Intelligent Unmanned Explorer for Deep Space Exploration 
T. Kubota and T. Yoshimitsu 
 
Institute of Space and Astronautical Science 
Japan Aerospace Exploration Agency 
Sahamihara, Japan. 
e-mail: kubota@isas.jaxa.jp 
 
 
Abstract 
In recent years, such small body exploration missions as 
asteroids or comets have received remarkable attention in 
the world. In small body explorations, especially, detailed 
in-situ surface exploration by tiny rover is one of effective 
and fruitful means and is expected to make strong 
contributions towards scientific studies. JAXA/ISAS is 
promoting MUSES-C mission, which is the world’s first 
sample and return attempt to/from the near earth asteroid. 
Hayabusa spacecraft in MUSES-C mission took the tiny 
rover, which was expected to perform the in-situ surface 
exploration by hopping. This paper describes the system 
design, mobility and intelligence of the developed 
unmanned explorer. This paper also presents the ground 
experimental results and the flight results. 
1 
Introduction 
Small planetary bodies such as asteroids, comets and 
meteorites in deep space have received worldwide 
attentions in recent years. These studies have been 
motivated by a desire to shed light on the origin and 
evolution of the solar system. Hence, exploration missions 
for small bodies have been carried out continuously since 
the late 1990s. To date, the missions of NEAR [1], Deep 
Space 1 [2], Deep Impact [3], and Stardust [4]  have been 
successfully performed, while MUSES-C[5] and Rosetta[6] 
are currently in operation. These missions have mainly 
provided remote sensing in the vicinity of the small body, at 
a distance which cannot be attained from the earth. In-situ 
observations of small bodies are scientifically very 
important because their sizes are too small to have 
experienced high internal pressures and temperatures, 
which means they should preserve the early chemistry of 
the solar system. For future missions, in-situ surface 
observation by robots will make strong contributions 
towards those studies.  
In such deep space missions, ground based operation is very 
limited due to communication delays and low bit-rate 
communication. Therefore, autonomy is required for deep 
space exploration using rovers. On the other hand, because 
the gravity and surface terrain are not known in advance, 
robotics and artificial intelligence technology must be used 
for rovers to explore a minor body. 
The Institute of Space and Astronautical Science (ISAS) of 
Japan launched the MUSES-C spacecraft toward Asteroid 
1998SF36 in May 2003. After the launch, the spacecraft 
was renamed "HAYABUSA" which means falcon. The 
MUSES-C project is demonstrating key technologies 
required for future sample return missions from extra-
terrestrial bodies as shown in Figure 1. The launch date 
was May 9th in 2003 and arrival at the asteroid was 
September 12th in 2005. Leaving the asteroid in April 2007, 
the spacecraft will return back to the earth in June 2010. 
HAYABUSA spacecraft has stayed for approximately three 
months around the asteroid and both mapping and sampling 
operations were carried out during that short period. For the 
MUSES-C mission, an asteroid surface exploration rover 
was developed, called MINERVA for MIcro/Nano 
Experimental Robot Vehicle for Asteroid. MINERVA is the 
first asteroid exploration rover to ever be developed and 
deployed. It is one of the technical challenges for the 
MUSES-C mission and its major objectives are as follows: 
establish a mobile system in the micro gravity environment 
of a small planetary body, demonstrate the autonomous 
exploration capability the rover is equipped with, and 
perform the first-ever planned scientific observations on an 
asteroid surface. On November 12th in 2005, the spacecraft 
Hayabusa deployed MINERVA at higher velocity than the 
escape velocity, and MINERVA could not arrive at the 
asteroid 
surface. 
However 
the 
spacecraft 
could 
communicate with MINERVA and it was confirmed that 
the health of MINERVA was very good and some 
autonomous functions were performed very well. 
 
