Sample Fetch Rover and Sample Recovery Helicopters

NASA/JPL-Caltech/AeroVironment. Public domain (NASA / US government work).

The problem

NASA and ESA describe Mars Sample Return as a joint campaign to bring Perseverance’s cached samples to Earth [17][16]. It needs a machine that finds sealed sample tubes lying on regolith, picks them up and delivers them to a lander carrying the ascent vehicle. Perseverance sealed 27 tubes, including witness tubes, between sols 109 and 1088 [24]. As of 21 March 2024 it had filled 21 rock cores, 2 regolith tubes, 1 atmospheric tube and 3 witness tubes [7]. The Orbiting Sample container takes up to 30 tubes [7], and the returned set is cataloged against astrobiology and geology questions in a traceability matrix for 36 cores and witness tubes [21].

The environment sets the design. Night temperatures reach -130 C under a thin CO2 atmosphere [14]. Tube positions are known from orbital imagery that is too coarse to drive a final approach, so the vehicle has to localize each tube from its own cameras [5]. Communication with Earth is not real time, so the fetch vehicle drives and picks autonomously [10]. Sample science adds a clock: four processes (organic degradation, headspace gas change, mineral and volatile exchange, redox change) alter samples once tubes are opened, which is why the sealed tubes must stay cold and the chain of host vehicles must hold a temperature [22][8]. Every kilogram of the fetching vehicle is paid for in landed mass: the Sample Retrieval Lander design point was 5500 kg at entry against 3400 kg for Mars 2020, with a 24 m parachute against 21.5 m for Mars 2020 [20].

What has flown

No fetch vehicle has flown. The flown systems are the ones the architectures lean on.

Perseverance sampling and caching. The Sampling and Caching Subsystem (a robotic arm, turret and corer outside the rover and an Adaptive Caching Assembly inside it) has a returnable design capacity of 37 samples [18], and a 40 sample design capability including blanks [19]. Perseverance is the baseline carrier of tubes to the lander in the helicopter architecture [15].

Ingenuity. The 1.8 kg, 1.21 m rotor technology demonstrator was planned as a 30 sol campaign with up to five flights [1]. It completed 72 flights from 19 April 2021 [13]. Flight 4 covered 271 m, flight 9 covered 631 m at 10 m height, and flight 12 lasted 170 s over 448 m [35]. Flights 68 and 69 in December 2023 were the first system identification flights of an aerial vehicle on another planet, reaching up to 15 m/s airspeed without instability or control saturation at an atmospheric density of 0.0137 kg/m3 [31]. Mastcam-Z video of six flights showed rotor downwash lifting dust below about 3 m altitude on takeoff, with 0.4 to 1.65 g of dust airborne 2 s after takeoff [32]. These are the only flight data on rotorcraft near a Mars surface, and the SRH design inherits its rotors, legs and architecture from them [15].

Autonomous driving. The AutoNav algorithm flown on Perseverance averaged 92 m/h against 17.3 m/h for Curiosity, and handled 90 percent of 32.1 km of odometry through sol 1312 [25]. The first global localization runs used the helicopter base station’s Snapdragon 801 as a coprocessor, a non-ECC part, with a 36 cm average position error and two successful Mars runs on sols 1762 and 1775 after memory-fault quarantine software was added [33]. Both bear on a fetch requirement of a long traverse with positions known only from orbit.

European precedent. The ExoMars Schiaparelli demonstrator, whose landing failed on 19 October 2016, is the European precedent for landing a heavy vehicle on Mars [2].

What is being tried now

None of the following has flown as a fetch system.

Dedicated fetch rover (Sample Fetch Rover)

ESA and ASI funded the SFR, built by Airbus Defense and Space at Stevenage [9][10]. Its 2020 status was 150 to 180 kg and a 20 km traverse in 150 sols [10], and a 2018 to 2019 mission analysis planned up to 20 km one way from the lander to the depot, with 5 sols for deployment and checkout and 20 sols for sample transfer [23]. Launch was set for 2028 and arrival for mid 2030 [14]. A 2009 JPL trade study of sample handling for a notional return rover had proposed rock cores about 1 cm by 5 cm and at least 20 cores [4]. The vehicle was never built; the mission was canceled before its thermal testing [14]. Locomotion and vision reached bench and simulation stage, described below. The 2022 architecture replaced it with two helicopters, saving mass and cost [15]. The end-2023 campaign architecture is a snapshot taken before an independent review [7].

Rotorcraft with wheels and an arm (Sample Recovery Helicopter)

The SRH is a proposal backed by ground-test rotors and a ground mobility model. It is two Ingenuity-derived coaxial helicopters, backups for Perseverance, deployed from the top of the lander [15]. Design details are in the sections below. Next-generation Mars rotorcraft concepts beyond it include a 17.7 kg hexacopter, for which thin ROAMX blades were analyzed for torsional stability and loads [28]. These are analysis only.

