Training source matters more than depth features for lunar terrain safety
Evaluating Depth-Augmented Visual Features for Cross-Dataset Lunar Traversability Estimation
Computer Vision and Pattern RecognitionRobotics
Summary
Assessing if a robot can safely move on the moon is crucial but hard due to limited real moon data. The authors tested if adding 3D shape info to camera images helps predict safe paths better than just using images alone. They found that simply adding depth details doesn't reliably improve predictions compared to using visual data by itself. The bigger impact came from training the models on similar environments to the lunar surface rather than relying on generic data. This insight can help design smarter, practical lunar robots that run with limited computer power.
What this means in practice
- •For space robotics engineers: Choose training data from lunar-like environments to improve robot terrain safety predictions over adding geometric depth features.
- •For robotics software developers: Develop perception models that rely primarily on visual data for terrain assessment on resource-limited lunar exploration devices.
Authors
Jan Baturo, Bartosz Ptak
Abstract
Autonomous mobility is essential for sustained lunar surface operations, as communication delays and line-of-sight occlusions limit continuous teleoperation. Although recent traversability estimation methods increasingly rely on deep learning, the domain gap between terrestrial analogue environments and real lunar terrain can limit their generalisation, and real lunar traversability labels are scarce, making fully supervised training difficult. To address these challenges, we build a weakly supervised framework trained on sparse pairwise annotations and study whether augmenting visual features with depth-derived geometric information: depth, surface normals, and combined geometric descriptors. We evaluate the models in both in-domain and cross-domain settings using two terrestrial analogue datasets and the real lunar Chang'E PCAM dataset. Contrary to expectations, we find that naive geometric augmentation yields no consistent improvement over a strong visual-only baseline in either setting, with effects small relative to seed-to-seed variance. Instead, the choice of training source emerges as the dominant factor in transfer quality: models trained on domain-relevant sources transfer substantially better to real lunar imagery than the geometric fusion strategy improves any single model. Our evaluation runs onboard a resource-constrained edge device and provides a controlled baseline for designing more expressive multimodal fusion architectures in future lunar perception systems.