YAOKI

Dymon Co., Ltd.

The problem

A rover that rides to the surface in a commercial lander’s payload bay shares that lander’s power and radio link, has no ground crew to right it if it lands wrong, and is delivered at a price per kilogram that pushes it toward very low mass. Astrobotic advertised Peregrine delivery at about 1.2 million dollars per kilogram in 2019 [12]. A teleoperated vehicle that tips over cannot be recovered quickly, so YAOKI’s stated design aim is to survive an uncontrolled drop and keep working in either orientation.

YAOKI is Dymon Co., Ltd.’s answer: a 498 g two-wheeled rover, 15 x 15 x 10 cm, with both wheels on a common axle either side of a central body, rated to withstand 100 G [10], [11]. The name comes from the proverb “Nanakorobi Yaoki”, to get up again however many times one falls [10]. It was carried on the Intuitive Machines IM-2 Nova-C lander, launched 26 February 2025 [1], and was not deployed because the lander came to rest on its side [2]. Nothing about its mobility has been demonstrated on the Moon.

The base rate for the class is unfavorable. A 2024 review counts nine lunar lander and rover missions launched from 2019 onward, four of them failed by that point [8]. The range of surface hazards a small vehicle faces, from soft soil to dust on thermal surfaces, is compared in that review with operating limits compiled from earlier vehicles: ground pressure below about 7 to 10 kPa and a slope capability around 25 degrees [9].

What has flown

YAOKI (IM-2, March 2025). Athena touched down on 6 March 2025 about 250 m from its target at Mons Mouton, on its side inside a crater [2]. Intuitive Machines did not expect it to recharge, given the sun direction, the panel orientation and the crater temperatures, and declared the mission concluded the next day. The lander’s inertial unit indicated it was on its side, but the company’s CEO would not accept the attitude determination without an image [2]. NASA’s PRIME-1 suite had several milestones accelerated before the batteries depleted, and the TRIDENT drill did not penetrate the surface in that orientation [5], [2]. YAOKI was not deployed. The PRIME-1 and YAOKI payloads were limited by the lander attitude, not by a fault of their own.

YAOKI’s earlier manifest. The rover was first manifested on Astrobotic’s Peregrine lander under an agreement announced 14 October 2019 for a 2021 flight [12]. Dymon stated eight years of development, base technology complete in 2018 and more than 100 tests passed, none documented [12]. Peregrine Mission One launched 8 January 2024 without it and was lost in cislunar space after a valve failure ruptured its oxidizer tank about 92 minutes after separation [6].

LEV-1 and LEV-2 (SLIM, January 2024). JAXA’s two nano rovers were released by the SLIM lander just before its touchdown on 20 January 2024 [14]. LEV-1 executed six hops, confirmed by the change in sensed gravity across each hop, with hop direction chosen autonomously from camera imagery, and transmitted directly to Earth on S band for 1 hour 51 minutes until battery depletion, with the onboard computer above 70 degrees C [14]. Its single wheel carries a 90-degree-tilted octagonal grouser selected with a resistive force theory model and single-wheel tests in silica sand, and the paper reports the choice confirmed by operation on the Moon [15]. LEV-2 (SORA-Q) is an 80 mm, 250 g ball that opens into two eccentric wheels [3]. It deployed itself, righted its attitude and drove away from the lander with no ground commands, imaging the lander and selecting frames onboard [13]. A rock common to its two returned images shows it moved about 0.13 m and rotated about 180 degrees, and telemetry shows at least about 108 minutes of operation [13]. LEV-2 is the one rover with an architecture comparable to YAOKI’s that has operated on the Moon.

The landers around them. SLIM landed with better than 10 m accuracy at about 200 kg landed mass, using crater-matching terrain relative navigation [16], [17]. HAKUTO-R Mission 1 failed on 26 April 2023 when its altitude estimate reached zero about 5 km above the surface [18]. Mission 2 struck the surface at 42 m/s in June 2025 after its laser range finder returned measurements too slowly for the descent speed to be corrected [19]. Both failures carried the payloads with them.

What is being tried now

Impact-tolerant, orientation-indifferent rovers. YAOKI (single axle, two wheels, never deployed) and LEV-2 (transformable sphere, flown) are the two vehicles in this family [12], [13]. Dymon’s own tests are not published. Mitsubishi Chemical supplied a cyanate ester CFRP body in place of aluminum, with polyamide-imide wheels chosen to survive drops from the deployer [11].

Small four-wheeled rovers. CMU’s MoonRanger carries no direct-to-Earth radio, so any traverse beyond lander Wi-Fi range must be autonomous [7]. ispace-EUROPE’s TENACIOUS and the Canadian Lunar Rover Mission are in the same class; the Canadian mission published its science payload and landing site down-selection after a June 2024 preliminary design review, with no mobility figures [31]. Neither has flown. JPL places this class within wider extreme-terrain mobility work [4], [8].