 
 
 
 
 
 
 
 
Figure 1: 
MUSES-C mission 
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
18
© 2007 ICIUS
This paper presents the challenges of mobilizing a rover on 
a minor body surface and discusses the necessary 
intelligence such an exploration rover must have. Section 2 
discusses possible mobility systems for the micro-gravity 
environment on a small body surface, and presents a 
hopping mechanism that was developed in MUSES-C 
mission. Section 3 describes the intelligent functions of the 
micro probe robot. Section 4 presents the asteroid 
exploration rover MINERVA and flight results, which has 
been designed to explore an asteroid surface for the 
MUSES-C mission. Conclusions are stated in Section 5. 
2 
Mobility system 
The gravity on small asteroids such as Itokawa is on the 
order of 10 μG, which is extremely weak compared with 
that on Earth (1G equals to one earth gravity). Naturally, 
the escape velocity from the surface is very small as well, 
on the order of 20 centimeters per second. The largest 
asteroid yet discovered, Ceres, has a diameter of more than 
900 km, however the majority of asteroids have diameters 
less than several kilometers in size, and surface gravity 
lower than one milli-G. Rovers that are intended to move 
over a small body surface need a mobility system suited for 
this milli- or micro- gravitational environment. 
2.1 Mobility under microgravity 
The main principle to be considered is that the traction 
force which drives the robot horizontally is very small in a 
micro gravity environment. Traction force is obtained by 
the friction between the robot and the surface. Assuming 
that Coulomb’s friction law holds true in a microgravity 
environment, the traction force f is given by  
f = μ N 
(1) 
,where μ denotes the friction coefficient and N is the 
contact force. In the case of a traditional wheeled 
mechanism, frequently used by planetary rovers on the 
Moon or on Mars, the contact force is opposite to the 
gravity, which results in the following equations,  
f = μ N=μ mg 
(2) 
In Eq.(2), m is the mass of the rover and g is the 
gravitational acceleration on the surface. Eq.(2) indicates 
that the available friction for wheeled mobility is very small 
in the microgravity environment ( approximately 10-4 m/s2 ). 
Thus, even if the friction coefficient is assumed to be large, 
friction between a robot and the surface becomes very low. 
Additionally, if the traction is larger than the maximum 
friction, the rover slips. Thus the available traction must be 
extremely small, which makes the horizontal speed 
extremely slow. Actually, NASA proposed a new robot [7] 
with wheeled mechanism as a payload of the MUSES-C 
mission, but it could have only driven only with the speed 
of 1.5 mm/s. To achieve distances of tens of meters, it 
would have used a mobility mechanism to hop. 
Moreover, the surface of small bodies is a natural terrain 
with uneven surfaces. When the robot begins to move, any 
small disturbance may push the robot away from the surface. 
So the mobility system such as traditional wheeled 
mechanisms is not optimal under a micro gravity 
environment, where continuous robot-surface contact is not 
guaranteed. This reason motivates the search for another 
mobility system that specializes in hopping. Once a robot 
hops with some horizontal speed, it can move with no 
contact on the surface. 
A hopping mechanism causes the robot to push the surface. 
This increases the contact force between the surface and the 
body artificially with the pushing force, which makes the 
friction larger and can provide mobility at a higher 
horizontal speed that cannot be attained by a wheeled 
mechanism. Once it has hopped into free space with some 
lateral speed, the robot moves in a ballistic orbit under the 
influence of the weak gravity. Lateral speed is also obtained 
by the friction acting during the short period after the 
actuator starts and before separation from the surface. In the 
hopping process, the friction force exerted by the hopping 
mechanism is expressed as: 
f = μ ( mg + F ) 
(3) 
where F denotes the artificial pushing force which makes 
the robot hop. In the case mg << F, the friction force is 
approximated as: 
f ~ μ  F  
(4) 
Thus, a robot can use a friction force independent on the 
value of the gravity acceleration. In order to press its mass 
to the ground, some mechanism is required. By increasing 
the pushing force, the hopping mechanism is able to attain 
higher speeds which cannot be attained in principle by a 
wheeled mechanism. The maximum allowed speed is the 
escape velocity from the surface. If a larger speed than the 
escape velocity is achieved, the robot may never return to 
the surface again. 
2.2 Hopping mechanism 
There are several ways [8-16] to make a robot hop. The 
hopping mechanisms are categorized into two groups. One 
is to use repulsed force by striking, sticking, or kicking the 
surface. The other is to use reaction force by rotating an 
internal torquer. The newly developed hopping mechanism 
is shown in Figure 2. MINERVA installed such a 
mechanism and incorporates a torquer inside of its structure. 
All the previous studied designs have had a moving part 
outside of the rover body. However, MINERVA has no 
apparent moving parts outside, which has a big advantage in 
reliability for long-term outer-space use. After hopping into 
free space, the robot moves ballistically and comes back to 
the surface again, so long as its launch speed is less than the 
escape velocity from the surface. After a few bounces on 
the surface, it will settle and be ready to hop again. This 
action can be repeated, allowing the robot to explore the 
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
19
© 2007 ICIUS
surface of the small planetary body. This proposed 
mechanism has several significant advantages: 
1. The actuator is not outside the rover. Since the actuator is 
sealed inside the body, no consideration is required for 
dust, which may exist on the small body surface. 
2. The torquer can also be used for attitude control during 
the ballistic orbit. 
3. The contact force between the surface and the body is 
increased with help of the artificial pushing force made 
by the torquer, which makes the friction larger and 
provides for mobility at larger horizontal speeds. 
4. DC motors can be used as a torquer, which are easily 
controlled. The imposed torque must be adjustable in 
order to control the hopping speed. A DC motor driven by 
PWM is used to provide torque adjustability. 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 2: 
Hopping mechanism 
 
