The CAFNN-Based Identification Technique for Autonomous Vehicle Dynamics Model

In traditional vehicle modeling processes, physical models often fail to accurately capture complex and nonlinear vehicle behaviors. To address this limitation, this paper proposes a vehicle dynamics modeling method based on Control Affine Feedforward Neural Network (CAFNN) identification. The proposed approach leverages neural networks to extract features from the current vehicle state data and predict the vehicle's state at the next time step. The constructed dataset is used for model training and testing. Findings demonstrate that the vehicle dynamics model identified through CAFNN offers higher modeling accuracy than traditional physical models, particularly under varying road adhesion conditions.