Wheel and grouser design on the ground. Single-wheel testbeds with discrete element models are the main tool. A Rashid-1 style grousered micro-rover wheel was reproduced at the flight load of 24.5 N, and extrapolation to lunar gravity left traction coefficient and tractive efficiency essentially unchanged while forces fell by about one sixth; Toyoura sand understated performance against FJS-1 by 15 to 20 percent in traction coefficient [20]. A single-wheel testbed run at up to 1.0 m/s, about a hundred times flown rover speeds, found that a slick wheel in Tohoku Silica Sand No. 5 needs a slip ratio of 0.25 before drawbar pull turns positive, against about 0.10 at 0.2 m/s [21]. For crawler mechanisms, a 45 degree lug alignment reduced slip against 60 and 90 degrees in a single-wheel rig [22]. A 20.4 kg four-wheeled testbed with passive suspension sweeps wheel geometry and center of gravity, but its results are scored pass or fail by observation [23]. Simulators now add deformable wheel traces from discrete element runs [24]. All of this is ground work at 1 g, and parabolic-flight data on a single wheel grades reduced-weight testing as overstating mobility, with drawbar pull to weight falling 20 percent or more [30].

Thermal and dust tolerance. Solar absorptivity can rise by as much as 50 percent at only 12 percent area dust coverage, and Apollo rover brush cleaning that passed at 1e-6 torr failed on the Moon [25]. Candidate lander and rover coatings are being screened for abrasion and adhesion: seven commercial ceramic and cermet coatings were tested against wear, simulant adhesion and thermal shock [26], and a simulant-bound Taber wheel produced a higher wear index than the standard wheel, 3.64 against 3.29 mg per 1000 cycles [27]. MoonRanger’s thermal model, entirely predicted, sets a surface operating band of -10 to 35 C and finds dark transit needs at least 9.5 W of lander heater power [28]. A wide-temperature lithium-ion battery for the CADRE lunar rovers went through qualification vibration and thermal vacuum [32]. YAOKI’s lens coating belongs to this family, with no published data [11].

Autonomous navigation. Dead reckoning on a lunar analog drifted about 5 percent of distance traveled in MoonRanger work [7]. Crater-based localization from lidar or stereo beat a 2 percent relative-navigation baseline in simulation, with particle filter error near 0.75 to 2 m, and moved from TRL 2 to TRL 4 without a result on real surface data [29]. A one-tonne rover carried cargo between a simulated lander and habitat by lidar teach and repeat over 18 taught segments in a Canadian analogue terrain [33], and resource-aware scheduling for the Endurance concept was tested in a Mars Yard with a physically swapped battery [34]. None has flown.

Communications. Through the lander. On IM-2 a Nokia 4G/LTE network demonstration planned lander to rover link ranges of 200 to 300 m and up to 2 km using Lunar Outpost’s MAPP rover as the mobile user [1]. YAOKI was not part of it, and its own band, data rate and range are not published [10].

The numbers

YAOKI specifications

Parameter	Value	Source
Mass	498 g	manufacturer [10], [11]
Dimensions	15 x 15 x 10 cm	supplier [11]
Wheels	2, on one common axle	manufacturer [12]
Impact tolerance	100 G	manufacturer [10], [11]
Body material	CFRP with cyanate ester resin matrix, about 30 percent lighter than aluminum, fivefold safety factor, about 4 million yen lower transport cost per unit	supplier [11]
Wheel material	polyamide-imide	supplier [11]
Camera lens treatment	anti-soiling coating	supplier [11]
Self-righting	yes, operates either way up	manufacturer [10]

IM-2 mission

Parameter	Value	Source
Lander	Intuitive Machines Nova-C, IM-2	[1]
Launch	26 February 2025	[1]
Landing	6 March 2025, Mons Mouton, about 250 m from target	[2]
Outcome	lander on its side, rover not deployed	[2]
Mission end	7 March 2025	[2]

Flown micro rovers compared

Vehicle	Date	Measured outcome	Source
YAOKI	March 2025	not deployed	[2]
LEV-1	20 January 2024	six hops, 1 h 51 min direct S-band link	[14]
LEV-2	January 2024	about 0.13 m moved, about 180 degrees turned, at least about 108 min	[13]

What is not established

Drive speed, slope limit, obstacle capability, power source, battery capacity, radio band, data rate, processor and thermal margins have not been published by Dymon or any third party, and no peer-reviewed paper on the rover was located. Every YAOKI number traces to manufacturer and supplier material [10], [11], including the 100 G rating, which has no stated test standard, axis or duration, and the more than 100 tests, which have no stated types or results [12]. The mass, safety factor and cost savings for the CFRP body have no published baseline [11]. YAOKI was never deployed, so its mobility, thermal and communications claims have no flight validation, and its performance against the ground pressure and slope limits of prior lunar rovers [9] is unmeasured.

For the wider field, LEV-2’s lunar record rests on two images and one short episode, with no traverse distance or terrain-interaction data [13]. Lunar-gravity wheel performance is model output, not reduced-gravity measurement [20], and the simulants used on the ground differ from regolith [21]. Crater-based localization has no end-to-end result on real surface data [29], and thermal predictions for micro rovers are untested in thermal vacuum [28].