2.3 Experiments 
In order to confirm the hopping mechanism, microgravity 
experiments [17][18][19] were conducted at the Japan 
Microgravity Center (JAMIC), where 10 seconds of 
microgravity is obtained inside a capsule that free-falls 490 
m. The experimental robot was developed and experimental 
results were compared with the simulation analyses. Five 
experiments were conducted, as shown in Table 1, each of 
which differs in the parameters of the motor-supplied PWM 
history. Video monitoring was performed for each 
experiment. For example, the sequence of images for 
experiment #5 is shown in Figure 3 at intervals of 1/3 sec, 
where a PWM duty cycle of 50 % pulses was supplied for 
0.5 sec. 
In each experiment, the images from every 1/30 sec were 
extracted. The robot has many optical markers on its 
surface, whose three-dimensional coordinates in the robot-
fixed coordinate system are known. The visible markers are 
tracked in each image and the robot position and attitude in 
the experimental apparatus are estimated in every image by 
using a least squares method. For example, the tracked 
visible markers are shown in Figure 4 by squares for the 
image of experiment #5 after 8/3 sec had passed since the 
start of the torquer (corresponds to Figure 3 (j)). 
 
Table 1:   Specification of MINERVA 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
Figure 3: 
Experimental results 
 
 
 
 
 
 
 
 
Figure 4: 
Tracking markers 
 
2.4 Experimental study 
After estimating the position of the robot in each image, the 
estimated positions are plotted with time. After the robot 
hops into free space, the robot goes straight as the gravity 
inside the apparatus is zero. The linear line is fitted on the 
plotted positions to derive the slope of the fitted line, which 
shows the estimated speed of the robot. For example, 
Figure 5 shows the estimated positions of experiment #5, 
where the horizontal and vertical coordinates of the robot in 
the apparatus are plotted. The fitted linear lines are also 
overlaid in Figure 5, the slopes of which show the 
horizontal and vertical values of the hop speed. In this way, 
the hop speed and direction can be estimated. 
hop
landing
attitude
controlled
torquer
stopped
(back spin)
rapid forward spin
torquer
robot
rover rotation
rotor rotation
observation
(image acquisition)
ballistic
movement
small body surface
case 
duty ratio 
dm [%] 
duration 
tm [sec]
voltage 
VP [V]
#1
100 
∞
4.53
#2
50 
∞
3.54
#3
25 
∞
4.80
#4
100 
0.5
2.33
#5
50 
0.5
4.05
(a) before the fall
(b) t=0(torquer starts)
(c) t=1/3[sec]
(d) t=2/3[sec]
(e) t=1[sec]
(f) t=4/3[sec]
(h) t=2[sec]
(g) t=5/3[sec]
(i) t=7/3[sec]
(j) t=8/3[sec]
(k) t=3[sec]
(l) after the fall
120
130
140
100
160
170
180
150
200
210
190
535
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
20
© 2007 ICIUS
 
 
 
 
 
 
 
 
 
 
 
 
Figure 5: 
Experimental result (case #5) 
 