Learned tube detection and pose

Four published pipelines address finding the tube; none has run on Mars.

• A YOLOv3-tiny detector trained on photorealistic synthetic scenes with classical contour pose estimation was closed-loop tested on the ExoTeR rover on an indoor Mars testbed [3].
• A benchmark on 824 Mars Yard images compared a Line2D template matcher with Mask R-CNN [5].
• A YOLOv8 segmentation network with angle-guided perspective-n-point localizes the tube glove for the lander’s Sample Transfer Arm. Results are from synthetic renders, with flight processor runtime scaled from a desktop by clock ratio [11].
• Localizing a rover in aerial maps built from helicopter imagery was tested in the Mars Yard across feature detectors, viewing angles and times of day, as a way to use helicopter products for ground navigation [34].

Terrain-aware traverse planning

A proposed three-step method predicts slip and speed from HiRISE and thermal inertia data, plans a sol’s traverse over the resulting map, and deploys a shear vane if onboard slip monitoring is inconclusive. Trials are simulations over one annotated map area, and the speed distributions were assumed from a single published mean and maximum per category [6].

Rover and helicopter operations

Two analog campaigns tested how a helicopter changes operations. A five sol simulation at Rainbow Basin had the helicopter fly 1322 m and acquire about 16 Gbit against the rover’s 7 Gbit, but only 552 Mbit reached planners in time to affect decisions, because the platforms shared a downlink [26]. A July to August 2022 field campaign at the Holuhraun lava flow ran a 14 sol rover mission against a helicopter-only mission, with the rover’s planned 200 m traverse revised to 245 m [27].

Rotor and airfoil development for Mars

Airfoil tests in the Tohoku Mars Wind Tunnel at Re 20,000 and Mach 0.60 gave a maximum lift-to-drag ratio of 12.2 for the roamx-0201 cambered plate [29]. The Ingenuity clf5605 section measured a lift coefficient of 0.770 and a drag coefficient of 0.0985 at 6 degrees, and two models of the same nominal geometry differed in lift by 11 to 19 percent [29]. Full-scale hover tests of ROAMX rotors in a vacuum chamber reached a peak figure of merit of 0.574 against 0.445 for an Ingenuity replica at 0.015 kg/m3 [30]. These are ground tests at room temperature.

Mission profile and design details

Parameter	Value	Source
SFR launch	2028 on the Sample Retrieval Lander	[14]
SFR arrival	mid 2030	[14]
SFR mass	150 to 180 kg, maximum expected value	[10]
Orbiting sample capacity	up to 30 tubes in an 8 kg design	[10]
SFR surface duration	150 sols	[10]
SFR traverse	20 km in 150 sols	[10]
SRH deployment	from the top of the Sample Retrieval Lander	[15]
SRH role	backup to Perseverance for depot retrieval	[15]

SFR locomotion

The locomotion subsystem is a four-wheel-drive, four-wheel-steer deployable design with NASA superelastic spring tires, unusual among Mars rovers, which have all been six-wheeled [12]. Lander volume forced the tires down from 70 cm to 55 cm. The vehicle has a 130 by 97 cm footprint, 37.5 cm ground clearance, a 20 degree slope capability and a 29.5 cm single-sided obstacle climb on flat compact soil, with a nominal speed of 6.7 cm/s that averages 3.7 cm/s once navigation pauses are counted, more than 200 m per sol, and an energy cost of 3.1 kJ/m against 6.9 kJ/m for a six-wheeled alternative [12]. These are dynamic-simulation figures from before the vehicle-level test campaign. The tire design still needed another iteration after a life test failure. The related ExoMars locomotion model was a breadboard in an 11 m by 11 m indoor sandpit, at Earth gravity rather than Mars gravity [12][2].

Tube detection results

The YOLOv3-tiny pipeline takes 2808.62 ms per inference on an ARM Cortex-A72 and 33.46 ms with a Coral USB TPU, a factor of 84; pose estimation costs 14.12 ms on the ARM [3]. Closed-loop tests detected the tube out to 5 m, with a mean orientation error of 3.23 degrees from the close-range camera and 10.75 degrees from the navigation camera [3]. On the Mars Yard set, Mask R-CNN reached an average precision of 0.911 to 0.918 at an IoU of 0.5 against 0.255 for the Line2D matcher, which fails where shadows and bedrock edges mimic the tube’s parallel contours; recall for the better method stayed near 0.55 to 0.58 [5].

