Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Full Stack Navigation, Mapping, and Planning for the Lunar Autonomy Challenge

Adam Dai, Asta Wu, Keidai Iiyama, Guillem Casadesus Vila, Kaila Coimbra, Thomas Deng, Grace Gao

Abstract

We present a modular, full-stack autonomy system for lunar surface navigation and mapping developed for the Lunar Autonomy Challenge. Operating in a GNSS-denied, visually challenging environment, our pipeline integrates semantic segmentation, stereo visual odometry, pose graph SLAM with loop closures, and layered planning and control. We leverage lightweight learning-based perception models for real-time segmentation and feature tracking and use a factor-graph backend to maintain globally consistent localization. High-level waypoint planning is designed to promote mapping coverage while encouraging frequent loop closures, and local motion planning uses arc sampling with geometric obstacle checks for efficient, reactive control. We evaluate our approach in the competition's high-fidelity lunar simulator, demonstrating centimeter-level localization accuracy, high-fidelity map generation, and strong repeatability across random seeds and rock distributions. Our solution achieved first place in the final competition evaluation.

Full Stack Navigation, Mapping, and Planning for the Lunar Autonomy Challenge

Abstract

We present a modular, full-stack autonomy system for lunar surface navigation and mapping developed for the Lunar Autonomy Challenge. Operating in a GNSS-denied, visually challenging environment, our pipeline integrates semantic segmentation, stereo visual odometry, pose graph SLAM with loop closures, and layered planning and control. We leverage lightweight learning-based perception models for real-time segmentation and feature tracking and use a factor-graph backend to maintain globally consistent localization. High-level waypoint planning is designed to promote mapping coverage while encouraging frequent loop closures, and local motion planning uses arc sampling with geometric obstacle checks for efficient, reactive control. We evaluate our approach in the competition's high-fidelity lunar simulator, demonstrating centimeter-level localization accuracy, high-fidelity map generation, and strong repeatability across random seeds and rock distributions. Our solution achieved first place in the final competition evaluation.
Paper Structure (28 sections, 10 equations, 18 figures, 3 tables, 1 algorithm)

This paper contains 28 sections, 10 equations, 18 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Lunar Autonomy Challenge
  3. Lunar Robot
  4. Simulator and Environment
  5. Objectives and Scoring
  6. Related Work
  7. Full Autonomy Stacks in Challenging Environments
  8. Visual SLAM
  9. Learned Feature Matching
  10. Approach
  11. Semantic Segmentation
  12. Feature Extraction and Matching
  13. Stereo Visual Odometry and Tracking
  14. Simultaneous Localization and Mapping (SLAM)
  15. Frontend
  16. ...and 13 more sections

Figures (18)

  • Figure 1: View from the simulator showing the lunar robot with the lander in the background.
  • Figure 2: Diagram of the lunar robot showing the arrangement of its eight cameras.
  • Figure 3: Mapping area for the challenge. The objective is to map both terrain height as well as rock presence within a 27 m $\times$ 27 m area around the lunar lander.
  • Figure 5: Block diagram of our overall approach. Stereo images are processed by the front-end for semantic segmentation, rock detection, and stereo visual odometry. Detected rocks are passed to the motion planner, while tracked 3D features and odometry estimates are used in the SLAM backend for pose graph construction and loop closure. Optimized poses and semantic landmarks are projected into a global frame to generate geometric and rock maps. High-level path design generates waypoints that encourage loop closures and full map coverage, while the motion planner selects safe arcs using geometric rock avoidance.
  • Figure 6: Semantic segmentation example. Given input image, each pixel is labeled with a semantic class of sky, ground, rock, lander, or fiducials.
  • ...and 13 more figures