NASA ARC Roverscape

NASA/Don Richey. Public domain (NASA / US government work).

A rover’s navigation and hazard-detection software has to be exercised on ground that looks like the terrain it will fly over before anyone trusts it on a mission, and orbit, thermal vacuum and radiation chambers do nothing for that: none of them puts a wheel on a slope, a boulder field or a shadow. NASA ARC answers that gap with an outdoor lot rather than an indoor rig. The Roverscape is a two-acre outdoor lunar analogue terrain field, designed to support testing of rover locomotion, navigation and operations, and also suitable for simulating planetary surface missions [4]. Its free surface is pea gravel rather than a regolith simulant, laid out with boulder distributions and an average surface albedo of 8 percent chosen to resemble lunar regolith [2]. It is the terrain the Intelligent Robotics Group drives on, and it is where the K10 planetary rover was operated from the International Space Station in 2013 [1][3].

A 2025 NASA state-of-practice survey gives the field’s area as 11,500 m2, larger than the “two acres” (about 8,100 m2) the site’s own operators state in the papers describing the 2013 and 2017 campaigns [4]; nothing published reconciles the two figures.

The Arc Jet Complex shares the center but nothing else, and has its own page: it is run by a different branch, in different buildings, for a different purpose.

Laboratory data

Parameter	Value
Operator	NASA ARC, Intelligent Robotics Group [1]
Location	Moffett Field, California, United States
Commissioned	Not published; in use for Surface Telerobotics in summer 2013
Type	Outdoor analogue terrain field [4]
Floor area	Two acres, approximately 8,100 m2; 11,500 m2 per a later NASA survey [2][4]
Capabilities	Terrain field
Simulant or terrain	Pea gravel free surface, boulder distributions, 8 percent albedo
Instrumentation	Not published as installed capability; users bring their own metrology
Ground truth	Site imagery at 0.75 m/pixel and a 1.5 m/post elevation model
Fidelity limits	1 g, Earth atmosphere, ambient daylight, pea gravel not simulant [2]
Access	Not published; no external user route, lead time or fee schedule located
Cited by	VIPER, photographed running its Moon Gravity Representative Unit test platform on the surface

Orbital-resolution imagery of the site at 0.75 m/pixel and a digital elevation map at 1.5 m/post were produced for mission planning, both deliberately degraded to match what is available for the Moon, which gives the site a ground truth layer that is independent of any rover driving on it: surface-level survey data can be compared against what orbital resolution would have shown [1]. Campaign metrology is otherwise brought by the user, for example the multi-dot laser projector and camera mounted on the leading edge of K-REX2 for the virtual bumper work [2].

Capabilities

Terrain field

Parameter	Value
Working volume	Two acres, approximately 8,100 m2; no layout plan published
Test article limits	Full-scale rovers K10 and K-REX2 run here; VIPER’s MGRU has been photographed on the surface [1][2]
Vacuum	Not applicable; outdoors at Earth ambient pressure
Temperature	Not controlled; Moffett Field outdoor conditions
Illumination	Ambient; night running used for directional lighting cases
Simulant or terrain	Pea gravel, boulder distributions, 8 percent average albedo
Slope	Not published; no slope inventory or settable slope rig described
Gravity offload	Not applicable; weight-equivalent test platforms used instead

Everything published about the field’s physical construction fits in two sentences of a computer vision paper: it is a two-acre outdoor planetary analog terrain with boulder distributions and an average surface albedo of 8 percent similar to lunar regolith, whose free surface is covered in a layer of pea gravel and not a regolith simulant, for reasons of practicality [2]. Obstacle test cases have been designed to mirror the geometric possibilities likely in craters at the lunar poles. No tonnage, grain size distribution, bearing capacity or preparation procedure between tests is published.

Illumination is not an engineered capability but it is used as a variable. Day and night tests have been run to recreate direct and indirect illumination conditions, the two chosen so that total incident illumination and dominant directionality were similar, which is how a structured-light hazard detector was evaluated on the site. There is no solar simulator and no settable solar elevation angle.

Lunar weight is approximated elsewhere by building a weight-equivalent test platform rather than by offloading; VIPER’s Moon Gravity Representative Unit is such a platform, its mass tuned to give Earth-gravity wheel loads matching the flight rover’s lunar wheel loads, and it has been photographed running on the Roverscape surface. The MGRU’s mobility and terramechanics test program itself runs at a different NASA test bed and is not described here.

Instrumentation

The site has no published installed metrology. What exists as facility infrastructure is the prior survey: satellite imagery of the site degraded to 0.75 m/pixel and a 1.5 m/post digital elevation model, both chosen to match the orbital data resolution available for the Moon, plus a control shelter from which operations are run [1]. Everything measured about a test article is brought by the campaign.

What it does not reproduce

Gravity and atmosphere. The field is at 1 g under Earth atmosphere and ambient daylight, and there is no offload rig [1]. Lunar weight is approximated instead by building a weight-equivalent test platform, the approach behind the VIPER Moon Gravity Representative Unit that has been driven here.

Regolith. The surface is pea gravel, an explicit practicality substitution by the operators [2], so nothing about dust, sinkage, cohesion or tribology transfers. The Roverscape tests navigation, perception and operations, not terramechanics.

Repeatability of the terrain. No preparation procedure between tests is published, and the field is outdoors, so surface state is a condition of the day rather than a set variable.

Campaigns run there

Surface Telerobotics, summer 2013. Three astronauts of ISS Expedition 36 remotely operated the K10 planetary rover on the Roverscape from orbit, simulating a mission in which an astronaut in lunar orbit deploys a radio telescope on the lunar far side [1][3]. A later paper from the same NASA group gives each of the three sessions as 3.5 hours [3], a figure the campaign paper’s own numbers, 40 minutes of crew training plus two hours of operations per session, do not add up to and that neither paper explains.

Session 1, on 17 June 2013 [1], surveyed the site and began deployment of a simulated telescope array, using surface-level survey data to assess terrain obstacles, slopes and undulations below the resolution of the orbital imagery or ambiguous because of its nadir pointing; session 2, on 26 July, deployed all three arms of the array and ran ahead of schedule into the inspection phase; session 3, on 20 August, started midway through deployment and performed remote visual inspection. Commanding ran over Ku-band by a combination of supervisory command sequencing and manual discrete commanding, for 11 hours of operation and 221.43 m, 170.14 m and 200.15 m driven across the three sessions [1].

Structured-light hazard detection, published 2017. A multi-dot laser projector and camera mounted on the leading edge of the K-REX2 rover were evaluated on the field against obstacle cases designed to mirror the geometry likely in lunar polar craters, in day and night runs chosen so that total incident illumination and dominant directionality matched [2].

VIPER, ongoing. The Moon Gravity Representative Unit, a weight-equivalent platform built to VIPER’s lunar wheel loads, has been photographed driving on the Roverscape surface. No source found states what test objective that run served, or whether it is a routine part of VIPER’s mobility and navigation development rather than a one-off imaging opportunity [4].

What is not established

The field’s own operators have published almost nothing about its physical construction: no tonnage or grain size distribution for the pea gravel, no bearing capacity figure, no preparation procedure run between test campaigns, and no slope inventory. Nothing published states when the site was commissioned; the earliest documented use is the 2013 Surface Telerobotics campaign [1][3]. The two published area figures, 8,100 m2 and 11,500 m2, are not reconciled by any source found [2][4], and no external user access route, lead time or fee schedule has been located. The MGRU photograph is the only evidence found that VIPER has used the site at all, and no source found states what that test was for [4].