A Simulated Multi-Sensor SLAM Dataset for Robotic Exploration in Planetary-Like Environments

SLAM is a key technology for robotic planetary exploration. Harsh conditions in planetary-like environments pose significant challenges to existing methods. Direct experiments in environments such as the Moon or Mars are difficult and costly. To address this, we collected a simulated SLAM dataset in a Lunar-like environment, including camera, LiDAR, and IMU data. The dataset features sparse visual landmarks, limited geometric structures, blurred imagery, and self-similarity, representing degraded scenarios that challenge current vision-based and LiDAR-based SLAM methods. It contains 9 sequences with a total trajectory length of 9 km, enabling targeted evaluation of different SLAM modules and tasks. We evaluate several state-of-the-art SLAM methods on this dataset and report their challenges and limitations, thereby helping to provide insights for future improvements of these methods.