SHRIMP: A ROVER ARCHITECTURE FOR LONG RANGE MARTIAN MISSION

Thomas Estier*, Ralph Piguet*, Roland Eichhorn**, Roland Siegwart*

*EPFL (Swiss Federal Institute of Technology Lausanne)
DMT-ISR, 1015 Lausanne, Switzerland
E-mail : [email protected]
**Helbling Technik AG, Schachenalle 29, 5000 Aarau, Switzerland
E-mail : [email protected]

ABSTRACT INTRODUCTION

Long-range robotic missions for Martian exploration NEW MISSION OPPORTUNITIES ON MARS
imply a high degree of autonomy. As a matter of fact,
human teleoperation from Earth reduce the mission Performing research at locations very distant from the
range due to the transmission delay and cannot be landing location (hundreds to thousands of kilometers)
considered anymore. Moreover, the low solar radiance climbing up mountains, volcanoes (e.g. Olympus Mons
combined with the dusty atmosphere allow only low 24 km high) or down valleys or craters (e.g. Valles
power consumption and the extreme temperature Marineris 6 km deep, 4000 km long), constitute abilities
encountered reduce drastically the energy storage that offer new mission opportunities because these
capabilities. New locomotion concepts have to be places might reveal much more geological and
developed and investigated allowing to increase exobiological information than everywhere else on
mobility and subsequently reduce calculation and power Mars.
consumption. In this case, autonomous robots with
Such missions pose new requirements on the rover
virtually no power storage ability can be considered.
system :
The most advanced locomotion concepts are based on • The navigation reference point cannot be a lander
wheels or caterpillars (e.g. Sojourner, NASA or but has to be either an orbiter or other means.
Nanokhod, ESA). These rovers have clear advantages
• The long operation time requires local power
regarding power efficiency and complexity if compared
generation rather than a one-time energy storage.
with walking robots. However, they still have quite
The short term energy storage needs a long lifetime,
limited climbing abilities. Typically they can only
high energy density at low temperatures.
overcome obstacle smaller than their wheel size.
• The integrated (solar) power generation restricts the
In this paper we present Shrimp, an innovative long- journey to zones with sufficient energy supply
range rover architecture with 6 motorized wheels. Using (sunny side of mountains or valleys)
a rhombus configuration, the rover has a steering wheel • Low power consumption per traveled distance
in both, the front and the rear, and two wheels arranged determines the average speed at reduced light
on a bogie on each side. The front wheel has a spring radiance.
suspension to guarantee optimal ground contact of all
• Due to the long mission duration high autonomy
wheels at any time. The steering of the rover is realized
reduces the mission control resources.
by synchronizing the steering of the front and rear
wheel and the speed difference of the bogie wheels. • Low power consumption and yet high autonomy
This allows for high precision maneuvers and even requires simple, reliable systems with a low number
turning on the spot with minimum slip. of consumers such as active control loops,
microprocessors etc.
The use of parallel articulations for the front wheel and • For low power consumption a locomotion system
the bogies enables to set a virtual center of rotation at using wheels is the best.
the level of the wheel axis while maintaining a high
• In long range missions zones with very rough terrain
ground clearance. This insures maximum stability and
cannot always be avoided. Hence, good mobility
climbing abilities even for relatively low friction
such as climbing ability and ground clearance is
coefficients between the wheel and the ground. This
necessary.
rover is able to passively overcome unstructured
obstacles of up to two times its wheel diameter. With • While traveling in very rough terrain tipping over
this high mobility, this architecture is the perfect cannot always be avoided. Therefore a recovery
candidate for long range planetary missions. measure must be provided.
• Moreover, the low mass, the high mobility and the
A well functioning prototype of 3.5 kg has been ability to recover after tipping over allow landing in
designed and manufactured. A demonstration in the rougher terrain than used in the past. Therefore the
Mars surface testbed of ESTEC will be done during the number of accessible research locations is even
conference. further increased.
Keywords : locomotion, roving vehicles, planetary rover,
Mars exploration
• The low mass and dimension allows several rovers lander- and rover-based cameras can resolve
to be landed in one or more landing sites during one structures in the cm to mm range);
mission. d) Item a) deserves particular attention, because
currently employed landing techniques and their
targeting accuracy do not permit to land in
apparently very rugged terrain, in closely confined
areas like craters or valleys, and in more elevated
areas. Thus, two of the most fascinating structures
on the surface of Mars, the Valles Marineris (a
canyon system extending over some 4000 km
length, whose floor lies up to 6 km below the level
of the adjoining plateaus)
Valles Marineris

