The invention discloses an online vehicle quality estimation method based on model iteration, aiming to solve the shortcomings of low estimation accuracy, poor robustness and few applicable conditions in the existing vehicle quality estimation technology. By obtaining engine torque, longitudinal vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, and steering wheel angle from the vehicle CAN bus and sensors, combined with recursive least squares and Kalman filter algorithms, using longitudinal model quality estimation methods or lateral The model quality estimation method estimates the vehicle quality according to the model arbitration method under different working conditions, and iteratively updates each other. The present invention refers to many vehicle-mounted signals, considers the longitudinal and lateral aspects, achieves the effect of mutual correction, and has the advantages of strong robustness, wide range of applicable conditions and high precision. It provides a relatively reliable vehicle quality signal input for the electronic control system of the automobile chassis.