SORA-Q

JAXA / TOMY Company / Sony Group Corporation / Doshisha University.

The problem

A rover carried by a precision lander gets the mass and volume left after legs, propulsion and guidance. SLIM landed by a deliberate two-step tip-over onto a main leg, a leg arrangement studied for stable touchdown in a two-step landing [8]. A vehicle of 80 mm diameter and about 228 g has to survive separation from that lander in free fall, fit a stowage volume with no room for a conventional chassis, drive on regolith with wheels a few centimeters across, and operate with no link to Earth of its own [1], [10].

The ground is loose: the top 15 cm of lunar regolith has a bulk density of about 1.50 g/cm3 [28]. A small wheel has little area to spread load, so sinkage drives the mobility design [1]. Wheel behavior also changes with gravity. In parabolic flight, drawbar pull to weight fell by 20 percent or more in lunar gravity at equal wheel load, so 1 g sand tests do not predict lunar sinkage directly [23].

What has flown

SORA-Q / LEV-2 (SLIM, 20 January 2024 JST). SORA-Q, formally Lunar Excursion Vehicle 2, was developed by JAXA, TOMY Company, Sony Group Corporation and Doshisha University [6]. TOMY answered the first request for proposals of the JAXA Space Exploration Innovation Hub in 2016 with the aim of an inexpensive insect-like robot built from consumer toy technology; Sony joined in 2019 for control boards and sensors and Doshisha University in 2021 for mechanism design [7]. The flight unit is an 80 mm sphere of about 228 g with two cameras [7], [10]. The ISAS account states that the LEV-M deployer released both rovers at about 00:19:50 JST, at the start of SLIM’s final free fall [9]. SLIM then touched down about 60 m east of its target and rested about 90 degrees from the planned posture after one of its two main engines lost most of its thrust [10]. Precision landing error was evaluated at 10 m or better [10].

On the surface the rover deployed from its sphere, righted itself, drove away from the lander taking images, selected frames onboard and sent them through LEV-1 with no ground command [6]. The first returned image, announced in January 2024, showed that the LEV-1 to LEV-2 link worked, that the transformation had occurred, and that the onboard algorithm had picked a frame with SLIM in view [5]. A rock common to the two returned images gives about 0.13 m of travel and about 180 degrees of rotation between them, and telemetry shows operation for at least about 108 minutes [6]. Anomaly detection and self-recovery functions acted autonomously during communication dropouts [6]. LEV-2’s behavior after release is partly inferred rather than observed [9].

LEV-1 (same mission). The 2.1 kg single-wheel hopper that relays LEV-2’s data hopped on the Moon six times by one JAXA account and seven by another [12], [9]. It transmitted for 1 hour 51 minutes from separation, and its onboard computer exceeded 70 degrees C [12]. JAXA describes the rover results only qualitatively in the peer reviewed landing paper [10].

MINERVA-II rovers (Ryugu, 2018). Each of the two is a disc of about 1.15 kg that hops by spinning and stopping a flywheel on one DC motor [14]. Rover 1A returned 609 images over 113 asteroid days and Rover 1B returned 39 before going silent after 10 days; hop distance could not be measured [14].

MASCOT (Hayabusa2, Ryugu). The 9.66 kg lander left its mothership with a release velocity of 4.94 +/- 1.7 cm/s (3 sigma) predicted for the flight model [16], then bounced and worked for 17 h 14 min [15]. Its hops were commanded from the ground, whereas SORA-Q’s imaging sequence was fully onboard [15], [6].

Surveyor 6 (1967). The one propulsive hop flown on the Moon before 2024 is the reference case for low-cost hopping mobility in a 2019 JPL briefing [17].

What is being tried now

Eccentric wheels and transformable stowage (flown on the Moon, ground tested before). The SORA-Q wheel rotates about a point 12 mm off the center of its hemispherical shell [1]. A motor aligns shaft protrusions with keyholes, a spring pushes the shafts out, and the head and tail stabilizer open passively, irreversibly [1]. Two gaits follow: a crawl averaging 24.4 mm/s and a butterfly motion averaging 21.6 mm/s on flat silica sand. Centric rotation left the rover fully entrenched after 20 wheel revolutions in that sand, while eccentric rotation climbed a 10 degree slope [1]. The tests ran at 1 g with a ground model of the flight design [1]. The flight image pair is the only on-Moon check [6].

Grouser design for light wheels (ground tested, flown on LEV-1). LEV-1’s single wheel used a 90-degree-tilted octagonal grouser selected with a resistive-force-theory model, validated in Tohoku silica sand No.5 and in discrete element simulation, then compared with lunar operation [30]. A separate single-wheel study of a grouser wheel from the Rashid rover design compared Toyoura standard sand with FJS-1 lunar simulant at slip ratios of 0 to 50 percent, on an Earth bench [22].

