To improve the path tracking accuracy and vehicle stability of virtual rail trains at different speeds, an active steering control algorithm is proposed based on the virtual rail train dynamics model and model predictive control theory, which adaptively adjusts the weight coefficients of the cost function according to the vehicle speed. Firstly, a dynamic model of a three-car virtual track train is established. To improve the accuracy of the model, genetic algorithm is used to identify the parameters of tire lateral stiffness. Secondly, shift registers are used to store the trajectory of the first vehicle as the target path for the following vehicle. Then, based on model predictive control theory, the steering angles of the first, fourth, and sixth axes are determined, and based on Stanley algorithm, the steering angles of the second, third, and fifth axes are determined. A multi-objective controller considering vehicle lateral error, center of mass lateral angle, and wheel angle variation is designed. Next, to address the issue of poor adaptability of multi-objective controllers with fixed weight coefficients at different vehicle speeds, a fuzzy adjustment strategy based on dynamic optimization of weight coefficients at vehicle speeds is proposed. Finally, the TruckSim/Simulink simulation platform is built to study the dynamic characteristics of vehicles under typical road conditions by setting different speeds.The results indicate that the adaptive weight controller can effectively improve the path tracking accuracy and vehicle stability of virtual rail trains at different speeds. Compared to fixed weight controllers, adaptive weight controllers can better adapt to changes in the operating speed of virtual rail trains and improve their operational performance.