【Objective】To address the challenge of achieving high-precision trajectory tracking control for heavy haul trains under complex track conditions, this paper proposes a multi-mass model for heavy haul trains and a radial basis function neural network sliding mode control (RBFNNSMC) method. 【Method】First, considering the constraints of air brakes and coupler devices, a multi-mass model of the heavy haul trains was established, and the model uncertainty problems caused by human measurement errors and vehicle parameter differences were estimated by using RBFNN. Second, a nonlinear disturbance observer (NDO) was designed to be utilized for re al-time estimation of strong wind, rain, snow, and other external fast time-varying disturbances during the opera tion of trains. Then, a Lyapunov function was designed to prove the stability of the entire system. 【Result】Based on actual track data from the Daqin Railway, speed tracking comparison experiments were conducted using the RBFNNSMC method, PID method, and SMC method. Simulation results show that the speed error of the RBF NNSMC method is within ±0.15 km/h, which is superior to the other two methods. Furthermore, the inclusion of the NDO significantly enhances the RBFNNSMC method's disturbance rejection capability. 【Conclusion】The tracking accuracy of the RBFNNSMC method based on NDO is improved by 27.3% and 28.9% respectively compared to the SMC method in the absence and presence of disturbances, with enhanced robustness as well.