Structure-Invariant Range-Visual-Inertial Odometry

TL;DR

This work addresses the challenge of accurate navigation for the Mars Science Helicopter during Mid-Air Helicopter Delivery by removing ground-planarity constraints in range-visual-inertial odometry. It extends the xVIO framework with online 1D-LRF range features that initialize precise depth and propagate scale through the EKF, enabling terrain-relative velocity estimation on non-flat terrain. Through a dedicated Mars MAHD simulation environment and a comprehensive Monte Carlo analysis, the approach shows improved scale observability and robustness compared to prior range-VIO methods, while highlighting the failure modes of planar-range models under large elevation changes. The results demonstrate the practical impact of terrain-invariant range-VIO for safe and accurate EDL navigation on diverse Martian landscapes, with potential applicability to future Mars helicopter missions and similar robotics scenarios.

Abstract

The Mars Science Helicopter (MSH) mission aims to deploy the next generation of unmanned helicopters on Mars, targeting landing sites in highly irregular terrain such as Valles Marineris, the largest canyons in the Solar system with elevation variances of up to 8000 meters. Unlike its predecessor, the Mars 2020 mission, which relied on a state estimation system assuming planar terrain, MSH requires a novel approach due to the complex topography of the landing site. This work introduces a novel range-visual-inertial odometry system tailored for the unique challenges of the MSH mission. Our system extends the state-of-the-art xVIO framework by fusing consistent range information with visual and inertial measurements, preventing metric scale drift in the absence of visual-inertial excitation (mono camera and constant velocity descent), and enabling landing on any terrain structure, without requiring any planar terrain assumption. Through extensive testing in image-based simulations using actual terrain structure and textures collected in Mars orbit, we demonstrate that our range-VIO approach estimates terrain-relative velocity meeting the stringent mission requirements, and outperforming existing methods.

Figures (10)

• Figure 1: Schematic representation of the proposed method. The red star corresponds to a range-feature whose depth is initialized with the 1D-LRF measurement and standard deviation $\sigma_{\rho_{LRF}}$.
• Figure 2: Schematic representation of the local planarity terrain assumption violation in the range-facet model in xVIO.
• Figure 3: Elevation map of Valles Marineris from NASA-JPL:valles_marineris_elevation. The elevation heat map shows that some areas in the site present abrupt changes in elevation, going from $-4000$ to $+4000$ m measured from the Mars Orbiter Laser Altimeter (MOLA).
• Figure 4: Block representation of the feature state covariance matrix. Through their cross-correlation block with the single ranged-feature, the depth estimates of the unranged feature converges to the correct metric depth under a large-enough translation motion. This is equivalent to a triangulation.
• Figure 5: The upper image shows a lateral view of a site in Valles Marineris, generated from a HiRISE DTM rendered in the representative simulation environment. The image below shows a camera image view on the model while running xVIO. Green dots correspond to SLAM features, while red dots to ranged-features.