Colorado School of Mines Center for Space Resources

Colorado School of Mines.

A rover that will excavate, drive or drill on the Moon has to be qualified against something, and lunar regolith is not on Earth to test against. The choice a university testbed faces is the one NASA’s own simulant program faced two decades earlier: no terrestrial feedstock reproduces lunar mineralogy, particle shape and strength at once, so a bed has to pick which properties it will match and accept that the rest will be wrong [6][7]. The Colorado School of Mines Center for Space Resources, in Golden, Colorado, runs the university’s answer to that trade: two indoor regolith beds, a set of thermal vacuum and cryogenic chambers at sample and component scale, and the characterization laboratories that produce and grade the simulant the beds are filled with [1][2][3].

Design space

A regolith testbed has to fix four things, and the facilities in the field differ mostly in which they leave out.

Simulant selection. Both Mines beds use a terrestrial basalt or highland-type material chosen for mechanical behavior. The earliest statement of that approach is a Berkeley study built for the Surveyor era, which argued that a simulant should be matched to measured friction angle, cohesion and density rather than to mineralogy, and that a single in-situ density measurement, from footprint depth or a cone penetrometer, would let the rest of the soil’s behavior be estimated [12]. NASA’s 2005 Lunar Regolith Simulant Materials Workshop cataloged Apollo hardware failures it attributed to untested materials, including rover slip and sinkage, tools that could not penetrate past tens of centimeters and seal failures, and proposed a small set of root simulants as compositional end members rather than one universal material [6]. The current NASA reference guide scores simulants on a common basis and still states the limit as a headline conclusion: none reproduces every aspect of lunar regolith and none is best for every purpose [7]. Delivered composition varies by product and by batch. One characterization of the commercial LHS-1 highland simulant reports a median particle size of 59.79 micrometers and a mineralogy of 74.4 weight percent anorthosite, with the authors warning that values change between production batches [13].

Geotechnical truth. What a bed’s soil does under a wheel or blade depends on density and preparation as much as on the material. Across seven lunar and Martian simulants on one direct shear apparatus, cohesion spanned 0.185 to 1.672 kPa over bulk densities of 1.32 to 1.97 g/cm3, so a cohesion value without its density is not usable, and the fits are valid only at normal stresses below 0.68 kPa [14]. A compilation of lunar returned-sample and in-situ data gives cohesion from negligible to about 5 kPa and friction angle of 20 to 55 degrees, mostly from penetrometer and trench data at equatorial and mid-latitude sites [15]. Adding chemical cohesion to the European EAC-1A simulant, an attempt to compensate for Earth gravity, produced a non-monotonic response and was reported as a negative result [16]. Particle shape is a separate gap: averaged shape statistics are dominated by the fines, and Apollo samples measure as less morphologically complex than the simulants compared against them, on different instruments [17].

Environment. Gravity, vacuum, temperature and dust adhesion are what an ambient bed cannot supply. The NASA Glenn Lunar Dust Adhesion Bell Jar was specified around them: a working pressure near 1e-7 torr, a 77 K cold wall, a xenon arc solar simulator and a plasma-activated simulant, because clean-body cohesion appears in simulant below about 5 torr while surface adhesion needs far lower pressure [18]. Vacuum soil mechanics has its own facility literature: at NASA Glenn, interstitial gas escaping a one-metric-ton bed during pump down produced spouting and surface disruption unless the pressure decay rate was controlled [9]. The first vacuum triaxial and plane strain tests of a lunar simulant at lunar confining pressures date to 1970 [12]. Reduced gravity changes the soil as well: cone penetration resistance in the CUMT-1 simulant at one sixth g was 29.1 percent lower than at 1 g at 40 percent relative density but only 8.8 percent lower at 76 percent, so a 1 g bed over- or understates strength depending on how densely it is prepared [20]. Parabolic flight with a single wheel in GRC-1 simulant found that preparing the 1 g bed to the same cone index gradient as the lunar-gravity soil overestimated drawbar pull to weight by 44 percent mean squared percentage error at 20 and 70 percent slip and underestimated sinkage by 13.5 percent [30]. Gravity offload for suited crew has its own artifacts: the ARGOS single-point cable rig leaves the limbs at 1 g and needs a pivot placed 2.5 to 6.4 cm above the system center of gravity, a zone found over 68 configurations on 14 subjects [21].

