Deep Learning-Based Lunar Crater Terrain Relative Navigation

2026-06-09

AI Summary

This study addresses the challenge of achieving high-precision autonomous localization for lunar landers in regions with sparse terrain features. To this end, the authors propose a visual navigation method that integrates deep learning with filtering techniques. Specifically, they combine, for the first time, a deep learning–based crater detector—originally developed for NASA’s Crater Detection Challenge—with Hungarian matching, consistency-based outlier rejection, and an extended Kalman filter (EKF). By fusing elevation-aided information within the Local Level Coordinate Frame (LCLF), the approach effectively mitigates radial drift and enhances robustness. Experimental results demonstrate that the system can converge localization errors to within a few hundred meters, even when starting from an initial position error as large as 5 kilometers, significantly outperforming existing methods.

Abstract

Accurate position estimation is crucial for the successful implementation of future lunar landings using autonomous vehicles, especially in dangerous environments with sparse terrain features. In this paper, we propose a terrain relative navigation (TRN) algorithm combining our deep-learning crater detector, which was designed specifically for the NASA Crater Detection Challenge problem, and an Extended Kalman Filter (EKF). Our detector analyzes crater features from the monocular images acquired from orbit, and their matches with craters from a global database are identified via a Hungarian assignment approach followed by the consensus-based outliers removal method. The estimated measurements are then used to refine an EKF, where spacecraft pose estimation in the Lunar-Centered Lunar-Fixed (LCLF) frame of reference, augmented with altitude aiding information, constrains radial drift. The simulation results indicate that even if the spacecraft is off from its actual location up to 5 km, TRN could recover from this situation, achieving navigation error reduction to a few hundred meters. It should be noted that in order to maintain crater feature correspondences, it is important to match the image resolution and the scales within the scene to the detector training set distribution.

Problem

• Lunar Navigation
• Terrain Relative Navigation
• Crater Detection
• Position Estimation
• Autonomous Landing

Innovation

• Terrain Relative Navigation
• Deep Learning Crater Detection
• Extended Kalman Filter
• Lunar Navigation
• Feature Matching