Unformatted text preview: In Proceedings of the 8th ESA Workshop on Advanced Space Technologies for Robotics and Automation
'ASTRA 2004' ESTEC, Noordwijk, The Netherlands, November 2 - 4, 2004 DEXARM – a Dextrous Robot Arm for Space Applications
A. Rusconi(1), PG. Magnani (1), T. Grasso(2), G. Rossi(2), J.F. Gonzalez Lodoso (3), G. Magnani(4)
(1) Galileo Avionica
via Montefeltro 8, 20156 Milano (Italy)
Email: [email protected]
Email: [email protected]
(2) Tecnomare SpA
S. Marco 3584, 30124 Venezia, ITALY
Email: [email protected]
Email: [email protected]
(3) SENER Ingeniería y Sistemas, S.A.
Avda. Zugazarte, 56 48930 Las Arenas (Spain)
Email: [email protected]
(4) Politecnico di Milano
Dipartimento di Elettronica e Informazione
Piazza L. da Vinci, 32, Milano, Italy
Email: [email protected]
In the recent past, several proposed European missions have shown the need of a robot arm comparable in size, force
and dexterity to a human arm. Therefore ESA has embarked on the development of a dextrous robot arm (herein
referred to as DEXARM) which could be used for diverse space robotics applications in which the
manipulation/intervention tasks were originally conceived for humans. These applications are typically external or
internal servicing of orbiting platforms or robotics for planetary exploration.
Beyond the design contraints posed by the tough space environment, the main challenges of this development lay in the
minimisation of resources (mass, volume and power) that the applications require. A target is to achieve a flight model
design of an arm with full joints instrumentation and harness, having a total system mass of 20 kg, with a 1-g payload
capacity of 10 kg. To achieve this goal, ESA has encouraged the exploitation of innovative approaches and
technologies to drastically minimise mass, volume and power consumption while providing adequate performance
(output torque capability and positioning accuracy/repeatability).
The current phase of the DEXARM project is focused on system requirements, arm architecture, design and
development of a joint prototype. The activities related to this phase are being executed in two parallel contracts, one of
which has been awarded to Galileo Avionica. The final objective of the project (next phase) is the development of an
engineering model of DEXARM, suitable for customisation and full qualification testing by user programmes. The first
user programme currently envisaged for DEXARM is EUROBOT , a three-arm robot complementary in capability to
an EVA crewmember to be used for ISS extravehicular operations.
This paper describes the system concept, architectural solutions and joint design solutions proposed by Galileo
Avionica and their partners to the accomplishment of this objective.
The DEXARM will provide dexterous manipulation capabilities which can be used for space robotics applications. The
primary objective of DEXARM is to serve as EUROBOT arm, thus supporting EUROBOT operations like translation
along the space station structure utilising standard EVA interfaces, execution or support to routine EVA tasks, removal
and replacement of ORUs, assistance in contingency operations that require EVA.
The main requirements which will most importantly drive the design of DEXARM are summarised here below:
• Functional and performance requirements:
positioning accuracy of 0.3 mm;
force-torque capability of 200 N and 20 Nm at the arm tip;
payload handling capability of 500 kg at 0-g, 10 kg at 1-g; 1 • • • Physical characteristics:
mass of 20 kg;
power consumption of 100 W;
length of 1 m;
dimension less or equal than EVA suit arm;
7 joints, spherical wrist;
Capability of performing 1-g operations without using any special off-loading device;
EVA operability (attach-detach DEXARM from EUROBOT body, back-drive the joints without the use of any
space station safety requirements are applicable, requiring special attention to robotic related hazards like
possibility of collision with other space station elements. ARM CONFIGURATION
Geometrical and kinematics configuration
A preliminary and heuristic evaluation of the existing 7 degrees of freedom (d.o.f.) arms over the system requirements
has been carried out to select basic architectures to be further analysed, according to criteria mainly related to arm
dexterity, cable passing, folding capabilities and human arm similarity.
In particular, considering arm dexterity, the most appealing kinematics solutions are the ones that try to maximise joint
angular range by implementing joint offsets or make use of pitch joints with asymmetrical range with adjacent roll
joints. Examples of these two basic architectural solutions can be found in the CAT arm , the SPIDER arm , the
SPDM  and the SSRMS  for offset joints and in the ERA  and K-1207i  for asymmetrical range pitch joints.