Table 2 summarizes all the experimental results of hop 
velocity and direction. The hop time is not shown in this 
table because it is difficult to distinguish whether or not the 
robot is contacting the surface. Using the hopping model 
described, simulation analyses were performed, which are 
shown in Table 3. The values used are shown in Table 1. 
The gravitational acceleration was set to be zero (gx=gy=0) 
and the coefficient of friction was measured and found to be 
approximately 1. The shape of the robot is a rectangle as 
viewed from the side and the distance from the center of the 
robot to the contact point is 156.2 mm, and the angle is 
equal to 219.8 deg. 
Tables 2 and Table 3 compare the experimental results and 
simulation analyses respectively, and show consistency 
between the results. Of course both results will differ a bit 
because the simulation model is not completely correct, the 
experiments include a lot of errors, the adopted estimation 
method does not derive the true value of the hop speed, 
direction, etc. It can be concluded that the simulation 
method provides approximate estimates of the truth and the 
proposed strategy can deduce the hop speed. Based on these 
experiments, our conclusion is that the hopping control 
strategy is expected to work well on the surface of the small 
body. 
With the proposed hopping strategy, the hop direction 
cannot be controlled because it is mainly dependent on the 
friction coefficient between the robot and the surface, which 
can neither be controlled nor estimated a priori. The hop 
speed is also affected by the friction coefficient, but here is 
not a major factor as the imposed torque largely dominates 
the hop speed. Thus the hopping strategy enables control of 
the hop speed. It may be difficult to control the hop 
direction, but by using the internal sensors and the acquired 
images of the surface of the planetary body, the hop speed 
and direction can be estimated during the ballistic phase, 
which enables the prediction of the friction coefficient and 
how much distance the robot flies. These techniques are 
under development but are not discussed further in this 
section. 
 
Table 2:   Experimental results 
 
 
 
 
 
 
 
Table 3:   Simulation results 
 
 
 
 
 
 
 
 
3 
Intelligence of unmanned explorer 
In deep space missions, ground based operation is very 
limited due to communication delays and low bit-rate 
communication. Therefore, intelligence is required for small 
unmanned explorer. 
3.1 Intelligent control system 
The hierarchy of the onboard control system [20] is shown 
in Figure 6. The lowest levels close to the devices are 
conducted by independent controllers. The next level is the 
action layer controlling the actuators, sensors and 
communication devices. For example, “take the picture 
from camera (A)”, “rotate the motor with duty pulse of X” 
are the commands on this level. The discrete commands to 
MINERVA from the earth can directly designate the 
commands on this layer when the CPU is working in the 
tele-operation mode. 
The highest level is the behavior control layer, which calls 
the action level commands to compose autonomous actions. 
MINERVA has some operational modes [21], which differ 
from the program of this layer. When it is booted up, a 
certain Flash ROM area is first referred to in order to 
extract the operational mode and select the program. The 
switch of this operational mode is basically conducted by 
human command. The surface of the asteroid may be very 
cruel in temperature, which changes beyond the range of the 
onboard devices of MINERVA working well. The 
developed rover has the automatic capability to wake and 
160
180
200
220
240
260
280
300
0
1
2
3
4
coordinates [mm]
t [sec]
horizontal
vertical 
x = 19.8t + 164.9
y = 21.4t + 200.1
start of torquer
hop speed 
case 
horizontal
vhx [mm/s]
vertical 
vhy [mm/s] vh [mm/s]
hop 
direction
θh [deg]
#1 
50.3 
47.2 
69.0 
46.8 
#2 
38.7 
32.6 
50.6 
49.9 
#3 
32.0 
26.4 
41.4 
50.5 
#4 
19.8 
20.5 
28.5 
44.0 
#5 
19.8 
21.4 
29.1 
42.8 
hop speed 
case
horizontal
vhx 
 [mm/s]
vertical 
vhy  
[mm/s] 
vh  
[mm/s] 
hop 
direction
θ h [deg]
hop  
time 
th [sec]
#1 
60.5 
50.6 
78.8 
50.1 
0.70 
#2 
35.5 
29.9 
46.4 
49.8 
1.14 
#3 
28.2 
23.9 
36.9 
49.7 
1.40 
#4 
24.8 
23.8 
34.4 
46.2 
0.50 
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
21
© 2007 ICIUS
shut down the CPU and is sensitive to the surrounding 
temperature so as to not work during hot or cold periods. 
 
