SUPER ODOMETRY 2.0: Resilient Odometry via Hierarchical Adaptation

Authors/institution: Shibo Zhao and six co-authors. Status: arXiv version posted 26 August 2026; the record references Science Robotics 2025. Problem: autonomous systems lose localisation reliability when cameras or lidar degrade in smoke, snow, dust or low light. Contribution: hierarchical sensor fusion with adaptive feature, state-direction and engine selection, plus learned inertial odometry. Authors report validation over 200 km and 800 operational hours across aerial, wheeled and legged robots.

Practical significance: supports graceful degradation and sensor fallback. Limitations: results are author-reported; operating conditions, comparative baselines and reproduction should be checked in the full paper. The unusual gap between journal reference and arXiv posting should be noted when citing novelty.

Paper