Real-World Testing of LiDAR-Inertial Based Navigation and Mapping for Precision Landing

The fusion of LiDAR and inertial measurements during spacecraft descent and landing can be used to estimate a lander's navigation state and map the terrain below. Together, these data products can be used to enable safe and precise landing on celestial bodies for which a priori orbital reconnaissance is insufficient for hazard detection and avoidance. Unlike camera images used in visual terrain relative navigation, LiDAR scans are insensitive to changes in illumination; as a result, the technique can be used to land in poorly lit areas, or at times of day when the lighting conditions are incongruent with existing orbital imagery. In this paper, we extend previous work in which we introduced a factor graph based smoothing approach for LiDAR-inertial navigation and mapping. Whereas the algorithms were previously tested on simulated data, this paper presents testing on real-world data. Data from the Autonomous Landing Hazard Avoidance Technology (ALHAT) airplane flight tests in the Yucca Flats and Death Valley in 2009 (FT3), the Morpheus vertical take off and landing flight tests at Kennedy Space Center in 2014 (FT6), and the landing of Perseverance and Ingenuity on Mars in 2021 (M2020) were used to evaluate algorithm performance. In this paper, we extend our LiDAR-inertial technique to work with a variety of ranging technologies: single point laser altimetry (FT3), dense flash LiDAR (FT6), and six-beam radar (M2020). A thorough performance analysis for all three datasets is presented. Dataset preparation, improvements in algorithm robustness, and outlier rejection, which were necessitated by the transition to real-world data, are discussed.