Header for SPIE use
(Micro Mast Arm &
Advanced Control &
& Instrument Placement)
Robot manipulator technologies for planetary exploration
, X. Bao, Y. Bar-Cohen, R. Bonitz, R. Lindemann, M. Maimone, I. Nesnas, C. Voorhees
Jet Propulsion Laboratory, MS 198-219,
California Institute of Technology, Pasadena, CA 91109
NASA exploration missions to Mars, initiated by the Mars Pathfinder mission in July 1997, will continue over the next
decade. The missions require challenging innovations in robot design and improvements in autonomy to meet ambitious
objectives under tight budget and time constraints. The authors are developing design tools, component technologies and
capabilities to address these needs for manipulation with robots for planetary exploration. The specific developments are: 1) a
software analysis tool to reduce robot design iteration cycles and optimize on design solutions, 2) new piezoelectric
ultrasonic motors (USM) for light-weight and high torque actuation in planetary environments, 3) use of advanced materials
and structures for strong and light-weight robot arms and 4) intelligent camera-image coordinated autonomous control of
robot arms for instrument placement and sample acquisition from a rover vehicle.
Keywords: Robot analysis, Ultrasonic motors, Rover-mounted manipulation, Autonomous sample acquisition, Autonomous
This paper describes the activities of the Planetary Dexterous Manipulators (PDM) task at the Jet Propulsion Laboratory in
1998. PDM is an on-going NASA telerobotics research effort to develop and demonstrate new technologies to enable or
improve manipulation capabilities for planetary exploration. The target PDM space mission application is the NASA Mars
Surveyor Program - a planned series of missions to explore the climate, geology, and possible biology on Mars over the next
several years. Activities planned with rover vehicles on some of the missions include close-up viewing, analytic probing,
instrument placement, sample exposure, and acquisition and collection. These activities require stowage and manipulation of
instruments, tools and samples.
The PDM effort develops advanced algorithms, software, and hardware for lander- and rover-based manipulation with
robotic arms and attached end effectors for planetary surface and near-surface exploration. PDM addresses the development
and evaluation of new manipulator component and control technologies and design methodologies for reduction of mass,
volume, power consumption and cost while extending the capability of dexterous robot arms consistent with planetary
exploration needs. The specific path taken by PDM to achieve these goals is to design, fabricate, demonstrate and evaluate
technologies for advanced and innovative implementations and applications of manipulator arms. The PDM task develops
software design tools for manipulator analysis and optimization, develops component- and system-level manipulation
systems and develops algorithms for intelligent autonomous sensor-guided control of manipulators for rover vehicles and
landers. The areas of activity of the PDM task and their relationships are illustrated on Figure 1.
The four areas of activity in 1998 were:
• Development of robot design tools – a
software package called Robot
Computer Aided Analysis and Design
(RCAAD) was developed to assist
designers of robot arms for planetary
rovers and landers to analyze and
optimize their designs.
• Development of component
technologies – finite-element modeling,
prototype fabrication and testing of
ultrasonic motors (USMs) was
conducted to demonstrate their use as
actuators for rovers and manipulators.
Correspondence: E-mail: firstname.lastname@example.org , Telephone: 818 354 9174, Fax: 818 393 5007
Figure 1 PDM task activities and their relationships.
Proceedings of the 6
Annual International Symposium on Smart Structures and Materials, 1-5 March, 1999, Newport Beach, CA. Paper No. 3668-17.
SPIE Copyright 1999.
• Design, fabrication and demonstration of manipulators for planetary rovers – two robot arms were built and delivered to
a rover technology integration and field demonstration project at JPL.
• Demonstration of advanced control algorithms and autonomous manipulation – autonomous acquisition of a small rock
with a rover and robot arm from 1 meter away and autonomous placement of a science instrument on a target with a
rover and a robot arm from 5 meters away were demonstrated.
In the following sections, we describe the specific developments and results from each of these activities.