 
 
 
 
 
 
 
 
 
Figure 6: 
Intelligent control system 
 
3.2 CPU system 
A CPU with high performance and low power consumption 
is required for unmanned small body explorer. Based on 
radiation tests, the commercial CPU is guaranteed to work 
in the high radiation environment of outer space. However, 
compared with the latest CPU commercially available, for 
example, its processing speed is drastically slow with only 
10 MIPS when the clock is driven at 10 MHz. The CPU 
system has small memory which preserves the program and 
small RAM as a working area. Additionally it should have a 
Flash ROM in which acquired data are stored. The surface 
of the asteroid may be a very harsh environment in terms of 
its temperature, which changes beyond the optimal 
operating range of the onboard devices. The rover has an 
automatic capability to wake up and shutdown the CPU 
depending on the surrounding temperature, so that 
operation is suspended during periods of extreme hot or 
cold. 
3.3 Intelligent power supply system 
The rover is powered by solar energy, supplemented by 
secondary batteries, which enable a few minutes of activity 
with no help from the sun when the solar arrays are blocked 
by shadows. The solar cells are attached on each face of the 
rover. The power generated is dependent on the attitude of 
the rover, with a maximum power expected when the rover 
is on an asteroid surface at a distance of 1 AU from the sun. 
Extra power is stored in the electric double layer capacitors. 
The onboard CPU works with the instantaneous power 
available from the solar arrays, but that is not sufficient for 
motor rotation, image capture and communication.  
For such operations, additional power support from the 
charged 
capacitors 
is 
necessary. 
Unfortunately 
the 
capacitors gradually degrade when temperatures go higher 
than 130 deg C. After a couple of days on the asteroid 
surface, the capacitors can no longer be used. However 
even if the capacitors are fully degraded, as long as the 
solar energy is supplied, the computer and communication 
system works in a low consumption mode, but the hop 
performance is degraded. 
3.4 Autonomous thermal control system 
The operational temperature range of the onboard devices is 
from -50 deg C to +80 deg C. Beyond this range, the 
devices do not work well. The rover is equipped with an 
automatic capability to wake up and shutdown the onboard 
CPU depending on the surrounding temperature. When the 
temperature becomes more than +80 deg C or less than -50 
deg C, the power supply to the CPU is suspended. The 
capacitors are not degraded if it is very cold, but will 
gradually degrade when the temperature goes higher than 
+130 deg C. After a couple of days on the asteroid surface, 
the capacitors can no longer be used because they will have 
experienced a couple of overheated daytimes. Even so, as 
long as the solar energy is supplied, the CPU works, but the 
hop performance is degraded and the rover may not move. 
The 
onboard 
software 
also 
monitors 
the 
internal 
temperature. If this comes toward the threshold of working 
temperature, the actions which consume much power are 
suspended and the data in RAM are transferred to the Flash 
ROM in order not to be lost by the sudden shutdown of the 
CPU system.  
The onboard devices are covered by MLI heat insulation 
material. The heat capacity of the rover is very small as the 
temperature of the onboard devices is easily synchronized 
with the surrounding asteroid temperature. During the night, 
the asteroid temperature will fall below -100 deg C, but the 
temperature on the onboard devices will not drop as quickly 
because the heat transfer is blocked. This makes for a quick 
wake up of the rover at sunrise because the device 
temperature is not so low. Toward noon, the temperature 
may be rise higher than +80 deg C and the rover will take a 
nap for a while. A CPU with high performance and low 
power consumption. 
3.5 Onboard image selection system 
The taken pictures are compressed and evaluated by the 
onboard CPU. The images with the least amount of 
information, such as completely black images, are 
abandoned immediately. Images with greater information 
are saved in the Flash ROM, attached with its priority 
proportional to the amount of information. When the rover 
communicates with the spacecraft to send the acquired data, 
