<p>During the operation of high-speed train, the train basic resistance parameters (TBRPs) are the core parameters influencing train dynamic performance, typically characterized by a quadratic polynomial model. However, as TBRPs cannot be directly measured, existing methods rely on extensive repetitive experiments to obtain empirical parameters. This leads to deviations between the controller model parameters and the actual physical system, constraining the effectiveness of model-based control strategies. To address this challenge, this paper proposes an online asymptotic estimation method for basic resistance based on an adaptive observer. Firstly, based on Newton’s laws of motion and aerodynamics principles, a longitudinal train dynamics model considering the coupling characteristics between adjacent cars is constructed. Building upon this model, an observer architecture incorporating parameter adaptive laws is designed. By utilizing online measurements of train’s position, velocity and traction/braking forces, the observer dynamically updates the TBRP estimates. Furthermore, under appropriate persistent excitation assumptions, rigorous stability analysis of the estimation error system is conducted by constructing a Lyapunov function. It is proven that the estimation error achieves asymptotic convergence in the absence of external disturbances, while the estimation error remains uniformly ultimately bounded in the presence of disturbances. To validate the effectiveness of the proposed method, a simulation platform is established using CRH3 train parameters, and comparative experiments are conducted against a modified intermediate observer. Simulation results confirm the validity and superiority of the proposed estimation method.</p>