2. ROBOT COMPUTER AIDED ANALYSIS AND DESIGN
The design of robot arms for planetary applications requires the designer to make trade-offs between many different
performance criteria. These include increasing the dexterity, payload, accuracy, repeatability, stiffness and workspace of the
arm, reducing its weight and power needs, and providing for efficient means of stowage, deployment and manipulation of
science instruments and samples. This complex problem is usually solved sub-optimally using intuition and experience. The
PDM task developed a software tool, RCAAD, to assist the designer visualize the performance of robot designs according to
different criteria quickly.
There has been some work previously reported that relates to RCAAD. Low-level functions of use in the analysis of robot
kinematics for the Matlab and Mathematica commercial packages were reported in  and  respectively. More
recently, Hill  developed a analysis tool with goals similar to RCAAD. A related but more general system analysis and
optimization tool was developed by Katragadda . RCAAD runs within Matlab, a commercial software package. When
run, it displays a window on the computer monitor. The main window of RCAAD is shown on Figure 2. A pull-down menu
under File is used for loading a previously created robot model, saving a modified or newly created model and quitting
RCAAD. Items under the Model pull-down menu are used to create:
• the model of the robot,
• models of tools to be installed
on the robot,
• a geometric model of a rover
for mounting the robot arm,
• geometric objects and
obstacles in the environment,
• gravitational fields in which
the analyses are to be
Items under the Analysis pull-
down menu select different
analyses to be performed on the
robot model. Analyses
implemented in 1998 are
manipulability, stiffness, accuracy
& repeatability, tip force/torque to
joint torque, and joint trajectory
There are four panels in the
display window. The top-left
panel is a 3-D graphic display of
the robot model and its
environment. On the bottom-left is
a move control panel to be used by
the designer to specify the posture of the robot. It includes buttons and sliders to select modes of controlling motion of the
robot. The bottom-right panel is used to control the view parameters of the 3-D graphic display. The panel on the top-right
changes as the user selects different analyses to perform on the robot model. Each analysis has its own unique set of buttons,
sliders, etc. that are used to perform its functions. In RCAAD, the designer creates a model of the robot design by specifying
the geometry, material properties and sensors for the robot arm. The parameters entered are visualized as in a 3-D graphic
display as component links of the robot arm are created.
Figure 2 The main display window of RCAAD.
The model of the robot may be specified using
either the Denavit-Hartenberg  notation or
with detail geometric, material and sensor
parameters for the links and joints. The display
corresponding to the Denavit-Hartenberg (D-H)
model is a stick-figure of the robot. A 3-D solid
model is displayed if the detail information is
entered instead. The models shown on Figure 3
illustrate the 2 displays. The robot model is
created by specifying the parameters for each of
its links. A link edit form – a window that pops-
up when a new link is to be created or a
previously defined link is to be modified – has
edit boxes for the parameters of the link. The
edit form can display a graphic preview of the
link to confirm the values entered. Similarly,
parameters for tools to be mounted on the robot,
components of the rover, obstacles and
geometric constraints in the environment can
also be created, modified or deleted.
Once the robot is defined, its
configuration can be specified
by entering its joint angles or
the position and orientation of a
selected tool mounted on the
robot. The configuration can be
entered as absolute position of
the respective parameters or
relative positions with respect to
the current position. This is
done with the Move panel. The
display panel is used to specify
the view of the 3-D display of
the robot and its environment.
RCAAD can automatically
determine the view to display.
However, the user can specify
the elevation, azimuth and zoom
settings of the display manually.
The analyses currently
implemented in RCAAD are
accuracy & repeatability, tip
force/torque to joint torque, and
joint trajectory planning of the
robot model. The analysis to be
performed is selected by picking
it out from the Analysis pull-
down menu. The selected
analysis has a corresponding
unique panel display that
appears on the tip right of the
RCAAD window. For example,
Figure 3 3-D gaphic of robot with D-H parameters (left) and detail
geometric .parameters (right).