Dust. Dust is both a test variable and a hazard to the test article. Apollo records show that 11 percent areal coverage doubles the solar absorptance of a thermal control surface and that a suit’s leak rate rose from 0.15 to 0.25 psi/min across two EVAs against a 0.30 psi/min limit [19]. Exposure limits and toxicity work follow from the Apollo samples [23][24], and a Mines-authored review surveys the seal, bearing, coating and barrier hardware and its maturity for lunar use [10].

Mines chose a feedstock at the coarse, mechanically matched end of this space rather than the compositionally matched end, and its director of engineering states the trade plainly: the 2019 bed is not built to the standard NASA and larger government facilities reach, but is a start on it [1]. None of the Mines beds flew anything; they are ground qualification, and status here means what has been run on the ground, not on the Moon.

What changed

The current bed is the Mines Lunar Surface Simulator, about 120 square meters of highland-type simulant inside a dust-tight and waterproof enclosure with an overhead gantry that follows the rover under test, reported operational in August 2025 and in daily use for the NASA-funded ASPECT autonomous site preparation project [2]. It succeeds rather than replaces the original Mines Lunar Testing Facility of 2019, a roughly 6 ft by 12 ft bed of milled Merriam Crater basaltic cinder in the Earth Mechanics Institute, which is still described here as its own capability [1][2][3]. The change enables full-vehicle autonomy testing, not only component or tool testing: a rover can drive, excavate and be tracked over a bed of about 120 square meters rather than about 7 square meters.

Laboratory data

Parameter	Value
Operator	Colorado School of Mines, Center for Space Resources [1][3]
Location	Golden, Colorado, United States [1]
Commissioned	2019, Lunar Testing Facility; August 2025, Lunar Surface Simulator [1][2]
Type	Indoor regolith beds, plus sample-scale thermal vacuum and cryogenic chambers
Floor area	About 120 m2, Lunar Surface Simulator bed [2]. Total not published
Capabilities	Simulator, 2019 bed, Chambers
Simulant or terrain	Highland-type Mines simulant, over 110 t [2]; Merriam Crater basaltic cinder, 20 t [1]
Instrumentation	Rover motion capture; overhead gantry [2]
Ground truth	Rover motion capture, independent of the vehicle. No accuracy published
Fidelity limits	No vacuum, thermal or reduced gravity in either bed [1][2]. See below
Access	Open to startups, industry and academia [2]. No lead time or fee published
Cited by	mapp

Capabilities

Lunar Surface Simulator

Parameter	Value
Working volume	About 120 m2. Bed depth and enclosure height not published
Test article limits	Not published. Rover scale in practice
Vacuum	Not applicable. Ambient pressure
Illumination	Not published. No solar simulator described
Simulant or terrain	Highland-type Mines simulant, over 110 t
Slope	Not published
Gravity offload	Not applicable
Instrumentation	Overhead gantry tracking the rover; rover motion capture

Source: [2].

The enclosure had to be dust-tight and waterproof, and it carries a gantry that follows the rover under test, with a motion capture system built to model rover mobility from observation [2]. Neither the gantry travel and payload nor the motion capture accuracy is published. The operators describe the simulator as in use almost every day.

The simulant selected for the testbed was CSM-LHT-T-30, one of four Mines highland variants characterized by standard ASTM methods. Its triaxial cohesion is about 355 percent above the lunar highland estimate of 3.8 kPa, density, compression index and friction angle differ by less than 30 percent, and matching the lunar average particle size distribution did not make shear strength or compressibility match [28].

The fill figure is internally inconsistent in the only source that gives it: the Mines news release reads “over 110 tons, that is 100 metric tons” [2]. Both numbers are repeated here as published rather than reconciled. Nothing is published about the simulant’s composition, grain size distribution, relative density control or preparation procedure between runs. The NASA reference guide’s occupational hazard is respirable crystalline silica, and other simulant programs report it as a lot-dependent contaminant rather than a fixed property [7][5]; the Mines materials are not characterized in public against that framework.