Predicting reduced-gravity mobility from 1 g tests (ground and parabolic flight). Two parabolic flight campaigns graded Earth-based prediction methods against lunar-gravity wheel data, and the equal cone index gradient method overstated mobility [23]. NASA’s mobility test practice for planetary rovers and tires is a recommendation set with no new measurements [29].

Consumer and commercial electronics (flown on the Moon). The control and image processing run on Sony’s Spresense processor board, a consumer IoT part, and JAXA frames the result as a demonstration of consumer devices in the space environment [5]. The release gives no radiation, thermal or vacuum qualification data [5].

Onboard image selection (flown on the Moon). A score built from the size and position of the detected lander rectangle favors frames showing the whole spacecraft and penalizes frames where it is too large or cut by the image edge; images are sent in descending score order [1]. The detector was tested only against a subscale lander model under a sun simulator [1].

Other hoppers and tumblers (flight pending or none). The Intuitive Machines Micro-Nova, called S. P. Hopper on IM-2, targets permanently shadowed terrain; the held descriptions are pre-flight [19], [18]. YAOKI is a small tumbling rover intended to be dropped into lunar caves [27]. A 1968 study designed a ballistic Lunar Hopping Transporter that was never built [35].

Small-body and polar rovers (in preparation). IDEFIX, the CNES/DLR rover for Phobos, is pre-flight [20], and its separation from the MMX mothership is analyzed in multibody simulation [21]. MoonRanger’s wheel module operated after thermal vacuum and vibration testing [26]. The Canadian Lunar Rover Mission is a planned micro-rover mission to the south polar region [25]. Coverage path planning for lunar micro-rovers fusing orbiter elevation and onboard LiDAR SLAM was simulated on an Apollo 14 elevation model and field tested on the CLOVER rover [24].

Legged and learned locomotion (simulation and bench only). A reaction-wheel bipedal hopper was simulated over seven hops on synthetic cratered terrain [31]. A tethered 3-DOF hopping leg with a custom gearbox hopped on a bench [33]. A reinforcement-learning controller for the Magnecko legged robot was tested in lunar gravity simulated by a constant-force spring offload [32], and a quadruped on granular lunar surface was analyzed in simulation [34]. None has flown.

The numbers

Parameter	Value	Source
Stowed form	sphere, 80 mm diameter	[1], [7]
Mass, flight unit	228 g	[7], [10]
Mass, design value in the mobility paper	250 g	[1]
Deployed size	123 x 136 x 90 mm	[1]
Wheel	78 mm diameter, axis offset 12 mm	[1]
Drive	brushed DC motor, gear ratio 256, 109 mNm at the wheel axes	[1]
Battery	6000 mWh, non-rechargeable, no solar cells	[1]
Design operating time	at least 2 h at -20 C	[1]
Cameras	two, one forward, one aft	[7]
Processor	Sony Spresense board	[5]
Link	short-range wireless to LEV-1, which relays to Earth	[9]
Crawl speed, silica sand, flat	24.4 mm/s	[1]
Butterfly speed, silica sand, flat	21.6 mm/s	[1]
Flight displacement between two images	about 0.13 m	[6]
Flight rotation between two images	about 180 degrees	[6]
Flight operating time	at least 108 min	[6]

Rover	Mass	Surface motion data	Source
SORA-Q	228 g	0.13 m and 180 degrees between two images	[10], [6]
LEV-1	2.1 kg	6 or 7 hops, no distance published	[12], [9]
MINERVA-II 1A/1B	about 1.15 kg	hop distance not measurable	[14]
MASCOT	9.66 kg	ground-commanded hops	[15]

The battery, torque and operating-time figures are design values from the mobility paper, not flight measurements [1].

What is not established

The lunar mobility result rests on two images, both partly corrupted by dropouts. It shows that the rover moved but gives no traverse rate, slip, sinkage, obstacle capability or power data [6]. The flight analysis comes from a JAXA release, not from the Science Robotics paper [6]. The mobility paper reports no on-Moon performance and leaves comparison with ground tests to future work [1]. Its ground tests used silica sand at 1 g, one 10 degree slope and one campaign [1]. The two JAXA accounts of LEV-1’s hop count disagree [12], [9]. Spresense radiation and thermal qualification is not reported [5]. No flight thermal data for LEV-2 are published, and the -20 C figure is a ground design point [1]. Wheel performance on lunar gravity and simulants is known from bench and parabolic data on other wheels, not this one [22], [23].