Figure 4 Manipulability (left), stiffness (middle) and trajectory (right) analyses panels.
the stiffness analysis panel has buttons, sliders and edit boxes for specifying the force and torque on a specified tool and for
displaying the resulting deflection due to the applied force and torque and gravity acting on the tools and the links of the
robot arm. The applied force and torque vectors are graphically displayed on the 3-D model of the robot as are the resulting
position and orientation deflection.
3. ULTRASONIC MOTORS
Efficient miniature actuators that are compact and consume low power are needed to meet the NASA needs of future
missions. Ultrasonic rotary motors are an emerging actuation technology that offer compact, light motors with many
advantages for robotic applications including self-braking. These motors have high torque density at low speed, high holding
torque, simple construction, can be made in annular shape (for optical application, electronic packaging and wiring through
the center), and have a quick response. To assure the compliance of this materials with the harsh and demanding space
applications, rigorous analytical tools as well as adequate attention are need to determine the effect of extreme temperatures
and vacuum . Generally, ultrasonic motors can be classified by their mode of operation (static or resonant), type of motion
(rotary or linear) and shape of implementation (beam, rod, disk, etc.). Despite the distinctions, the fundamental principles of
solid-state actuation tie them together: microscopic material deformations (usually associated with piezoelectric materials)
are amplified through either quasi-static mechanical or dynamic/resonant means. Obtaining the levels of torque-speed
characteristics of USMs using conventional motors requires adding a gear system to reduce the speed, thus increasing the
size, mass and complexity of the drive mechanism. The emphasis of the current efforts is on rotary type motors. In Figure 5
the principle of operation of an ultrasonic motor (flexural traveling wave ring-type motor) is shown. A traveling wave is
established over the stator surface, which behaves as an elastic ring, and produces elliptical motion at the interface with the
rotor. This elliptical motion of the contact surface propels the rotor and the drive-shaft connected to it. Teeth on the top
section of the stator are intended to form a moment arm to amplify the speed. The operation of USM depends on friction at
the interface between the moving rotor and stator, which is a key issue in the design of this interface for extended lifetime.
Recently, a 3-D finite element analytical modeling was developed to examine the excitation of flexural plate wave traveling
in a rotary piezoelectrically actuated motor (Figure 6). The model was used to predict the excitation frequency and modal
response and it incorporates the details of the stator, which include the teeth, piezoelectric crystals, stator geometry, etc. The
theoretical predictions were corroborated
experimentally for the stator. Parallel to
this effort, USMs are made jointly with
QMI (Costa Mesa, CA) and they are
incorporated into the robotic arms as well
as being programmed for computer control.
Examples of the motors that were
developed jointly with QMI are shown in
Figure 7. The different diameters allow
various torque-speed levels. In addition,
the effects of low temperatures and
vacuum were investigated it was shown
that a motor with a novel actuation can be
operated at temperatures as low as -150
and 16-mTorr pressure (see Figure 8).
Figure 7 Three sizes of ultrasonic motors that were made using flexural
traveling wave as a drive mechanism.
Figure 5 Principle of operation of a rotary travelling wave motor.
Figure 6 A segment of a 3d finite element
model of a USM stator showing two stages.
Orbit of surface particles
Also, a USM that was tested in thermal cycles of 0 to -90
a period of 210 cycle did not show any significant change in
performance. The use of segmented ring for the actuation of the
motor showed a significant improvement in the longevity of the
motor with about 6 time longer durability as compared to a
continuous ring drive. Currently, issues associated to the
interface, longevity, efficiency and miniaturization are being
4. MICRO MANIPULATORS FOR PLANETARY ROVERS
The PDM task designed two robot arms for the
Exploration Technology (ET) Rover task, a
collaborating rover technology integration and
demonstration effort at JPL. The ET Rover Task will
perform integration of new technology and field tests
on a prototype vehicle to demonstrate concepts for the
Mars `03/`05 rover missions. The specifications for the
Mars `03/`05 rover require it to have an instrument
carrying robot arm to deploy and place up to four
science instruments against soil and rock samples for
analysis and a mast-like arm for panoramic viewing and
self-inspection. The arms developed by the PDM task
towards these specifications are the Micro Instrument
Arm (MIA) and the Micro Mast Arm (MMA).
