MINERVA

Norimaki, via Wikimedia Commons. CC BY 3.0.

Overview

MINERVA was the surface rover carried by the Hayabusa (MUSES-C) spacecraft to asteroid 25143 Itokawa [1]. It was the first asteroid exploration rover ever built and deployed [1]. Its objectives were to establish a mobile system in the microgravity environment of a small body, to demonstrate onboard autonomy, and to make the first planned scientific observations from an asteroid surface.

Specifications

Parameter	Value
Body	Hexadecagonal cylinder, 120 mm diameter, 100 mm high
Mass	591 g
Mobility	Hopping, 9 cm/s maximum
Power generation	Solar cells on every face, 2.2 W at 1 AU
Energy storage	Condenser, 4.6 V, 20 F
Processor	32-bit CPU at 10 MHz, 10 MIPS
Memory	512 kB ROM, 2 MB RAM, 2 MB flash ROM
Operating temperature range	-50 to +80 C
Communication	9600 bps, maximum range 20 km
Payload	3 CCD cameras, 6 sun sensors, 6 thermometers
Power, actuators	2.6 W maximum
Power, communication	1.8 W
Power, camera	8 W
Power, onboard computer	0.8 W

Source: [1].

Mission profile

Parameter	Value	Source
Host	Hayabusa (MUSES-C), 510 kg, core 1.0 x 1.1 x 1.6 m and 5.7 m across the deployed solar paddles	[1], [13]
Carrier launch	9 May 2003, M-V-5 from Kagoshima Space Center, Uchinoura	[13]
Target	asteroid 25143 Itokawa, reached 12 September 2005	[1], [13]
Carrier landings on Itokawa	two, in November 2005	[13]
Release	12 November 2005, during the second descent rehearsal	[2]
Outcome	released above the escape velocity of Itokawa; never reached the surface	[1]
Survival after release	about 18 hours, with autonomous functions working	[1], [12]
Data returned	one autonomously taken image, of part of the Hayabusa solar paddle, relayed through the spacecraft	[1], [12]

Why hopping

The argument is set out explicitly. Traction is friction, f = mu N, and for a wheeled vehicle the contact force is the weight, so f = mu m g [1]. At Itokawa’s surface acceleration of order 1e-4 m/s2 that product is negligible whatever the friction coefficient, and if commanded traction exceeds available friction the vehicle slips. A wheeled rover proposed as a MUSES-C payload could have driven at only 1.5 mm/s, and would have needed a hopping mechanism to cover tens of meters [1].

Adding an artificial pushing force F changes the scaling: f = mu (m g + F), and where F is much larger than m g this reduces to f = mu F, which is independent of the gravitational acceleration [1]. The upper bound on hop speed is the escape velocity of the body, about 20 cm/s at Itokawa against the 6e-5 to 9e-5 m/s2 surface acceleration that makes the weight term negligible in the first place [3].

The same relation is why cohesion, not weight, governs the surface material a hopper lands on. At 6e-5 to 9e-5 m/s2 on Itokawa, van der Waals attraction between grains exceeds their weight across the relevant particle size range [3], and Itokawa’s surface has minimum particle sizes of millimeters to centimeters with finer gravels migrating into geopotential lows.

Hopping mechanism

MINERVA hops by spinning an internal torquer and using the reaction on its body [1]. Unlike the contemporary JPL hopping rover, which charges a spring between hops and swings an off-axis camera mass through a one-way clutch to aim the next one [9], MINERVA has no moving part outside the body, which removes any requirement to consider asteroid dust in the mechanism. The torquer doubles as attitude control during the ballistic phase, and increases the contact force between body and surface, raising available friction and therefore achievable horizontal speed. Torque is adjustable because the DC motor is driven by pulse width modulation.

The mechanism was validated at the Japan Microgravity Center, which provides 10 s of microgravity in a capsule falling 490 m [1]. Five runs differing in PWM history were recorded on video at 1/30 s intervals, with the rover’s position and attitude recovered in each frame by least squares fitting of optical markers whose body-frame coordinates were known.

