To improve the intelligence and safety of the rehabilitation walking training robot, a time-limited learning iterative control method for the rehabilitation training robot based on motion speed decision is proposed to reduce the influence of trainer's pose uncertainty and human-robot velocity incoordination on the safety performance of the system. A dynamic model of the rehabilitation walking training robot with system uncertain offset is established. By comparing the current movement velocity of the rehabilitation training robot and the actual walking velocity of the trainer, a movement velocity decision method for the robot is proposed. Thereby, the trainer can realize the coordination of human-robot movement velocity in the active training mode. Furthermore, a tracking error system is established by using the movement velocity of the robot decision-making and dynamic model, an iterative control method with limited learning time is proposed, and the stability of the tracking error system is verified based on Lyapunov theory. The simulation comparative analysis and experimental results show that the proposed velocity decision method and tracking control method can enable the human-robot system to coordinate rehabilitation training in the active mode.