The approach taken with the design of the two robot
arms was to share as much as much as possible in their
components. Consequently, there are a number of
similarities between the two arms. They both have four
degrees of freedom (DOF) – a shoulder yaw joint, a
shoulder pitch joint, an elbow pitch joint and a wrist
pitch joint. The joints are all fabricated of aluminum.
The links between the shoulder and elbow and the
elbow and wrist are a 1.5 inch diameter graphite-epoxy
The elbow joint configurations, however, of the arms
are different. The two tubular links of the Micro
Instrument Arm are side-by-side in the stowed
configuration while they are one-above-the-other in the
MMA. Two distinct types of joint designs are used
throughout each system. The first is a yoke and clevis
Figure 9 Micro Instrument Arm (front middle) and
Micro Mast Arm (back left) mounted on the ET
Figure 8 Torque-speed performance of a JPL/QMI
USM subjected to 150
C and 16 mTorr.
joint. This joint consists of a stationary clevis, which houses the actuator and gear train, and a yoke, which surrounds the
clevis on both sides and provides the output of the joint. The second joint configuration is an internal stator arrangement. In
this configuration, the stator hardware remains on the interior of the joint, while the rotor output rides on the outside. The
yoke and clevis joint was used on the shoulder, elbow, and wrist pitch joints of the MMA as well as the shoulder and wrist
pitch joints of the MIA. The internal stator arrangement was used in the shoulder yaw joint of both arms as well as the elbow
pitch joint of the by the PDM task MIA.
Joint angle sensing on all joints of the two arms is with potentiometers and magnetic encoders. For both arms, the gear ratios
of the shoulder yaw joints are 6600:1, shoulder pitch joints are 19800:1, elbow pitch joints are 19800:1 and on the wrist joints
are 6600:1. The joints were designed to allow for the internal routing of up to fifty-five 26AVG teflon insulated cables. The
MIA weighs 2.35kg. and the MMA weighs 2.42kg. The load carrying capacity of the arms is 20 Newtons for the MIA and 10
Newtons for the MMA.
These designs are the result of previous experience from the design of robot arms for rover vehicles [12,13] and of new
innovations made to increase performance and meet tight design specifications. The use of harmonic drives in the joint gear
reduction, the implementation of internal cable routing, the use of graphite-epoxy tubular links and the arm joint
configuration are legacies of the Micro Arm I and Micro Arm II designs developed at JPL. The concept of a mast arm on a
rover for panoramic viewing and self-inspection was first demonstrated on the Rocky 7 rover . Innovations implemented
in the MIA and MMA are:
• greater mechanical robustness needed for rugged field testing,
• increased range of motions on the joints and corresponding increased workspace,
• greater capacity for internal cable routing,
• incorporating hard stops on the joints range limits,
• increased payload capacity, and
• modular joints and corresponding cable layout for ease of disassembly and repair.
The mounting of the MIA and the MMA on the ET Rover is shown on Figure 9.The MIA is mounted on the front of the rover
under its solar panel. It stows under the panel and deploys to reach beyond the panel as shown on Figure 9. The MMA is
mounted on the rear side corner of the solar panel and it stows along the back edge of the solar panel. It deploys vertically
above the corner of the solar panel and reaches a height of 1.9 meters.
5. AUTONOMOUS SAMPLE ACQUISITION AND INSTRUMENT PLACEMENT
The PDM task also developed algorithms for autonomous sample acquisition and science instrument placement in 1998. The
demonstrations performed were the acquisition of a small rock sample designated by an operator autonomously from one
meter away using the Rocky 7 rover and the placement of a science instrument on a target designated by the operator
autonomously from five meters away. These capabilities enable a rover-manipulator system to perform science operations
with minimal input from the operator. In the Mars planetary exploration scenario, each command cycle between the Earth
and the rover on Mars occurs over one Mars sol – the equivalent of approximately an Earth day. Reducing the number of
command cycles required to perform science operations can greatly increase the accomplishments of the mission.