data with highest priority is transmitted first. Using this 
selection strategy, scientifically important images are 
transmitted effectively within a limited bandwidth. 
3.6 Autonomous localization system 
Localization to calculate the current attitude and position of 
the rover is needed for robust navigation. Estimation of 
absolute position on the asteroid can only be attained if the 
attitude of the rover is known. However MINERVA can not 
directly sense its attitude. So localization is a very difficult 
problem and onboard localization is only an optional 
function. If the rover were equipped with gyroscopes, once 
the attitude of the rover is calculated in one moment by 
some means, propagation would enable one to find the 
attitude at a future time. For MINERVA, however, a 
controller
camera
DC motor
photo
detector
thermal
probe
transimitter/
receiver
A/ D
controller
Onboard CPU
Flash ROM
Asteroid Parameters
Operational Mode
Stored Telemetry
action
behavior control
function calls
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
22
© 2007 ICIUS
localization technique [22] is introduced for a rover not 
equipped with gyroscopes. After the rover stops in the 
evening and before the rover begins to move in the morning, 
the position and the attitude are reserved, when the solar 
directions are observed for a couple of times. The attitude 
of the rover with respect to inertial space can be estimated 
using the solar observation. Then the gravitational 
directions, which are estimated during the hop, as well as 
the hours of sunset and sunrise, are used to derive the 
absolute position and attitude in the asteroid-fixed frame.  
4 
Asteroid Exploration Rover 
4.1 MINERVA 
ISAS has developed an asteroid rover MINERVA [23] 
which is an experimental payload of the HAYABUSA 
spacecraft. Figure 7 shows the flight model of MINERVA. 
MINERVA had planned to explore the asteroid surface 
after being deployed from the HAYABUSA spacecraft. Its 
shape is a hexadecagon-cylinder with a diameter of 12 cm 
and a height of 10 cm, its weight was 591g. Table 4 shows 
the specification of MINERVA. It is powered by solar 
energy, supplemented by condensers which enable a few 
minutes’ worth of power when the sunlight is blocked by 
shadows. During cruise, MINERVA is also supplied with 
power by wired line from the OME(Onboard Mounted 
Equipment). 
OME holds MINERVA during cruise. When MINERVA-
deployment is triggered, the cover of OME is thrown away 
to allow MINERVA to be pushed towards the surface. After 
deployment, OME plays the role of a communication relay 
between MINERVA and the HAYABUSA spacecraft data 
recorder. Telemetry from MINERVA is received by OME 
via the antenna mounted on the asteroid-oriented face of the 
spacecraft. Commands from Earth are stored in the OME, 
and 
when 
communication 
becomes 
possible 
with 
MINERVA, the stored commands are transmitted. The 
communication bandwidth is 9,600 bps within a distance of 
20 km. 
4.2 Onboard sensors 
The appearance of MINERVA and the sensor allocations 
are shown in Figure 8. MINERVA has three CCD cameras 
mounted on the turntable. The camera sight direction can be 
controlled by rotating the turntable. There are camera 
windows at the center of the side faces of the rover, through 
which the cameras view the outside. All the cameras are 
commercially available and are sensitive at visible 
wavelengths. They have also passed radiation level tests 
and have been slightly tuned considering the hottest 
operational temperatures expected. The focal length cannot 
be adjusted onboard, thus the two cameras are set to look 
very closely and another one is to look far away. 
MINERVA has six photo detectors (PDs) used as sun 
sensors. A single sun sensor senses the intensity of the 
incoming light, not the solar direction. Thus by using two 
outputs from the directly illuminated sensors, the solar 
direction can be detected. The allocation of the sun sensors 
is offset 90 deg from each other. At least three detectors 
will sense the light simultaneously and, by comparing the 
outputs of these detectors, the solar direction is computed. 
There are also pins sticking out from the body to protect the 
solar panels. Four of the pins are also used as thermal 
detectors, by which the temperature of the surface is 
measured. 
 