From these two basic architectures five tentative configurations for DEXARM have been designed, equipped with
spherical and non spherical wrists. The spherical wrist is a system requirement, mainly justified by the fact that the
inverse kinematics problem allows an explicit, computationally efficient solution for the arm orientation, thus obtaining
the required orientation with high accuracy and without the need of iterative algorithms. On the other hand, as it will be
shown, a non spherical wrist can lead to a significant improvement in arm dexterity and therefore it has been considered
in the analysis.
The five configurations are depicted in Fig. 1 to Fig. 5 together with their symbolic names and have been preliminarily
dimensioned to cope with workspace and folding requirements. Fig. 1. Offset arm Fig. 4. Human arm like Fig. 2. Offset Arm, Non Spherical Wrist Fig. 5. Human arm like, non spherical wrist 2 Fig. 3. Multi Offset arm In order to select a reference architecture, the following analyses have been carried out:
• Workspace and dexterity analyses, aiming to understand the positioning capabilities of the candidate arms inside
the nominal workspace. A significant set of workspace points, belonging to a plane including the first joint axis has
been considered and for each point a significant set of approach orientations to it, covering the whole solid angle
around the point, have been defined;
• Trajectory analyses on typical paths. The chosen trajectories are 2D grids of linear paths parallel to the X and Z
axes of the base reference frame and with specific length. Trajectory locations have been chosen in order to
replicate as much as possible the operations of an astronaut;
• Kinematics and static performances at the end effector. These analyses aimed to understand the capability of the
arm to produce speed and force at the end effector, given the speed and torque performances at joint level. Since
the kinematics and static performances at the end effector are linearly related to the relevant performances at the
joints through the Jacobian matrix, then it is possible to give specifications of the performances at the end effector
and compute the needed performances at joint level simply re-scaling results obtained at the end effector starting
from ‘unitary’ performances at the joints.
A summary of the arm performances, based on the above analyses, is given in Table 1.
Table 1. Kinematics analysis summary
Wrist Offset Non
++++ ++ Human like,
Dexterity ++ Execute
Trajectories ++ +++++ +++++ ++ +++++ Static
Performances ++ ++ + ++ ++ Human like,
Sph. Wrist From a mechanical point of view a basic design of the five selected arms has been drafted, assuming for the joints a
• From the arm force/torque requirements the required output torque of the joints has been preliminarily
• Three sizes for the seven arm joints have been assumed; the larger size has been used in the arm shoulder, the
medium size in the elbow, and the smaller one in the wrist;
• The joints have been schematised as cylindrical subassemblies with hollow shaft to allow the cable passing and
having a base rotating with respect to the other. The volume of the joint electronic boards has been included in
the joint envelope;
• The joints have been assembled with link elements to reproduce the selected kinematics.
Three types of joint connections used to preliminary design the five arms have been analysed and compared: fork-type
connection, L-shaped connection and cup-type connection. Table 2 shows the mechanical solutions preliminary drafted
and the cable routing in the three cases. The fork type connection implies a more complex shape of mechanical
components and a more difficult cable routing. The L-shaped connection can be obtained with a simpler shape of
mechanical components and also cable routing is very simple. The cup-type connection requires a minimum transversal
envelope, but the angular range is limited to about +/- 90°.
A critical aspect of arm design is the cable routing. In DEXARM every single joint has its own harness to connect the
actuators and sensors to the electronic boards, so the electronic boards should be placed as near as possible the relevant
joint. Power and signal cables pass through the whole arm and must not be damaged during the joint rotation. The
cables are axially attached to the two joints bases, the second one rotating with respect to the first one. The cable
remains loose inside the joint hollow shaft so that it can be axially twisted during the joint rotation.
A preliminary evaluation of the folding configuration and volume required for the five selected arms has been
performed. For each of them a folded configuration has been supposed and compared with the other ones. In the
evaluation both the dimension of the arm envelope and the complexity of the assumed arm pose have been considered.
The Offset Non Spherical Wrist arm is the one that best fits the volume requirements.
As far as wrist type is concerned, spherical wrist configurations are the primary choice, in accordance with the
Results of the qualitative evaluation performed on the architecture candidates are summarised in Table 3. 3 Table 2. Mechanical solution, cable routing and comparison of the joint connection types
Fork-type connection L-shaped connection Cup-type connection Table 3. Qualitative arm architecture comparison
Wrist Multi Offset,
Non Sph. Wrist Human
Sph. Wrist. Human
Non Sph. Wrist Folding ++ ++++ ++ ++ ++ Cable routing +++ +++ +++ ++ ++ Wrist Envelope ++ + + ++ + Complexity/mass
of joint-link S/S
Wrist type +++ +++ +++ ++ ++ +++ +++ +++ ++ ++ ++++ + + ++++ + Both offset type and human arm like configurations exhibit equivalent behaviour in terms of dexterity and kinematics
performance. A preference should be given to the offset configuration in terms of manufacturing complexity and cable
routing, while probably the human arm like configuration has a smaller transversal encumbrance.
