Abstract: The optimisation of high-speed train operation process is a multi-objective, non-linear optimisation problem. In order to study the multi-objective optimisation problem of automatic train operation (ATO), a multi-objective optimisation model for trains is established with on-time train operation, stopping accuracy, comfort and energy consumption as the control objectives. The model incorporates the train dynamics model as the constraint conditions, while considering the train's coasting neutral zone, and an improved biogeography-based optimization (BBO) method for optimizing the ATO speed curve. To improve the optimisation performance of the algorithm, a migration model with hyperbolic tangent variants that is more inclined to natural laws is used. A differential evolutionary (DE) variation strategy is used in the variation process to improve population diversity, while random numbers of the Cauchy distribution are added to help the algorithm jump out of the local optimum. Opposition-based learning (OBL) is used to improve the diversity of individuals after variation to ensure the full domain search of the algorithm. The superiority of the algorithm's convergence speed and global optimisation capability is verified by benchmarking functions. Simulation experiments are carried out with a CRH3 high-speed train and a line of the Hanyi passenger dedicated line, and the results show that the proposed method in this paper can make the train tracking operation more efficient, comfortable, energy-saving and safer. The comfort level is increased by 39.24%, and the energy consumption is reduced by 3.5653%.