Lunar Testing Facility, 2019

Parameter	Value
Working volume	About 6 x 12 ft bed [1]. Depth not published
Simulant or terrain	Merriam Crater basaltic cinder, 20 t, milled to 150 to 200 um

The 2019 bed was built around a deliberate choice of feedstock and grading rather than a purchased simulant [1]. Twenty tons of basaltic cinder were bought from the Merriam Crater volcanic ash deposit in Arizona, the same source NASA has used for the BP-1 simulant line [1][5], and milled on campus from 8 mm down to roughly 150 to 200 micrometers, then laid as a fine surface layer over the unmilled coarse fraction, because, as the operator puts it, “the lunar surface isn’t just the fine stuff. There are rock fragments and layering of the subsurface” [1]. A geological audit of the related BP-1 material found that the same Merriam Crater feedstock, drawn as washing waste from a road-aggregate quarry rather than purpose-mined rock, carries 9 to 24 weight percent secondary weathering and contamination phases, including calcite that varies from 9 percent to nil between lots, so the batch a bed is filled with is not chemically identical to the last one [5]. A separate outgassing study of the sibling JSC-1A material traces its own trace carbonate and sulfate contamination to the same class of weathering products [8]. Depressions in the bed stand in for impact craters, and the operators note that a real crater would also contain ejected material and glassy subsurface material, so that building a testbed is about the simulated environment as well as the simulant, and that the bed simulates mare regions rather than highlands [1].

Chambers

Parameter	Value
Temperature	Cryogenic. Range, ramp rate and uniformity not published [1]

The Center holds thermal vacuum and cryogenic chambers alongside the beds. Commercial users have cited access to those chambers, alongside the regolith bed and the faculty expertise around both, as what brought them to Mines rather than elsewhere [1]. No chamber inventory, dimension or base pressure is published. Nothing published states whether the Mines chambers hold regolith beds at the scale where the pump-down disruption described above becomes a problem [9], or are reserved for sample and component testing.

Instrumentation

Published instrumentation is limited to the gantry and the motion capture system in the Lunar Surface Simulator [2]. The program’s other measurement capability sits in separate laboratories rather than in the bed: mineral and elemental characterization of simulant and returned-sample analogues, and the penetrometer work carried out for lunar surface measurement, which was adapted for a robotic arm under a NASA-funded sample acquisition project [1]. Other facilities add measurements the Mines beds do not report: a probe measurement of the dry real permittivity of the NU-LHT-2M highland simulant, 2.75 at 50 MHz over 46 readings, informs radio-frequency work on regolith [22], and component-level qualification of the NASA IPEx excavator actuators reached 1,402,223 accumulated motor revolutions in an accelerated life test, over four times the mission requirement [26]. Neither is a Mines result.

What it does not reproduce

The clearest statement of what the facility does not do comes from its own director of engineering, comparing the 2019 bed with the Kennedy Swamp Works regolith bin: Swamp Works is about ten times the area, fully enclosed and ventilated with air scrubbers, worked in bunny suits and masks, and filled with a much finer regolith that generates more dust in use, in the director’s words, “we’re not there yet, but this is a start” [1]. The Mines material was deliberately coarser and the facility was neither enclosed nor scrubbed at that time, and the material itself is milled terrestrial basaltic cinder matched to composition, mineralogy and particle size distribution rather than to the mechanical history of lunar soil. The Mines-authored dust review makes the general case for why that matters at this scale: dust generation and dust tolerance have to be designed together, because every mechanism that moves regolith both generates dust and is damaged by it, and no single mitigation layer reaches full effectiveness on its own [10].

Simulant quality is a cost problem as much as a knowledge problem, and the operators say so: building a large facility with a high-quality geotechnical lunar regolith simulant “presented logistics challenges to make a large amount of simulant at a reasonable cost” [2].