Taking into account all the arm requirements, the Offset Spherical Wrist Arm has been preliminarily selected (see Fig. 1
and Fig. 6), as the best trade off among the candidates over the selected criteria.
Furthermore, if the spherical wrist requirement could be relaxed, the arm dexterity could be improved by choosing one
of the non-spherical wrist configurations.
It should be underlined that the candidate configurations should be evaluated in the frame of the specific tasks and
scenario of the user application, that currently is the EUROBOT system. Based on the evaluation results, an update of
the preliminarily selected configuration could be envisaged. 4 Fig. 6. Preliminarily selected configuration
The communication architecture of DEXARM will be based on a power bus and a digital data bus, composed of the
following nodes (see Fig. 7):
• the Robot Control Unit (RCU) not part of DEXARM itself, acting as a master; this is the central controller,
performing co-ordinated control of multiple axes;
• one slave node for each joint, with the joint electronics integrated in the mechanics of the joint (mechatronic joint);
the joint electronics is in charge of servo control and can therefore be called Servo Control Unit (SCU);
• one slave node for the End Effector
(not part of DEXARM itself); in this
way the local electronics of the End Robot Control
.... Joint 7
Effector is integrated in the End
Effector itself, allowing the optimal
digital data bus
allocation of control functions
between the local controller and the
central controller (RCU).
Fig. 7. DEXARM communication architecture
This distributed architecture, with the electronics incorporated into the joints, eliminates the need for multi-wire cable
harnesses, thereby saving system mass and increasing system reliability, improves robustness with respect to EMC,
having joint electronics located close to the sensors and the actuators, and still provides the needed data exchange
frequency, satisfying control system bandwidth requirements.
Examples of candidate data buses can be CAN, SERCOS, SpaceWire or MIL-STD-1553B.
A trade-off analysis has been conducted, indicating that CAN bus should be the preferred choice, because it gives a
greater advantage in terms of hardware simplicity (in terms of harness, only two wires are needed inside the arm and in
terms of electronics, some DSPs are equipped with CAN interface without the need of additional components) and
permits a significant mass and dimension saving with respect to the other candidates, providing a communication
bandwidth that satisfies the requirements. If the DEXARM is to be integrated in a SpaceWire based system, it is always
possible to implement a CAN/SpaceWire bridge at the base of the arm, therefore the DEXARM as a whole could be
seen as a SpaceWire node from the external system.
Manipulator dynamics and control
A number of control strategies and algorithms have been proposed in the scientific literature for controlling robot free
space and contact operations, among which:
• Implicit impedance control;
• Explicit impedance control;
• Implicit hybrid control;
• Operational space control.
In the implicit control schemes, the Cartesian level loop generates joint position setpoints, while in the explicit control
schemes, the Cartesian level loop generates joint torque setpoints. The DEXARM architecture allows the incorporation
of any of those schemes, because the joint is equipped with both position and torque sensors, allowing the selection
between joint position control and joint torque control. Mixed control schemes can also be envisaged, controlling joint
impedance, in which the Cartesian level loop generates position, torque and impedance setpoints. 5 The DEXARM control analysis is supported by full
dynamic simulation. A rigid body, flexible joint
model of DEXARM has been developed in
Modelica/DYMOLA environment, consisting of the
arm model and the controller model. The arm model
is built by assembling seven joint models with the
model of the seven d.o.f. multi-body rigid mechanical
chain (see Fig. 8).
It is worth mentioning that the DEXARM control
system will be designed to be fully compatible with
previous ESA developments, such as the Common
European Space A&R (CESAR) controller . In
fact, it will be possible to install the CESAR software
(with small adaptation such as the new kinematics)
on the RCU and use the full CESAR controller
software to execute the DEXARM performance tests
(accuracy, repeatability, etc.) and during final
demonstration in ESTEC.
Fig. 8. DEXARM multi-body chain, for free motion operations
Besides standard safety hazards (structural failures, contamination, etc.) which are generally applicable to a space
system (e.g. a payload or a facility of the space station), there are some specific hazards which are applicable to a
robotic system. The most important ones are the collision of the arm with the environment and the inadvertent release of
payloads. Of the two above hazards, only the first one is directly applicable to DEXARM, because the second hazard
depends to a great extent on the design of the end effector and will not be discussed in this paper.
