SAMPLR

Overview

SAMPLR is a lunar robotic arm payload built by Maxar with the Colorado School of Mines and NASA Goddard, selected as one of twelve payloads under NASA’s Lunar Surface Instrument and Technology Payload program for delivery on a Commercial Lunar Payload Services flight [11][15]. It consists of an arm, a sieve-scoop, a penetrometer and a camera system.

The arm is a second-generation Instrument Deployment Device, an evolution of the arms Maxar built for the Mars Exploration Rovers Spirit and Opportunity, with motors compatible with the lunar environment replacing the Martian ones [11]. The scoop descends from the same company’s scoop built for the canceled 2001 Mars lander and later flown on InSight. The penetrometer is a different lineage, derived from the In Situ Experimental Probe developed at the Colorado School of Mines under NASA’s SSERVI IMPACT team.

The payload’s primary objectives are to demonstrate the next-generation arm on the lunar surface, to capture regolith geotechnical data with the penetrometer, and to demonstrate a sieve-scoop that isolates and delivers particles of a wanted size to a lander or rover [11].

Specifications

Parameter	Value	Source
Degrees of freedom	5, reducible	[11]
Designed arm lengths	1 m and 2 m
Supported mounting height	0.5 to 1 m for the 1 m arm, up to 2 m for the 2 m arm
End effector	force-torque sensor and a turret with up to four radial instrument mounts
Scoop sieves	2, different mesh sizes, in the scoop side walls
Turret positions used	2 of 4, by the penetrometer and the sieve-scoop

Mass, joint torques, tip speed, work volume, power and positioning accuracy are not published, and neither arm length has been selected. A mass budget, once published, would be checked against the lunar surface itself rather than a simulant bin: the Lunar Sourcebook’s recommended intercrater bulk density runs from 1.50 g/cm3 in the top 15 cm to 1.66 g/cm3 by 60 cm depth, with a recommended shear strength of friction angle 49 degrees and cohesion 1.6 kPa averaged over that same 60 cm [3].

Mission profile

The payload was manifested on Masten Mission One, a CLPS delivery to the lunar south pole that did not fly [11][15]. No mass, power or accommodation figures from a current assignment have been published.

Mobility

None. The arm is fixed to its host lander and reaches within its own work volume.

Manipulation

Five degrees of freedom is the baseline, with the explicit option of building fewer joints where a mission’s task set allows it, which the designers frame as a mass and cost trade rather than a capability one [11]. The end of the arm carries a six-axis wrist force-torque sensor, the most accurate of five contact-force techniques compared on an arm kinematically equivalent to the MER Instrument Deployment Device, holding within 1 percent of its rated range against a load-cell reference where joint-torque and motor-current sensing suffered cross-axis errors up to 15 N; link strain gauges, the next most accurate technique compared, still trailed flexibility modeling and were judged useful only for detecting an arm collision; the wrist sensor’s own RMS error against the load-cell reference ran 0.41 to 1.19 N across runs [6]. The turret has as many as four radial mounting locations. Two are taken by the payload’s own tools, and the remaining two are offered to other instruments, which is one of the ways SAMPLR is described as enabling for other payloads.

Arm length is set by the host rather than by the task. Both 1 m and 2 m configurations were designed against mounting heights of 0.5 to 1 m and up to 2 m above the surface, so the arm adapts to whatever deck height the commercial lander provides [11].

Power and energy

Not published.

Thermal

Not published.

Compute and avionics

Not published.

Autonomy

Demonstrating variable autonomy to aid telerobotic missions is listed as a secondary objective [11]. What that means in implementation, and what the arm does without an operator, is not described.

Communications

Through the host lander.

Payload and instruments

The sieve-scoop collects regolith and sieves it. Two sieves of different mesh size are built into the scoop’s side walls, so a particle size can be bracketed between a large and a small cutoff rather than merely capped, and flow through the sieves is induced by rotating and vibrating the scoop [11]. Sample mass is measured with the wrist force-torque sensor rather than with a dedicated instrument. The scoop also removes surface material to expose regolith below it for other instruments, and can move rocks out of the work volume to a limited extent.

The penetrometer uses the arm as its linear translation mechanism and the wrist force-torque sensor as its instrument. From penetration and stress-relaxation forces it yields bulk relative density, cohesive behavior and surface strength. The supporting data published for it is from the predecessor In Situ Experimental Probe rather than from SAMPLR hardware: penetration force curves separating low-density JSC-1A from high-density GRC-3 simulant [11], measured at Earth gravity. How far that carries to the Moon depends on how dense the regolith already is: magnetically counter-weighted static cone penetration tests show the drop in penetration resistance from 1 g to lunar gravity runs from about 29 percent at 40 percent relative density to only about 9 percent at 76 percent relative density, and in situ lunar regolith reaches roughly 90 percent relative density by 60 cm depth [4]. Two-dimensional discrete-element modeling attributes the gravity dependence to force chains: reduced gravity does not remove the interlocking and friction that angular, densely packed grains supply, which is why normalized penetration resistance rises sharply as gravity falls even though raw resistance drops [4]. The lunar reference values the curves would be checked against, a cohesion of 1.6 kPa averaged over the top 60 cm by the Lunar Sourcebook’s synthesis of Apollo and Lunokhod data or 0.52 to 3.0 kPa by depth in NASA’s simulant guide, are not targets JSC-1A or GRC-3 is shown to hit: a standardized comparison of six simulants matched to the lunar particle size distribution found cohesion still differing by as much as 355 percent between them [3][12][7]. The abstract also proposes, citing that earlier IEP work rather than demonstrating it on SAMPLR hardware, that the rates of change of penetration resistance and of stress relaxation can indicate the water ice content of a regolith-ice mixture in a permanently shadowed region [11]; LCROSS ejecta modeling of one such region puts the ice-bearing layer at about 5.8 m thick with a water mass concentration of roughly 4 to 8 percent by weight, for a total pre-impact water ice mass in that permanently shadowed region of roughly 5 times 10^11 kg [8], the scale of signal the technique would need to resolve.

