CraterGrader: Autonomous Robotic Terrain Manipulation for Lunar Site Preparation and Earthmoving

TL;DR

CraterGrader presents an autonomous lunar site-preparation system that integrates online perception, GPS-free localization, and an online optimal transport planner to reshape deformable lunar-like terrain. The transport planner extends Earth Mover's Distance concepts into a MILP framework handling unequal source-sink volumes, enabling energy-efficient material movement; planning is tied to a 2.5D perception map and executed via kinematic planning and Stanley-based trajectory control. Demonstrations in a lunar analog (MoonYard) show CraterGrader achieving low-grade and low-smoothness metrics (around 0.1° and sub-centimeter levels) and substantial area-of-spec reduction, with live demonstrations and larger simulations validating scalability and generality. The work offers a practical benchmark for planetary site-preparation robotics, highlighting the importance of external localization, generic environment representations, and online planning for deformable terrain manipulation, while outlining future paths toward multi-robot and multi-modality earthmoving.

Abstract

Establishing lunar infrastructure is paramount to long-term habitation on the Moon. To meet the demand for future lunar infrastructure development, we present CraterGrader, a novel system for autonomous robotic earthmoving tasks within lunar constraints. In contrast to the current approaches to construction autonomy, CraterGrader uses online perception for dynamic mapping of deformable terrain, devises an energy-efficient material movement plan using an optimization-based transport planner, precisely localizes without GPS, and uses integrated drive and tool control to manipulate regolith with unknown and non-constant geotechnical parameters. We demonstrate CraterGrader's ability to achieve unprecedented performance in autonomous smoothing and grading within a lunar-like environment, showing that this framework is capable, robust, and a benchmark for future planetary site preparation robotics.

Figures (10)

• Figure 1: CraterGrader in the MoonYard. Top left: Pseudo-sun fiducial tag used solely for bearing estimation. Center: CraterGrader. Top right: Leica Viva TS16 Total Station.
• Figure 2: An example crater cross-section, depicting key parameters such as grade, smoothness, and positive/negative height, with respect to a fit ground plane. In particular, grade and smoothness metrics are used to evaluate CraterGrader's performance against the LuSTR RFP baseline LUSTR . Grade is defined as the angle between the fit plane and a plane normal to gravity. Smoothness is calculated by constructing a distribution of positive and negative terrain heights relative to the fit plane and computing its standard deviation. The roughness of the topography and grade angle are not to scale and are exaggerated for effect.
• Figure 3: The CraterGrader belly-mounted grading blade assembly. The actuator shown in red is kept inside the robot chassis perimeter to be protected from dust and rock strikes. The blade translates vertically and is rigidly constrained to a linear rail bearing.
• Figure 4: The worksite map is updated as new perception data are acquired. The grid cell color is overloaded in this figure to show both the height of the mapped grid cells and whether a cell has been observed. The dark blue shows map grid cells that have not yet been observed by the robot. Two observed craters are shown alongside the 3D robot model.
• Figure 5: The Kinematic Planner incorporates topography cost to avoid the crater with kinematically feasible solutions for the double Ackermann mobility platform. Green arrows show the starting and goal locations, the green spline shows the resulting trajectory, and the dark blue arcs show path candidates for the expanded A* lattice nodes.