Lunar rover slip estimation improved by fusing expert models

Fleet-To-Lab: A Transfer Learning Framework For Lunar Rover Slippage Estimation Via Model Fusion

Robotics

Summary

Estimating how much a lunar rover’s wheels slip is important for safe movement, but models trained on Earth don't work well on the Moon because the terrains are very different. The authors created a system called Fleet-to-Lab that uses data from previous lunar rovers to improve slip prediction for new rovers, even with limited new data. They combine multiple expert models into one using a special algorithm, AcoMerge, which smartly finds the best way to mix them. Their tests in a detailed simulation showed better and more balanced predictions compared to other methods. This approach may help future space missions deal with small datasets when estimating rover slippage.

What this means in practice

  • For space robotics teams: Improve slippage prediction for lunar rovers deployed on the Moon with limited new data by fusing known expert models.
  • For autonomous vehicle engineers: Incorporate model fusion techniques to enhance slip estimation in robots or vehicles transitioning between different terrains with scarce new data.

Tested on simulated data.

Authors

Riccardo Viviano, Saki Omi, Andrej Orsula, Miguel Olivares-Mendez

Abstract

Accurate wheel slip estimation is essential for autonomous lunar rover mobility and navigation. Machine Learning models trained on terrestrial data generalize poorly to lunar terrain, and real lunar datasets are scarce due to the limited number of missions and costly data acquisition. We present Fleet-to-Lab, a transfer learning framework that leverages proprioceptive data collected by previously deployed heterogeneous lunar rovers to mitigate the Earth-Moon domain gap in slip estimation for a future deployable unit. We fuse several heterogeneous expert models into a single architecture, using a modest dataset collected after the rover deployment. We propose AcoMerge, a new hybrid swarm-intelligence algorithm that performs model fusion by searching for an optimal combi- nation of expert parameters. Experiments conducted in a high- fidelity physics simulation show balanced accuracy and macro- F1 improvements compared to deep model fusion baselines. AcoMerge exhibits competitive performance with joint training on deep architectures, while achieving higher macro-F1 and balanced accuracy on a smaller model. Overall, our framework shows model fusion as a possible transfer learning alternative for slippage estimation in space robotic missions with limited data.
View PDFOpen arXiv