 
 
 
 
 
 
Figure 7: 
Flight model of MINERVA 
 
Table 4:   Specification of MINERVA 
Body size 
Hexadecagonal cylinder 
φ: 120[mm], height: 100[mm] 
Weight 
591[g] 
Mobility system
Hopping ability [9cm/s(max)] 
 
Power supply 
Solar cells (2.2[W] at 1[AU]) 
Condenser (4.6[V], 20[F]) 
Onboard CPU 
32bit CPU(10M[Hz]) 
Memory 
ROM:512kB, RAM:2MB,  
Flash ROM:2MB 
Temperature  
Range 
Min:-50[deg C],  
Max:+80[deg C] 
Communication
9600bps (max range: 20[km]) 
Payloads 
CCD camera×  3,  
Sun sensor×  6,  
Thermometers ×  6 
Power  
Consumption 
2.6[W] for actuators (max.),  
1.8[W] for communication 
8[W] for camera,  
0.8[W] for on-board computer 
 
 
 
 
 
 
 
 
 
Figure 8: 
Onboard sensors 
sun sensor
solar panel
pin
sun sensor
pin
antenna
camera
window
thermal probe
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
23
© 2007 ICIUS
4.3 Flight results 
MINERVA was deployed on 12 November 2005, when 
HAYABUSA was very close to the target asteroid Itokawa 
[24][25] as shown in Figure 9. Unfortunately MINERVA 
did not reach the asteroid surface, because the relative 
velocity and position of the deployment was not so good. 
Thus the surface exploration by the rover was not 
conducted. However MINERVA survived for about 18 
hours while the autonomous capabilities worked very well. 
MINERVA succeeded in taking the picture autonomously 
and sending the image data via Hayabusa to the earth. 
Figure 10 shows the obtained image data, which is a part of 
solar paddle of Hayabusa spacecraft. 
 
 
 
 
 
 
 
 
Figure 9: 
Target asteroid Itokawa  
 
 
 
 
 
 
 
 
 
Figure 10: Image data taken by MINERVA  
 
5 
Conclusions 
This paper has presented the design of an exploration robot 
which can move over the surface of small bodies with very 
weak gravity. Firstly this paper discussed issues related to 
mobility in a micro-gravity environment. A new mobility 
system with hopping function was introduced. This paper 
also described the intelligence of deep space explorer. Then 
this paper presented the world first asteroid exploration 
rover "MINERVA", on board the HAYABUSA spacecraft. 
MINERVA could not explore the surface of the target 
asteroid Itokawa, but the effectiveness of the most of 
intelligent functions were confirmed by the flight data. 
 