The Sample Transfer Arm is a 7-joint arm with a force-torque sensor, a placement budget of 9.35 mm lateral at the orbiting sample, a 350 N axial preload requirement and an insertion envelope of 15 N tip force [9].

SFR thermal design

The rover carried no radioisotope heater units and relied on electrical heating against a -130 C night. The dominant leak is conduction and convection through the CO2 atmosphere, so a gas insulation design around the hot zones was required [14]. Viking used 10 to 20 cm of fibrous insulation, Pathfinder a 50 mm honeycomb filled with foam plus another 50 mm, Sojourner silica aerogel, and MER 25 mm carbon-opacified aerogel bricks. From MSL onward JPL used a 25 mm CO2 gas gap, with 50 mm proposed for the Sample Return Lander [14]. JPL found convection onset at about 76 mm, while the ExoMars campaign found none at 30 mm, weak convection at 60 mm and significant convection at 120 mm and above [14]. The proposed SFR test of five configurations was never run.

The orbiting sample container is a passive thermal system whose cruise predictions are -70 to +15 C, with a -20 C steady state requirement for up to ten years [8].

SRH vehicle

The SRH keeps Ingenuity’s two counter-rotating coaxial rotors, solar panel, mast and carbon fiber legs. Two identical vehicles are carried because a failure anywhere in a coaxial rotor system grounds the aircraft [15]. Changes are a ground mobility system with tank-like steering and four lightweight wheels, a manipulation arm specified to grasp tubes of up to 150 g while massing under 100 g, higher-capacity batteries, a larger solar array, eight additional servos, an additional camera and in-flight absolute localization. Two Ingenuity parts, the radio and the 801 Snapdragon computer, were unavailable and are replaced by newer models [15]. The vehicle is about 2.5 kg against Ingenuity’s 1.8 kg [15].

Parameter	Value	Source
Rotor radius	0.6 to 0.7 m trade range; 0.7 m baseline from March 2023	[15]
Rotor speed range in test	2043 to 2500 RPM	[15]
Total mass	about 2.5 kg	[15]
Ground mobility	tank-like steering, four wheels	[15]
Arm target mass	under 100 g	[15]
Tube capacity	up to 150 g per tube	[15]
Servos added over Ingenuity	8	[15]
Vehicles	2, identical	[15]
Optimized rotor peak figure of merit	0.55 to 0.59 at blade loading 0.12 to 0.13	[13]

Rotor testing

CAMRAD II was used to optimize rotor size and planform, giving a larger radius, a similar chord and a twist close to Ingenuity’s [15]. Testing in the JPL 25 ft Space Simulator ran three campaigns: Ingenuity Engineering Design Model 1 with and without a cruciform box standing in for the lander takeoff platform, a transonic single-rotor rig, and an SRH dual rotor test [13]. The Ingenuity rotor peaks in figure of merit at blade loading 0.13, loses efficiency above tip Mach 0.75, and did not meet the SRH requirement [13]. The redesigned 0.706 m rotor rose from a figure of merit of 0.55 at 0.0120 kg/m3 to 0.59 at 0.0185 kg/m3 and met the requirement, so no further blade iterations were run [15][13]. Runs were limited to about 60 s by motor thermal limits. Two campaigns used a single force-torque sensor for the coaxial pair, so per-rotor forces were not resolved [15].

Takeoff from the lander

Ingenuity took off from the ground after deployment from the underside of Perseverance [1]. The SRH lifts off from the top of the lander through a Lift-off Adapter and Inverted Retention box, with no flight precedent. Mid-fidelity RotCFD runs show 15 m/s outwash developing over 8 to 20 s of a rapid collective ramp, and blade clearance from the lander must be shown for the whole sequence [15].

What is not established

Nothing in the SFR or SRH architectures has flown or been qualified. The SFR was canceled before its gas-insulation test, so none of the five proposed configurations has data [14]. Its locomotion figures are simulation [12]. Tube detection results come from indoor testbeds, the Mars Yard and synthetic renders, with no Martian imagery tested [3][5][11], and recall near 0.55 has not been translated into a fetch success rate [5]. The accelerator power savings that motivated the hardware choice were not measured [3].

For the SRH, rotor results are hover only at room temperature, with per-rotor forces unresolved in the single-sensor campaigns [13]. Figure of merit in the preliminary study came from estimated motor efficiency rather than measured shaft power, and the ground mobility system and arm had not been demonstrated [15]. Airfoil data at the relevant Reynolds number disagree between nominally identical models [29]. Takeoff from the lander box, outwash and blade clearance remain analysis [15]. The helicopter-augmented operations results are from analog simulations with shared downlink [26][27]. Every campaign-level mass, date and orbit is a snapshot of a plan under review [7][10].