Modes of operation

Not published as a set. The described activities are work volume mapping with the camera system, scooping, sieving and delivery, penetrometry at the surface and at depth beneath removed material, and imaging of the surrounding terrain and of the lander and its payloads [11]. Scooping into regolith of unknown consistency is a known source of delay for arms of this class: Phoenix’s 4-DOF arm found indurated icy layers 3 to 10 cm below the surface across nearly its whole workspace, detected through rising joint torque and motion-impeded events rather than by a dedicated sensor, and needed a rasp on the back of its scoop to cut through them [1]. SAMPLR’s published activity list has no equivalent subsurface-hardness step: the camera system maps the work volume and the force-torque sensor weighs what the scoop collects, neither of which is described as flagging a hard layer before the scoop reaches it [11]. An off-nominal mode of the kind InSight improvised over hundreds of sols after its self-hammering mole stalled, pressing and chopping with the arm and scoop to recover a different instrument, is likewise not among the modes SAMPLR’s abstract describes; that recovery campaign, run on a dedicated anomaly-response testbed at JPL, spanned frangibolt firing on sol 87 through the end of the recovery effort on sol 754 [13][14]. A lighter arm handling icy soil has managed sample delivery before: Phoenix’s arm delivered 18 samples across 53 dig sequences, including 8 to an optical microscope and 6 to a thermal evolved-gas analyzer, over more than 150 sols [1], a cadence against which SAMPLR’s own future delivery rate would be compared.

Ground operations

Not published for SAMPLR itself. Geotechnical characterization from a bin of simulant is sensitive to how the bin was prepared: a 150 kg BP-1 test bed for an unrelated plume-surface test found that unbaked simulant erupted below 2 Torr regardless of compaction, that vibratory compaction tripled to eightfold the scatter in measured shear strength while producing an inverted density profile, and that the most repeatable bin state came from pouring the simulant in from a scoop with no compaction step at all [9]. Whatever ground rig eventually qualifies SAMPLR’s penetrometer and sieve-scoop against JSC-1A or GRC-3 inherits that sensitivity: a reading is only as good as the bin it was taken from, and the same fidelity problem applies one level up, since a standardized six-simulant comparison found cohesion differing by up to 355 percent between simulants matched to the same lunar particle size distribution [7]. The Lunar Sourcebook and NASA’s own simulant guide agree that no terrestrial feedstock reproduces lunar mineralogy, density and shear strength together, which is why both recommend that a reported cohesion or friction angle always be published alongside the relative density it was measured at [3][12]. NASA’s simulant guide scores fourteen commercial and NASA simulants against Apollo and Luna reference data with a quantitative Figure of Merit across mineralogy, chemistry, particle size, particle geometry, density, magnetic susceptibility and shear strength, and its headline conclusion is that no single simulant wins across all seven [12]. Even the simulant itself needs upkeep between runs: BP-1 baked at 250 F for 2 hours stays usable for a month if kept sealed with desiccant, but a vacuum eruption during pump-down invalidates any geotechnical reading taken before it [9], and a standardized comparison found a highland simulant selected for a 110 metric ton testbed running 355 percent higher in cohesion than the lunar highland estimate despite matching density, compression index and friction angle within 30 percent [7].

Technologies developed

The program’s engineering result is a reuse argument made concrete. A Mars Exploration Rover arm design, twenty years old and flight proven across two rovers, is re-motored for the lunar environment and offered with a configurable joint count and two lengths, so that the same arm fits landers whose deck heights differ by a factor of four [11]. The scoop follows the same pattern from a different ancestor, the 2001 Mars lander scoop that eventually flew on InSight, modified with two side-wall sieves. That earlier scoop’s flight record includes years of off-label use: after InSight’s self-hammering HP3 mole stalled at 31 cm instead of the planned 3 to 5 m, the mission spent from sol 87 to sol 754 using the arm and scoop to press on and pin the mole and chop at the cohesive duricrust that had stopped it, turning the recovery attempt into a soil-mechanics experiment in its own right [13][14]. What the arm and scoop measured while chopping at the duricrust, a cohesion of 5.8 kPa by scoop-blade shear against 4 to 25 kPa from cone-penetration theory, and a penetration resistance of 0.5 to 1.2 MPa in the crust rising to 5.3 MPa below 30 cm [14], is the kind of scoop-derived geotechnical reading SAMPLR’s own scoop and penetrometer are meant to produce on the Moon.

The second result is the pairing of the wrist force-torque sensor with the arm’s own linear motion to make a penetrometer, so that geotechnical measurement costs a probe on a turret rather than a dedicated instrument with its own actuator [11]. Whether it works on the Moon is untested: every published measurement comes from the predecessor probe in Earth-gravity simulant, and how much of a penetration-resistance reading survives the trip depends on the regolith’s relative density in ways not yet worked out for this design [4]. The permanently shadowed regions the water-ice mode targets are not uniform either: refitted LCROSS ejecta data from Cabeus put the ice concentration there at 8.2 percent by weight near the surface, rising with depth over a roughly 5.8 m ice-bearing layer, a profile a single penetrometer insertion would only sample a slice of [8].

References