No published capability exists for vacuum, thermal or reduced gravity inside either bed. The program’s chambers provide those conditions at sample and component scale, but a rover tested in the bed is tested at one atmosphere, room temperature and one g [1][2]. NASA has since put a Break the Ice Challenge rover through thermal vacuum testing of its own, at the Marshall Space Flight Center Environmental Test Facility rather than at Mines, indicating where in the qualification chain that step currently sits for hardware developed on this bed [11].

Outdoor endurance testing is done on a wholly different material. For the NASA Break the Ice Lunar Challenge the site at the Colorado Air and Space Port used low-strength concrete to stand in for the properties of lunar ice, broken up with an impact hammer before excavation [4].

A NASA consolidation of rover mobility test practice adds two limits that apply to any 1 g bed, Mines included: matching lunar soil strength by cone index gradient is of uncertain validity under lunar gravity, and no known facility reproduces the inertial and dynamic effects of a vehicle specified to 15 km/h, such as the Lunar Terrain Vehicle [29]. Beds like these therefore support quasi-static mobility and excavation work, not fast-vehicle qualification.

Flight heritage is the other absence. A maintainability review tabulating deployed mass, traverse distance and failure mode across Lunokhod, Apollo, Yutu and the Mars rovers is the reference against which ground endurance campaigns like the 15-day run below are judged [27]. The NASA IPEx excavator, designed for up to 10,000 kg of regolith per lunar day, is the same class of machine and has also been qualified on the ground rather than flown [25].

Campaigns run there

Lunar Outpost MAPP development. Lunar Outpost used the 2019 test bed during development of its MAPP rover; the company’s chief executive identified the quality of the regolith simulant, plus access to thermal vacuum and cryogenic chambers, as what the facility offered that others did not [1]. See mapp.

ASPECT, Autonomous Site Preparation: Excavation, Compaction and Testing. Selected under NASA’s Lunar Surface Technology Research program as one of three university-led projects, led by Mines with Lunar Outpost, Michigan Technological University and Bechtel, with Lunar Outpost supplying the rover mobility element derived from its MAPP and HOUND platforms. The goal is autonomous preparation of a landing site by relocating rocks, moving regolith, and leveling, grading and compacting the surface, culminating in a terrestrial demonstration of autonomous landing pad construction on the Mines campus [3]. Rover investigations for ASPECT were the facility’s main occupant when the Lunar Surface Simulator was reported operational [2].

Outpost Digger System endurance run, September 2023. For the NASA Break the Ice Lunar Challenge, Mines and Lunar Outpost ran a two-rover excavation system for 15 consecutive days from 9 September 2023 at the Colorado Air and Space Port, against a requirement to excavate, transport and dump at least 12 metric tons of concrete-hard simulant [4]. Each rover was planned to excavate five hours a day and drive 38 km over the competition. Mines optimized the excavation implements and Lunar Outpost built the rover platform, based on its Hound platform at more than ten times the size of MAPP, with a high-capacity battery and rapid charger sized for the impact hammer, and airless tires [4].

Commercial rover evaluation, 2025. Neurospace GmbH of Berlin evaluated its modular HiveR rover platform in the Lunar Surface Simulator, working towards the minimum technology needed for an inexpensive, scalable, self-repairing rover [2].

What is not established

No dimension, base pressure, temperature range, ramp rate or uniformity figure is published for the thermal vacuum and cryogenic chambers, and it is not established whether they can hold a regolith bed at rover scale or only samples and components [1]. The Lunar Surface Simulator’s fill quantity is given only as an internally inconsistent figure in a single news release, and neither its grain size distribution, mineralogy nor batch-to-batch consistency is published [2]; the geological record for the related Merriam Crater feedstock used in the 2019 bed shows that consistency cannot be assumed across lots [5]. Neither bed offers vacuum, thermal control or reduced gravity, so no published Mines test establishes how a rover characterized here would perform once those conditions are added. Because cohesion and friction angle vary with density and preparation [14][20], and no Mines preparation procedure or density measurement is published, the bed’s strength cannot be compared with the lunar values compiled elsewhere [15].