Olympus Mons South Pole Cap

Fig. 1: Overview of interesting research sites on mars ~6 km

Long-range scientific exploration should consist of an
orbiter and one or several rovers with long-range
capabilities. This concept facilitates communication to
earth (rover Æ orbiter Æ Earth) and allows to track the
rovers. Such a mission scenario also results in a more
efficient use of the resources because the orbiter can be
reused for a whole family of subsequent rover missions
that could land in very distant areas around a planet or Fig. 2: Part of Valles Marineris
moon.
and the Olympus Mons (a volcano extending ca. 24 km
In absence of a lander, the landing of the rover is above the surrounding lava plains and with this being
realized through a combination of much smaller reentry the highest mountain in the Solar system) can not easily
shields and parachutes with airbag cushions. However, be reached with current technology landers. A long-
these landing technologies have the risk that the rover range rover might land in a safe place in the vicinity of
might not touch down on its wheels and the landing area these structures and then attempt to drive into/onto
is hardly controllable precisely. Thus the rover has to them.
have sufficient recovery capabilities after landing which
could be integrated with the means of recovery after
tipping over.

SCIENTIFIC GOAL
In analogy to the journeys of terrestrial discoverers there
are three broad and partly interrelated topics:
1) Geography / topology / climate
2) Geology / mineralogy
3) Biology - in this case exobiology ~24 km
According to Dr. R. Rieder, Max Planck Inst. for
Geochemistry, the role of a long-range rover can be
seen in this context as an important complementary
research element to landers with short-range mobile
robots. It enables: Fig. 3: Olympus Mons towering over a cloud layer
a) discovering areas potentially not accessible by An attempt to reach the floor of Valles Marineris is of
stationary landers because of technical constraints, particular interest for two reasons:
b) exploring areas further remote from, or along 1) Strata in the walls of this canyon may present in
stretches between, stationary landers, their vertical arrangement a unique record of
c) obtaining more detailed geographic and topographic Martian geological history, otherwise only
information (e.g. images of higher resolution; the accessibly by deep drilling - a wealth of
best images from orbiting cameras are currently information and the dream of every geologist.
limited to ca. 1.5 m per pixel. Images taken by
2) If there has ever been water on Mars, and if the 8
presence of this water has enabled the development 6W
of life, then the likelihood of any of these life forms 4.5 h
being preserved to the present day is probably the 4W
highest somewhere in this low lying canyon, where
atmospheric pressure and temperatures are still 8.5 h
higher than average and even traces of liquid water
may have been preserved in underground layers.
0 time [h] 24.37
However, for operation in a valley the availability of
Fig. 4: Solar power on SHRIMP [array 0.2 m2]
solar energy must be checked.
If we assume a maximum efficiency of the solar cells
A long-range rover, equipped with a camera for high
including the power converter of 15% we can calculate
resolution panoramic imaging and with a properly
the solar power generation of the 0.2 m2 solar array of
selected, highly miniaturized instrument package to
SHRIMP (fig. 11). Presuming a power consumption of
address some of the key issues of geology/mineralogy
6 W for full speed movement, the rover will be able to
and exobiology (see also NASA's Athena-2 project or
operate during a period of 4.5 hours. During cloudy
ESA's exobiology initiative) would thus constitute an
days or at a meridian fare from equator, this operation
excellent research tool for a further refinement of our
time might be drastically reduced. Furthermore, as
knowledge of Mars.
experiments during the Pathfinder mission have shown
ENERGY HOUSEHOLD [LAN97], dust deposition on the solar array might
reduce the power with a rate of 0.3% per day. To avoid
As Shrimp will run on solar power only, the power this deterioration the solar array will be equipped with a
management is of high importance. cleaning system e.g as a special function of the
The orbital solar intensity on Mars is only ~43% of that manipulator arm. The inclination angle of the solar
on Earth because of its bigger distance to the Sun. The array has also an important influence on the available
average solar intensity in the orbit of Mars is in the solar power. However, it might anyway be advisable to
order of 600 W/m2. The intensity on Mars’s surface is reduce the speed of the robot during climbing
then further reduced to 50% of this value on a clear day operations. Additional mountain climbing has to be
or even to 20-30% on cloudy days. Additionally it also planned in such a way that the solar array get most
depends on the meridian of the actual rover position. energy.
Mars Power Operational Mode
Red. Factor