Case	PWM duty	Voltage	Duration	Horizontal speed	Vertical speed	Hop speed	Direction
1	100 percent	4.53 V	continuous	50.3 mm/s	47.2 mm/s	69.0 mm/s	46.8 deg
2	50 percent	3.54 V	continuous	38.7 mm/s	32.6 mm/s	50.6 mm/s	49.9 deg
3	25 percent	4.80 V	continuous	32.0 mm/s	26.4 mm/s	41.4 mm/s	50.5 deg
4	100 percent	2.33 V	0.5 s	19.8 mm/s	20.5 mm/s	28.5 mm/s	44.0 deg
5	50 percent	4.05 V	0.5 s	19.8 mm/s	21.4 mm/s	29.1 mm/s	42.8 deg

All values from [1]. The measured friction coefficient in these tests was approximately 1, the distance from the rover center to the contact point was 156.2 mm at an angle of 219.8 degrees, and gravity was set to zero in the matching simulation. Measured hop speeds of 28 to 69 mm/s sit below the 9 cm/s design maximum [1]. A drop tower reproduces the acceleration but not the target: restitution and ejecta behavior of a granular bed both change by an order of magnitude between 1e-2 g and below 1e-4 g [6].

The successor MINERVA-II rovers, which did reach a surface, use the same principle with a flywheel on the motor shaft and achieve about 10 cm/s for a 1 kg body [4].

Autonomy

Autonomy on MINERVA is built where inertial sensing is absent, which is the same design position MINERVA-II took thirteen years later on a rover that also could not localize itself [4].

The onboard control system is a three-layer hierarchy [1]. Independent controllers sit closest to the devices. An action layer commands actuators, sensors and the radio, at the granularity of “take the picture from camera A” or “rotate the motor with duty pulse X”, and these are the commands a ground operator addresses directly in teleoperation mode. A behavior layer composes those actions into autonomous sequences; the operational mode is read from a flash ROM area at boot and is normally switched by human command.

Image selection runs onboard. Pictures are compressed and evaluated by the CPU, images with the least information such as completely black frames are discarded immediately, and the rest are stored in flash ROM tagged with a priority proportional to their information content, so the highest priority data is transmitted first within the limited bandwidth [1]. MINERVA-II carried a developed form of the same filter, rejecting frames on the luminance histogram of a decimated channel against four thresholds settable from the ground [4].

Localization has no inertial reference. MINERVA cannot sense its attitude directly and carries no gyroscope, so onboard localization is an optional function [1]. The method used instead observes the Sun direction several times while the rover is stationary, in the evening after it stops and in the morning before it moves, using six photodetectors offset 90 degrees from one another so that at least three see the Sun simultaneously and their outputs give the solar direction. Rover attitude with respect to inertial space follows from those observations; the gravitational directions estimated during a hop, together with the times of sunset and sunrise, then give absolute position and attitude in the asteroid-fixed frame.

Thermal design

An asteroid surface at 1 AU swings across the whole operating range of commercial parts in a single rotation [3], so thermal control here is a duty cycle problem rather than a radiator sizing problem.

Thermal control is autonomous and consists of shutting down. Onboard devices operate from -50 to +80 C, and outside that range the power supply to the CPU is suspended [12]. The devices are covered in multilayer insulation, and the rover’s heat capacity is small enough that device temperature tracks the surrounding asteroid temperature. Night on the asteroid falls below -100 C, but the insulation slows the drop so the rover wakes quickly at sunrise; toward noon the temperature may exceed +80 C and the rover suspends operation [12]. The onboard software also watches internal temperature and, as it approaches a threshold, suspends power-hungry actions and copies RAM contents to flash ROM so that data survives a sudden shutdown.

The condensers are the limiting component. They are undamaged by cold but degrade above +130 C, and after a couple of days on the surface, having been through a couple of overheated daytimes, they can no longer be used [12]. The CPU and radio still work on direct solar power after that, but hop performance degrades and the rover may not move.

Four of the pins protruding from the body to protect the solar panels double as thermal detectors that measure surface temperature [12]. The 80 C shutdown threshold was carried forward to MINERVA-II, where it was never triggered because Ryugu’s short rotation period keeps the surface below it despite a very low albedo [4].

Payload

Three commercial CCD cameras sensitive at visible wavelengths are mounted on a turntable, so sight direction is set by rotating the table, and they view through windows at the centers of the side faces [12]. Focal length cannot be adjusted in flight, so two cameras are focused close and one far. All passed radiation level testing and were tuned for the hottest expected operating temperature.