References 
[1] R.Farquhar, NEAR SO FAR, IAA Int. Conf. on Low-
Cost Planetary Missions, No.IAA-L-0201, 2000. 
[2] M.D.Rayman, P.Varghese, D.H.Lehman, L.L.Livesay, 
Results from the Deep Space 1 Technology Validation 
Mission, Acta Astronautica Journal, vol.47, pp.475-
487, 2000. 
[3] M.C.Chiu, J.Veverka, E.L.Reunolds, The Contour 
Mission – Status of Implementation, IAA Int. Conf. on 
Low-Cost Planetary Missions, No.IAA-L-0205, 2000. 
[4] K.L.Atkins, 
B.D.Martin, 
J.Vellinga, 
R.Price, 
STARDUST: Implementing a New Manage-to-Budget 
Paradigm, IAA Int. Conf. on Low-Cost Planetary 
Missions, No.IAA-L-0202, 2000. 
[5] J.Kawaguchi, K.Uesugi, A.Fujiwara, The MUSES-C 
Mission for the Sample Return - Its Technology 
Development Status and Readiness, IAA Int. Conf. on 
Low-Cost Planetary Missions, No.IAA-L-0306, 2000. 
[6] K.Wittmann, B.Feuerbacher, S.Ulamec, H.Rosenbauer, 
JP.Bibring, D.Moura, et al., Rosetta Lander in situ 
Characterization 
of 
a 
Comet 
Nucleus, 
Acta 
Astronautica Journal, vol.45, pp.389-395, 1999. 
[7] B.Wilcox et al., Nanorover Technology and the 
MUSES-C Mission, Proc. of 4th International 
Symposium on Artificial Intelligence, Robotics and 
Automation in Space, W-6-2, 1997. 
[8] T.Yoshimitsu, I.Nakatani, T.Kubota, New Mobility 
System for Small Planetary Body Exploration, Proc. 
of the IEEE International Conference on Robotics and 
Automation, pp.1404-1409, 1999. 
[9] P.Fiorini, S.Hayati, M.Heverly, J.Gensler, A Hopping 
Robot for Planetary Exploration, IEEE Aerospace 
Conference , 1999. 
[10] S.Schell, A. Tretten, J.Burdick, S.B.Fuller, P.Fiorini, 
Hopper on wheels: evolving the hopping robot 
concept, IEEE Field and Service Robots, 2001. 
[11] http://www.planetary.org/mars/tpr_rover-rus.html 
[12] T.Yoshimitsu, T.Kubota, I.Nakatani, T.Adachi and 
H.Saito, Hopping Rover MINERVA for Asteroid 
Exploration, Proc. 5th i-SAIRAS, pp.83-88, 1999. 
[13] E.Hale, N.Schara, J.Burdick, P.Fiorini, A Minimally 
Actuated Hopping Rover for Exploration of celestial 
Bodies, IEEE Int. Conf. on Robotics and Automation, 
pp.420-427, 2000. 
[14] E.Hale, N.Schara, J.Burdick, P.Fiorini, A Minimally 
Actuated Hopping Rover for Exploration of celestial 
Bodies, IEEE Int. Conf. on Robotics and Automation, 
pp.420-427, 2000. 
[15] K.Yoshida, The Jumping Tortoise: A Robot Design 
for Locomotion on Micro Gravity Surface, Proc. of 
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
24
© 2007 ICIUS
5th International Symposium on Artificial Intelligence, 
Robotics and Automation in Space, pp.705-707, 1999. 
[16] Y.Nakamura, S.Shimoda, S.Shoji, Mobility of a 
Microgravity Rover Using Internal Electro-Magnetic 
Levitation, Proc. of the 2000 IEEE/RSJ Int. Conf. on 
Intelligent Robots and Systems, 2000, 1639-1645. 
[17] T.Yoshimitsu, T.Kubota, I.Nakatani, Nano-Rover 
MINERVA for Deep Space Exploration, Journal of 
Machine Intelligence and Robotics Control, Vol.3, 
No.3, pp.113-119, 2001. 
[18] T.Yoshimitsu, 
T.Kubota, 
I.Nakatani, 
T.Adachi, 
H.Saito, “Analysis on Hopping Mechanism by 
Microgravity Experiments,” Drop Tower Days 2000, 
pp.63-65, 2000. 
[19] T.Yoshimitsu, 
T.Kubota, 
I.Nakatani, 
T.Adachi, 
H.Saito, “Microgravity Experiment of Hopping 
Rover,” Proc. of the IEEE International Conference on 
Robotics and Automation, pp.2692-2697, 1999.. 
[20] T.Yoshimitsu, 
T.Kubota, 
I.Nakatani, 
T.Adachi, 
H.Saito, 
“Micro 
Hopping 
Robot 
for 
Asteroid 
Exploration,” 4th IAA International Conference on 
Low-Cost Planetary Missions, IAA-L-1104, 2000. 
[21] T.Yoshimitsu, 
T.Kubota, 
I.Nakatani, 
T.Adachi, 
H.Saito, 
“Micro-hopping 
robot 
for 
asteroid 
exploration,” Acta Astronautica: vol.52, pp.441-446, 
2003. 
[22] T.Yoshimitsu, T.Kubota, S.Aakabane, I.Nakatani, 
T.Adachi, H.Saito, Y.Kunii, “Autonomous Navigation 
and Observation on Asteroid Surface by Hopping 
Rover 
MINERVA,” 
Proc. 
Sixth 
International 
Symposium on Artificial Intelligence, Robotics and 
Automation in Space, AS017, 2001. 
[23] T.Yoshimitsu, T.Kubota, I.Nakatani, J.Kawaguchi, 
“Robotic Lander MINERVA, Its Mobility and Surface 
Exploration,” Advances in the Astronautical Sciences, 
Vol.108, pp.491-502, Univelt Inc., 2001. 
[24] H.Yano, 
T.Kubota, 
H.Miyamoto, 
T.Okada, 
D.Scheeres, Y.Takagi, K.Yoshida, M.Abe, S.Abe, 
O.Barnouin-Jha, 
A.Fujiwara, 
S.Hasegawa, 
T.Hashimoto, M.Ishiguro, M.Kato, J.Kawaguchi, 
T.Mukai, 
J.Saito, 
S.Sasaki, 
M.Yoshikawa, 
Touchdown of the Hayabusa Spacecraft at the Muses 
Sea on Itokawa, SCIENCE, Vol.312, pp.1350-1353, 
2006. 
[25] M.Uo, 
K.Shirakawa, 
T.Hashimoto, 
T.Kubota, 
J.Kawaguchi, Hayabusa Touching-Down to Itokawa - 
Autonomous Guidance and Navigation - , The Journal 
of Space Technology and Science, vol.22, no.1, pp.32-
pp.41, 2007. 
ICIUS 2007
Oct 24-25, 2007
Bali, Indonesia
ICIUS2007-A004-OP
ISBN 978-979-16955-0-3
25
© 2007 ICIUS