Worst Case

Locom Measure Commu

Best Case

Budget otion ments nication

Actuators 4W 1W
Controller 1W 1W 1W
Solar Constant Navigation Sensors 1W -
W /m2 485 705
(Distance to sun) Communication - - 3W
Clear Athmosphere Scientific Instruments APX
% 50% 353 - 1.5 W -
Transparence (one sensor MOS (1.5 W)
Cloudy Athmosphere at a time) CUI (1 W)
% 30% Additional sensor
Transparence (2 W)
Global Storm Total consumption 6W <4W 4W
% 15% 73 Max. Solar power,
Transparence
Solar Cell Area m/2 0.2 15 71 peak hours (4.5h) ~6 W ~6 W ~6 W
Solar Cell Efficiency, side hours (2x2h) ~4 W ~4 W ~4 W
% 15% 10.6
new, clean Margin
Solar Cell Efficiency, during peak hours (4.5h) 0W 2W 2W
% 10% 1.5
degraded during side hours (2x2h) - 0W 0W
average Noon Sun Table 2: Power budget of SHRIMP
deg 70 1.4 9.9
Angle wh.driving
Table 1: available Power on Mars [LAN91] As can be seen from the power budget table, there is
little power margin in locomotion mode. Fast
Figure 4 represents a typical solar intensity distribution locomotion is thus only possible during peak hours on
during a Martian day (24h37’). non-cloudy days. It is therefore unavoidable to focus on
this point during the project to optimize the power
consumption during locomotion mode.
 LOW TEMPERATURE ENERGY STORAGE Sojourner used one-way Lithium-Thionyl Chloride (Li-
SOCl2) batteries with an energy density of ~500Wh/kg.
Usually with the use of solar power generation an There are now Li-Ion cells for space application with
energy storage is used for bridging passing shadows and about 120 Wh/kg and double layer caps have about 2-3
night operation and to deliver peak power. Due to the Wh/kg all at room temperature. A trade-off has to show
intended challenging mission the requirements for such the best compromise between lowest operating
an energy storage are also challenging. temperature, useful capacity at low temperatures and
• Due to the necessary long lifetime a high number of deep discharge limitation and lifecycles compared to the
charging cycles is required. mass of the thermal protection system defining the
• The low availability of sun energy and the wish for lower operating temperature.
low mass require deep discharging to make full use
of the storage capacity. MECHANICAL DESIGN
• In order to save as much mass as possible with the
thermal control system the energy storage should not
be its design driver. Thus the energy storage should
be operable at the lower temperature limit of the
payload with the highest low temperature limit. e.g.
the electronics at about –50°C.
• For climbing over obstacles or for recovery after
tipping over discharging with peak power is
required.
However, there are also some mission restrictions that Fig 5: Schema of the mechanical architecture
relax the energy storage requirement:
• Due to the autonomy the rover does not need to OVERVIEW
move, process images and broadcast at the same Using a rhombus configuration, the rover has one wheel
time. mounted on a fork in the front, one wheel in the rear and
• It is not necessary to drive at night; just to survive it. two bogies on each side. Although our bogies have a
• The stored energy should be sufficient to supply the special geometry, it is the same basic principle as used
rover for some minutes with peak power and for a for a train suspension : a couple of two wheels mounted
night length with survival power. on a support which can freely rotate around a central
The most used types of energy storage are pivot.
electrochemical batteries with NiCd or Li-Ions. The front fork has two roles : its spring suspension
However, depending on chemical reactions these guarantees optimal ground contact of all wheels at any
batteries are temperature sensitive which limits the low time and its particular parallel mechanism produce an
temperature capacity and peak power. Since the elevation of the front wheel if an obstacle is
chemical reactions are not completely reversible the encountered (fig. 6).
number of charging cycles is comparatively low.
However, the energy density is very good.
As an alternative solution double layer capacitors
typically use high surface carbon and sulfuric acid as
electrolyte and utilise the physical principle of moving
electric charges. This in itself is not directly temperature
dependent, but the motion of charges in a cold or frozen
electrolyte does affect and limit the performance.
However, the operation is possible at considerably
lower temperatures than batteries. The fully reversible
physical effect allows very high cycle numbers and very
high charging and discharging peaks. On the other hand, Fig 6: Working principle of the front fork
the energy density is only in the order of lead batteries. The parallel architecture of the bogies and the spring
As a second alternative there are mechanical energy suspended fork provides a non-hyperstatic configuration
storage such as springs or fly wheels. They can be made for the 6 motorized wheels while maintaining a high
appropriate in terms of charging cycles, peak power and ground clearance. This insures maximum stability and
low temperature operation. However the energy density adaptability as well as excellent climbing abilities.
would be very low and therefore either its mass very The steering system (explained later in this article)
high or the available capacity very low. allows the rover to carry out a pure rotation even in
these extreme situations.
 BOGIES trajectory under this line is needed to insure a good
stability when the rover is on a convex ground.
The bogies are the first key components of the rover.
They provide the lateral stability during the motion even
on very rough terrain. To insure good adaptability of the
bogie, it is necessary to set the pivot as low as possible
and in the same time to keep a maximum ground
clearance. This problem is solved by using the parallel
configuration showed on fig. 7 that bring the virtual
center of rotation of the bogie at the height of the wheel
axis.