Six photodetectors act as sun sensors, and six thermometers with four external pins measure surface temperature [12]. MINERVA-II reduced this to three or four cameras with photodiodes, thermometers, a gyro and an accelerometer, adding the inertial sensing MINERVA lacked [4].

Deployment and relay

The relay architecture is forced by the link budget: a 591 g rover cannot close a link to Earth, so everything passes through the spacecraft, exactly as for the later MINERVA-II and MASCOT deliveries [4].

MINERVA was held during cruise by the Onboard Mounted Equipment, which also supplied it with power over a wired line [12]. At deployment the OME cover is jettisoned and the rover is pushed toward the surface. After deployment the OME acts as the relay between the rover and the spacecraft data recorder: telemetry is received through an antenna on the asteroid-facing side of the spacecraft, and commands from Earth are stored in the OME and transmitted when a link to the rover becomes available. The link is 9600 bps to a maximum range of 20 km [12].

The body it was aimed at

Quantity	Value	Source
Principal axes	535 x 294 x 209 m
Mass	(3.51 +/- 0.105)e10 kg
Volume	(1.84 +/- 0.092)e7 m3
Density	1.90 +/- 0.13 g/cm3
Macroporosity	about 40 percent	[2], [3]
Surface acceleration	6e-5 to 9e-5 m/s2, 6 to 9 micro-g	[3]

Rows with no marker are from [2].

Itokawa is a rubble pile with no apparent monolithic component at the 100 m scale, decomposed into two ellipsoidal components resting on each other, and with tangential evidence for a non-homogeneous mass distribution [3]. The surface coefficient of restitution inferred from Hayabusa’s own landing altimetry and Doppler tracking is 0.84, high for unconsolidated gravel. Laboratory impacts into simulant below 1e-4 g give effective restitution of 0.15 +/- 0.04, an order of magnitude lower, which is the measure of how far a competent boulder field departs from a granular bed [6].

Returned Itokawa grains were later measured by nanoindentation: reduced Young modulus of 83.0 +/- 0.1 to 111.0 +/- 0.2 GPa on particles 131 and 149 micrometer across [5]. The material is competent silicate; the weakness of the surface is in the packing. Bulk cohesive strength for a regolith of Itokawa’s grain size is of order a few to tens of pascals [7], and simulant fidelity for this class of material is assessed against a figure of merit rather than any single property [8].

Operating mode sequence

MINERVA works only in daylight, on instantaneous solar power supplemented by the condensers, which cover a few minutes of activity when the arrays are shadowed [12]. Instantaneous array power alone is insufficient for motor rotation, image capture or communication; those require the condensers. Once the condensers have degraded, the machine survives as a camera and radio that cannot move.

Radiation

The commercial CPU and the cameras were qualified by radiation testing rather than by using radiation-hardened parts [12]. No dose figures are published. The interplanetary design environment for the two-and-a-half year cruise is the Badhwar-O’Neill galactic cosmic ray model [13].

What it established

The MINERVA hopping mechanism, an internal torquer with no external moving part, is the design that MINERVA-II carried to Ryugu and used successfully [4]. The friction relation f = mu (m g + F) and its microgravity limit f = mu F is the argument that wheels are unusable at this gravity, and it is stated in the MINERVA design paper before any asteroid surface had been reached [1].

The failure was in delivery, not in the rover. Deployment above escape velocity is a spacecraft state problem at release, and at an escape velocity of order 20 cm/s the margin between a successful release and a lost one is small [1]. The published account attributes it to the relative velocity and position at deployment rather than to any rover fault, and the 18 hours of subsequent autonomous operation is the only flight validation the design received [1], [12]. MINERVA-II demonstrated the same architecture at a body of comparable gravity thirteen years later [4].

A margin failure of this kind is not unique to Hayabusa. Philae’s hold-down system, a pair of harpoons and a cold gas thruster meant to pin the lander to 67P at 1.01 m/s touchdown speed, failed on both counts, and the lander bounced for nearly two hours before its second contact [11]. In both cases a small body’s near-absent gravity narrows the difference between a nominal release or landing and a lost one to margins that ground testing on Earth cannot reproduce, and in both cases the fault lay in the delivery mechanism rather than the surface package it carried. Hayabusa itself delivered its Earth reentry capsule to within one sigma of the predicted parachute opening point after a targeting sequence in which better orbit determination did not translate into better maneuver execution, an asymmetry between the two error sources that recurs in the deployment problem MINERVA depended on [10].