Abstract: In order to solve the problems like poor adaptability in the actual implementation of automatic train operation (ATO) system, and the tight coupling of vehicle traction and braking performance, an intelligent ATO system based on support vector machine (SVM) with training mode and operation mode is proposed. The ATO system directly takes on-board equipment performance parameters, vehicle performance parameters, ATP (automatic train protection)/ATO curve and line parameters as the system input, and uses the machine learning algorithm based on SVM to calculate ATO system traction and braking voltage/current output parameters. The optimal intelligent ATO control algorithm is obtained by training with the actual ATO control curve data and adjusting the SVM parameters. The algorithm can also be used to predict the vehicle control commands under the actual line environment.