Fig 7: Explanation of the parallel bogie architecture Fig 9: Trajectories of the wheel axis of the front fork
FRONT FORK STEERING
As shown on Fig. 6, a trajectory of the front wheel with The steering of the rover is realized by synchronizing
an instantaneous center of rotation situated under the the steering of the front and rear wheels and the speed
wheel axis is helpful to get on an obstacle. The second difference of the bogie wheels. This allows for high
goal for the fork is to provide a maximum vertical precision maneuvers and even turning on the spot with
amplitude for the wheel. To find the optimal minimum slip.
configuration for the fork, we established the following
kinematic model (fig. 8) :

Fig 8: Parametric model of the fork

With the parametric equations of ξ, α et ψ as function
of the angle A, Fig 10: Configuration of the wheels on ground
π With two motors less than the K9 (or Sojourner)
α ( A) = − A + φ
2 architecture, this configuration allows to follow curved
ª c − b ⋅ cos [α ( A) ] º
trajectories in motion, therefore increasing the mobility
ψ ( A) = a cos « »
¬« b + c − 2 ⋅ b ⋅ c ⋅ cos [α ( A) ] ¼» of the robot and its ability to keep its stability in critical
2 2

ª b 2 + c 2 − 2 ⋅ b ⋅ c ⋅ cos [α ( A) ] + d 2 − e 2 º situations.
+ a cos « »
«¬ 2 ⋅ d ⋅ b 2 + c 2 − 2 ⋅ b ⋅ c ⋅ cos [α ( A) ] »¼ As function of an angle of gyration alpha and an overall
ξ ( A) = A −ψ ( A) speed Vref, three distinct speeds must be assigned: V for
the front/back wheels, Vext for the wheels of the
we have all elements to establish the position of the
external bogie and Vint for the internal bogie.
point P as function of the angle A:
As it can be seen on Table 3, this can be easily
§ c ⋅ cos( A) + h ⋅ cos[ξ ( A) ]·
P ( A) = ¨¨ ¸¸ implemented on very simple micro-controllers defining
© c ⋅ sin( A) + h ⋅ sin [ξ ( A)] ¹
V, Vext and Vint as ratio of Vref. Using only relative
Finally, we chose the different parameters to get the speeds of n/8*Vref, it can be done only with byte shifts
trajectory shown on fig. 9. The horizontal line is the and byte additions, which are the most basic instructions
height of the wheel axis when the robot is on a of processors. That means that even with different
horizontal plane. Note that the characteristic of the speeds for each wheels, it can be done consuming little
calculation power.
 alpha Vext V Vint
0 1 1 1
10 1 7/8 6/8
20 1 7/8 5/8
30 1 7/8 4/8
40 1 7/8 2/8
50 1 7/8 1/8
60 1 1 -0
70 7/8 1 - 2/8
80 6/8 1 - 4/8
90 5/8 1 - 5/8
Table 3: Simple implementation of different wheel
speeds as function of the steering angle

RECOVERY STRATEGY FROM TIPPING OVER

Due to the extremely challenging terrain in which the
robot will move about, the risk of falling and tipping
over can’t be ignored and a recovery strategy has to be
established. Fig 12: Dynamic simulation results of the shrimp arm
An interesting solution lies in the use of the robotic arm helping the robot to recover
which has been designed especially for this robot with
the following guidelines: EXPERIMENTAL RESULTS
• Low position of the center of gravity and low inertia,
This section presents some of the tests we performed
even in fully deployed configuration.
with our first prototype manufactured at EPFL.
• Ability to position tools (micro-camera, micro-
spectrometer, drill, etc.) around objects of interest at OVERCOMING ABILITIES
the ground level.
To evaluate the getting over abilities, we performed
• Ability to mechanically clean the whole surface of
several test on a critical configuration: the overcoming
the solar panels.
of a vertical step. Fig. 13 shows the main sequences of
• Ability to move a panoramic-camera in high position the rover climbing a step. First, the front fork gets on
for reconnaissance purposes the step, compressing its spring (shown fig. 6), then the
• Possibility to recover the robot after a tipping over. energy accumulated in the spring helps the first wheel
of the bogie to climb. When the second bogie wheel is
in contact with the wall, the bogie turns around the step.
At this time the center of gravity reached almost its final
height. Finally, the last wheel can easily get on the step.

