Abstract: Accurate acquisition of vehicle state parameters is an important prerequisite for active safety control. However, the vehicle center of mass lateral deflection angle, as a key stability parameter for active safety control, is difficult to be measured directly. To address these challenges, this paper proposes a hybrid state estimation algorithm based on neural network and untraceable Kalman filter (UKF). First, a composite of neural network combining convolutional neural network (CNN) and bi-directional long and short-term memory network (BiLSTM) is designed to estimate the center-of-mass lateral deflection angle. Second, the dataset is constructed by different maneuvering conditions simulated by TruckSim. Then, the UKF estimator is built based on the three-degree-of-freedom vehicle dynamics model and the Magic Tire formulation, and the estimated values of CNN-BiLSTM are inserted into the UKF estimator as virtual observations. Finally, it is verified by simulation experiments, and the comparison results show that the proposed hybrid scheme outperforms the separate CNN-BiLSTM estimation and UKF estimation, and improves the accuracy and robustness of the center-of-mass lateral deflection estimation.