Related developments in visual guided manipulation reported for similar applications includes work at MIT  and at the
NASA Ames Research Center . The algorithms we developed use stereo processing algorithms developed by Matthies
and colleagues[8,9]. The advances made in our implementation of the sample acquisition and instrument placement
capabilities are modeling the operation on a realistic mission scenario
, performing all the autonomous computation on-board
the rover and performing the operations in relatively fast times.
The approach taken in the development of the autonomous manipulation capabilities in the PDM task was to design and
implement a control architecture to support high-speed visual guidance. The implementation of the capability addressed only
the demonstration of the autonomous behavior without considering other high-level functions of a planetary rover, for
example, obstacle avoidance or navigation functions. The sensory inputs used to perform the autonomous operations were
limited to those available on most rover platforms, i.e. odometry and black & white stereo vision. The portable, extendible
and re-useable object-oriented algorithms were designed to be easily ported to other platforms. The demonstration was
Sensors available on Mars rover missions include B&W cameras mounted on the rover and wheel odometry. The data is of
poor quality due to noise and wheel slippage.
successfully implemented on the Rocky 7 rover (see Figure 10). This development is retorted in greater detail in  and .
Rocky 7 was developed at JPL as a technology demonstration platform for future NASA Mars planetary exploration missions
The features of the development are:
• No assumptions were made that would prevent the use of the algorithms on NASA’s planned Mars rover missions to
explore the planet Mars.
• On-board cameras were used for vision.
• Noisy odometry was used for traverse estimates.
• The target location and an intensity threshold from one camera image are the inputs needed.
• Target designation can take place from anywhere in the world because the image and target positions are communicated
via a TCP (Internet) link.
• The autonomous software is all on-board the rover.
• The algorithm does not require a continuous view of the target.
The algorithms successfully demonstrated autonomous sample acquisition and instrument placement using the Rocky 7 rover
platform at operator designated targets from one meter and five meters away respectively.
These innovations extend the capabilities of robotic systems for a wide range of planetary exploration environments. The
activities reported above are continuing in 1999. The modeling of joint backlash will be included in the analyses performed in
RCAAD. Improved and faster 3-D visualization is also planned to provide real-time graphic updates of the robot models and
their environments. Analysis of lander or rover tilting from forces exerted during manipulation will be found. This feature of
RCAAD is expected to assist in the mission operations of the MVACS manipulator arm on the Mars `98 lander mission.
Our USM development will analytically and experimentally study the effects on operation longevity, variance in behavior
and the effect of vertical and tangential mechanical loads on USM performance. A finite-element model of the rotor-stator
interface and a miniaturized electronic driver and compact ultrasonic motor will be developed.
Replicas of the MMA and MIA robots will be installed on a rover mock-up. A control system for the arms will be developed
and research will be conducted into coordinated control of the arms to improve robustness and functionality. The autonomous
sample acquisition algorithm developed in 1998 will be extended to acquiring up to 3 samples autonomously from targets
designated by the operator from 1 meter away. One technology advance needed to enable this capability is a robust visual
odometry system for recovering the location of targets after loss of their view in the camera images.
This work was done at the Jet Propulsion Laboratory, California Institute of Technology under a contract with the National
Aeronautics and Space Administration. The development of ultrasonic motors was performed in collaboration with QMI, Inc.
of Costa Mesa, CA.
Figure 10 The Rocky 7 rover completing the small rock acquisition (left) and the science instrument placement (right)
1. Y. Bar-Cohen, X. Bao, and W. Grandia, "Rotary Ultrasonic Motors Actuated By Traveling Flexural Waves, " SPIE
Paper No.. Proceedings of the SPIE International Smart Materials and Structures Conference, SPIE Paper No. 3329-82, San
Diego, CA, 1-6 March 1998.
