposed to address the path tracking control problem of autonomous vehicles in the presence of external measurement disturbances. First, the preview-deviation-yaw angle based tracking method is proposed, which transforms the path tracking problem into the preview-deviation-yaw angle control problem. Second, a novel dynamic linearization technique is employed to convert the nonlinear dynamical model, based on preview-deviation-yaw angle, into a linear data model with pseudo partial derivative (PPD), and the proposed algorithm (PFDL-EMFAC) is designed based on this data model. Furthermore, a measurement disturbance suppression scheme is designed by introducing the decreasing factor. Notably, implementing the algorithm does not involve any model information; it is a purely data-driven control algorithm. Finally, the joint simulation results of MATLAB-Panosim platform demonstrate that the maximum tracking error of the autonomous vehicle controlled by the R-DMFAC in different scenarios can be reduced to 0.5-0.7 m, verifying the effectiveness of the control algorithm.

Addressing the lateral path tracking control issue of autonomous vehicles during data dropout, an improved model-free adaptive control system with data compensation (DC-EMFAC) is introduced. First, the method introduces a dynamic linearization technique with a time-varying factor pseudo gradient (PG) to linearize the dynamic process of an autonomous vehicle, and then designs a model-free adaptive controller. Moreover, addressing the issue of data dropout in the actual system, this paper employs an estimation algorithm to estimate the data loss at the present time based on the system 's input and output (I/O) from the past and PG. The advantage of the DC-EMFAC is that the controller design process is based on the I/O data of the controlled object, without the need for an accurate mathematical model. The effectiveness of the proposed algorithm is verified through a series of simulations on the Panosim platform.