Fig 13 Climbing sequence for a step of 22cm high

(twice the wheel diameter)
As the two bogies are independent from each other, it is
Fig 11: Various configuration of the robotic arm. even possible to climb the step if the robot is not
approaching perpendicularly or if only one bogie
This arm is based on a parallel structure actuated at the encounters a step. Although it was designed to climb
level of the solar panels (except for the rotating joint at steps up to 17 cm, the rover is able to climb even steps
the end of the arms). This allow to provide an excellent of twice its wheel diameter (22 cm).
workspace volume for the tools and even to reach the
ground with the elbow of the arms to put the robot on its The climbing ability is mostly given by the sequential
wheels after a fall (shown figure 12). rising of the center of gravity (CoG) provided by the
consecutive action of the wheels. Figure 14 shows the
trajectory of CoG for a step climbing of 17 cm. The
center of gravity goes up to 10% of the final height
when the front wheel is on the top of the step (fig 12b). CONCLUSION
Then the first bogie wheel, helped by the action of the
front fork, brings the CoG to 50%. The second bogie The presented concept demonstrator proves the
wheel and the rear wheel contribute each for feasibility of the locomotion concept using wheels for
approximately 25%. It can be seen that the mechanical low power consumption and yet achieving remarkable
structure allows a smooth movement of the CoG. rough terrain mobility. This is the basis for long range
missions to remote research sites even in very
challenging environments with important slope and
considerable number of obstacles. On Mars, it offers
new scientific opportunities to reach places which are
rich in geological and exobiological information.
90%
75%
The most crucial subsystems for this long range rover
50% are the power train including the energy storage, the
25% thermal subsystem, the autonomous navigation system,
10%
the recovery from tipping over, the communication
system and the miniaturized payload.
In order to integrate all these subsystems successfully
into a whole rover:
• a mission must be chosen,
• budgets and requirements defined and
Fig 14: Trajectory of the gravity center.
• solution concepts for all subsystems developed.
OFF-ROAD ABILITIES The work for the most challenging subsystems like the
thermal subsystem and the autonomous navigation
The rover demonstrates excellent stability in both system should be attacked first as they contain the most
smooth and rough terrain. It still moves with a lateral or development risk. Specially for the autonomous
frontal inclination of 40° (Fig 15, above) and is able to navigation subsystem another concept demonstrator is
overcome obstacles like rocks even with a single bogie needed for testing over long distances in rough terrain.
(fig.15, bellow). The rover was tested in various terrains
(sandy and gravelly soil) and showed that its For autonomous systems reliability and therefore
architecture was well adapted for fields motion even in simplicity must be the most important design guideline.
dunes or in furrows. Although a further development of the concept to a
space proof level still requires substantial efforts, the
new order of possibilities should justify it.

REFERENCES
[EST00] T. Estier, Y. Crausaz, B. Merminod, M. Lauria,
R.Piguet, R. Siegwart, “An innovative Space
Rover with Extended Climbing Alilities”,
Proceedings of Space and Robotics 2000,
Albuquerque, USA, February 27-March 2, 2000.
[LAN91] G. A. Landis, J. Appelbaum, NASA Lewis
Research Center,” Photovoltaic Power Options for
Mars”, Space Power, Volume 10 Number 2, pp
225-237 (1991)
[LAN97] G. A. Landis, P. P. Jenkins, NASA Lewis
Research Center, “Dust on Mars”, Proceedings of
Fig 15: Off-road stability test the 26th IEEE Photovoltaic Specialists
Using the models described in [WIL97], we computed Conference, pp. 865-869, 1997
the Mean Free Path (MFP) for the two Viking landing [STO96] H. W. Stone, “Mars Pathfinder Microrover: A
sites (VL1 and VL2). We set the height of passively Low-Cost, Low-Power Spacecraft”, Proceedings
of the 1996 AIAA Forum on Advanced
climbable obstacles to 17 cm instead of the true value of
Developments in Space Robotics, Madison WI,
twice its wheel diameter because there is at this time no 1996.
heavy payload on the rover. For VL1, we obtained a
[WIL97] B. Wilcox, A. Nasif, R. Welch, “Implications of
value of 35.7 (Sojourner: 9.6). For VL2, the MFP is 5.4 Martian Rock Distributions on Rover Scaling”,
(Sojourner: 2.4). This is a clue to say that this rover Planetary Society International Conference on
architecture is an excellent candidate for long range Mobile Robots and Rover Roundup, Santa
missions. Monica, 1997.