2. P.I. Corke, ‘A Robotics Toolbox for MATLAB’, IEEE Robotics and Automation Magazine, v3, n1, pp24-32, 1996.
3. J. Denavit and R.S. Hartenberg, ‘A Kinematic Notation for Lower Pair Mechanisms Based on Matrices’, Journal of
Applied Mechanics, v22, 1955.
4. S. Hayati, R. Volpe, P. Backes, J. Balaram, R. Welch, R. Ivlev, G. Tharp, S. Peters, T. Ohm, R. Petras and S. Laubach,
‘The Rocky 7 Rover: A Mars Sciencecraft Prototype.’ Proceedings of the IEEE International Conference on Robotics
and Automation, Albuquerque NM, April 20-25 1997.
5. B.M. Hill, 'Design of mechanical properties for serial manipulators', Ph. D. thesis, University of Texas at Austin, August
6. L.K. Katragadda, 'Synergy: A language and framework for Robot Design', Ph. D. thesis, CMU, December 1997.
7. M. Maimone, I. Nesnas, H. Das, “Autonomous rock tracking and acquisition from a Mars rover”, submitted to I-
SAIRAS `99 conference.
8. L. Matthies, "Stereo vision for planetary rovers: stochastic modeling to near real-time implementation," IJCV, 1992, Jul,
v8, n1,pp 71-91, 1992
9. L. Matthies, A. Kelly, T. Litwin and G. Tharp, "Obstacle Detection for Unmanned Ground Vehicles: A Progress
Report,” Robotics Research: the 7th International Symposium, Springer-Verlag, 1996
10. I. Nesnas, M. Maimone, H. Das, R. Petras, R. Steele, B. Balaram, “An object-oriented software architecture for planetary
dexterous manipulation,” submitted to I-SAIRAS `99 conference.
11. J. Nethery, and M. Spong, "Robotica: A Mathematica Package for Robot Analysis," IEEE Robotics & Automation
Magazine v1, n1, pp13-20, 1996.
12. P. S. Schenker, Y. Bar-Cohen, D. K. Brown, R. A. Lindemann, M. S. Garrett, E. T. Baumgartner, S. Lee, S.-S. Lih and
B. Joffe, "A composite manipulator utilizing rotary piezoelectric motors: new robotic technologies for Mars in-situ
planetary science," in Enabling Technologies: Smart Structures and Integrated Systems, SPIE Proc. 3041, San Diego,
CA, March, 1997
13. P. S. Schenker, E. T. Baumgartner, S. Lee, H. Aghazarian, M. S. Garrett, R. A. Lindemann, D. K. Brown, Y. Bar-Cohen,
S. S. Lih, B. Joffe, and S. S. Kim, Jet Propulsion Laboratory; B. H. Hoffman, Massachusetts Institute of Technology; T.
Huntsberger, Univ. of So. Carolina, "Dexterous robotic sampling for Mars in-situ science," Intelligent Robotics and
Computer Vision XVI, SPIE Proc. 3208, Pittsburgh, PA, Oct 14-17, 1997
14. D.A. Theobald, W.J. Hong, A. Madhani, B. Hoffman, G. Niemeyer, L. Cadapan, J.J.-E. Slotine, J.K. Salisbury,
“Autonomous Rock Acquisition,” Proceedings of the AIAA Forum on Advanced Development in Space Robotics,
Madison, Wisconsin, August 1-2, 1996.
15. R. Volpe, T. Ohm, R. Petras, R. Welch, J. Balaram, R. Ivlev, "Mobile Robot Manipulators for Mars Science." Space
Technology Journal, 17(3-4), pp. 219-229, April 1998.
16. D. Wettergreen, H. Thomas, M. Bualat , “Initial Results from Vision-based Control of the Ames Marsokhod Rover,”
Proc.IEEE/RSJ International Conference on Intelligent Robots and Systems, Control of Wheeled Robots, pp. 1377-1382,
Grenoble, France, September 7-12, 1997.