When a robotic manipulator enters in contact with the environment, either in case of a planned or an unplanned contact,
there is a potential for large impact loads resulting from the large effective inertia (or more generally effective
impedance) of many robotic manipulators. If inherent safety is to be achieved, manipulators must be designed to have
naturally low impedance. Quantitatively, this requirement can be expressed in different ways. In general, for space
station manipulators, this requirements is expressed in terms of maximum allowed values for impact load, energy and/or
Unfortunately, the performance requirements often lead to designing manipulators which have high effective
impedance, they necessitate the use of high gear ratio actuators and stiff structure which increase the effective
impedance of the system.
On the other hand, the effective inertia of the system depends also on other parameters, like the overall mass and the
impact velocity, and space manipulators are not usually required to move at a high velocity. Therefore an acceptable
trade-off can be found, ensuring system safety while maintaining adequate performance.
The methods that have been preliminary identified to achieve safety with respect to the collision hazard are:
• the lightweight structure;
• the joint torque control system, ensuring that the joint output torque is always below a predefined threshold;
• the joint speed control system, ensuring that the joint speed is always below a predefined threshold;
• the Cartesian workspace control system, a geometrical collision detection check which allows to avoid arm self
collisions and collisions between the arm and the environment. Geometrical collision detection algorithms can be
implemented in the on-board computer (e.g. ).
If needed, safety checks can be performed by software, employing a computer based control system, designed to satisfy
NASA requirements . Since for a robotic system it is practically impossible to rely only on hardware devices to
perform safety checks without jeopardising the performance of the robotic operations, nearly all space station/Shuttle
manipulators and robotic payloads make use of computer based control techniques as part of their safety approach (e.g.
  ). If needed by the application, computer based control techniques can be applied also for DEXARM.
In a robot arm with a size comparable to the human arm, a large portion of the mass is allocated to the robot joints.
From this consideration, it appears that in order to accomplish the demanding resource requirements, the development
of DEXARM must first focus on the making of an optimised joint. The DEXARM joint will be a highly integrated 6 mechatronic joint, with a minimum number of structural parts, where all the components are tightly integrated together
to minimise mass and volume requirements. The joint is designed with a hollow shaft to allow central cable passing.
The preliminary joint configuration includes:
• Reduction gear;
• Bearing system;
• Sensors: motor shaft position/speed sensor, output shaft position sensor, joint torque sensor, temperature sensor;
• Electronics: motor drives, sensor acquisition, DSP to perform low level, very high frequency control loops (motor
current, motor torque, and if needed, speed/position) and joint level health check functions, data bus interface.
A preliminary evaluation of main electromechanical components and of the joint electronics has been performed and
will be briefly described hereafter.
DC brushless motors are the candidates for the application of the DEXARM joint (robotic application: closed loop
control), because they offer many advantages with respect to other solutions like brushed motors, stepper motors,
permanent-magnet stepper motors, variable reluctance stepper motors, hybrid stepper motors. A customised version of a
rare earth permanent magnet motor will be developed, in co-operation with the manufacturer. This motor can achieve a
high torque density by means of a new winding configuration and novel non laminated soft magnetic materials. The
motor is produced on an innovative stratified winding system, unlike all conventional permanent magnet motors; with
this kind of motor is possible to achieve a power to mass ratio higher than the conventional motors.
Two typologies of fail safe brakes can be considered for the DEXARM application: with contact and without contact.
Contact fail safe brakes exploit the friction action of two parts (metallic or non-metallic) pushed in contact the one with
the other. Non contact fail safe brakes exploit the detent action caused by permanent magnets closing a magnetic circuit
between a rotor and a stator. Since a DEXARM requirement is to allow the EVA to back-drive the joint ‘by hand’ if the
brake is energised by a simple dedicated circuitry (with limited mass and dimension impact), it appears that a brake
based on a classical fail safe solution can be the best approach, starting from a good existing product, possibly
customised to achieve the best trade-off between braking torque and minimum power consumption.
