Simultaneous localization and mapping (SLAM) is a navigation technology used in scenarios where the surrounding environment is unknown. Although SLAM technology is highly advanced in atmospheric environments, it is not highly effective in underwater environments because of various constraints. In addition, experiments in underwater environments involve higher risks and costs compared with other environments. Therefore, in this paper, a simulator-based data collection method was proposed to reduce risks and costs for effective experimentation. By using the proposed method, sensor data can be acquired by adding and generating paths to control the movement of underwater robots depending on research purposes. In addition, collected data can be saved in various formats to facilitate data processing. Moreover, an experiment was conducted to verify that SLAM can be performed using the data collected.