The gear is one of the core components of the joint, since it drives the torque capability, accuracy, size and life of the
joint. The selection of the gear type to be used in the robotic joint is based on the torque capability at output shaft
(driving aspects are gear-teeth contact stresses, teeth strength, and bearing load capacity), geometrical envelope and
compactness (gear and motor are the most voluminous parts of the joint), lifetime and duty cycle in space, reduction
ratio (which combined with motor output torque and system efficiency determine output torque for each joint: shoulder,
elbow, wrist) and thermal performance. Existing gear concepts are spur gear, planetary gear, cyclo drive and harmonic
drive. Due to the stringent torque needs of the DEXARM, a high reduction ratio is mandatory, the Harmonic Drive
technology being capable to achieve the highest reduction ratio in one single stage, as well as the highest torque/mass
and stiffness/mass ratios and the lowest backlash. The selection between the different typologies of harmonic drive
(HFUC, CSD, CSG…) is part of the work to be accomplished.
Bearings for the joint must be carefully selected since they are critical components in the interface between the rotary
and the stationary part of the joint, and can drive the maximum loads to be supported by the joint. In addition, these
elements are needed to ensure the required stiffness and strength during both launch and in-orbit environments along
with the needed lifetime. Bearings are basically required at the input and output shafts of the joint and in the wave
generator of the harmonic drive (in fact, the latter bearing is considered part of the harmonic drive). Bearings at the
output shaft must ensure a high stiffness, since they are in direct contact with the load, as well as being able to support
combined radial and axial forces and bending moments. A preloaded pair of angular contact ball bearings seems the
optimal design solution. With respect to the input shaft, bearings must support reverse thrust loads to ensure good
performance of the harmonic drive, the level of radial loads and bending moments not being significant. A four-point
contact ball bearing or a super-duplex ball bearing could perform well for the high-speed part of the joint.
For robotic applications where a control loop concept is used to control the robot output at every condition of
movement, a position sensor must be joined to the motor, to provide motor commutation and to get motor speed for the 7 joint speed control loop (to damp high frequency oscillations). Additionally, another position sensor is needed to control
the joint output position. A driving requirement for the selection of the position sensor is the angular resolution.
Moreover, the lifetime and the environmental conditions are important requirements to be considered, along with the
need for minimum mass and volume. Several encoder technologies are being considered (optical, magnetic, electrical),
together with other options such as resolvers, potentiometers or inductosyns. Torque acquisition is also required in
DEXARM joints in order to determine the actual torque being delivered by the joint. Torque sensors are used in
combination with a torque control loop so that non-linearities derived from gearing transmission (cogging, backlash,
frictions) and other elements can be compensated. The optimum design of the torque sensor will result from a
compromise between the resolution it can provide, and the stiffness of the joint and load capability, since these two
requirements are directly in opposition.
The DEXARM joint electronics is integrated in the joint itself, by means of two circular boards, accommodated within
the joint structural case. The joint servo control is achieved by means of a DSP, capable of performing position,
velocity, torque and current loop, combined as needed by the selected control strategy.
The results achieved in the initial phase of DEXARM project have been presented. The main aspects and ideas of the
current DEXARM design are:
• accommodation of servo electronics in the joint assembly and employment of mechanical design solutions resulting
in a highly integrated mechatronic joint;
• employment of kinematics solutions that maximise robot arm dexterity;
• performance assessment and control design supported by full dynamic simulation;
• safety provisions which allow to control robotic hazards as collision, while maintaining adequate performance and
preserving operational capabilities;
• compatibility with previous ESA developments, such as the CESAR controller and other existing building blocks
of space robot arm systems, in order to serve future robotics missions.
More results will be available in the course of the project, when the joint prototype will be developed and tested.
 P.H.M. Schoonejans, “EUROBOT System Requirements Document”, MSME-RQ-EB-0001-ESA, Issue 1
Revision 1, 6 April 2004.
 S. Nicolodi, D. Surdilovic, J. Schott, R. Boumans and P. Putz, “SPARCO: a space robot controller with advanced
sensor-based control capabilities derived from an industrial controller”, 2nd IARP Workshop on Robotics in
Space, Canadian Space Agency, July 1994.
 R. Mugnuolo, F. Bracciaferri, F. Didot, G. Colombina, E. Pozzi and A. Rusconi, “EUROPA (External Use of
Robotics for Payload Automation)”, 50th International Astronautical Congress, Amsterdam, The Netherlands,
 P. Laryssa, E. Lindsay, O. Layi, O. Marius, K. Nara, L. Aris and T. Ed, “International Space Station Robotics: A
Comparative Study of ERA, JEMRMS and MSS”, Proceedings of 7th ESA Workshop on Advanced Space
Technologies for Robotics and Automation - ASTRA 2002, ESTEC, Noordwijk, the Netherlands, November 2002.
 K-1207i robot description, “http://www.robotics-research.com”.
 F. Fusco and R. Gallerini, “European Robotic Arm: The Problem Of Preventing Collisions”, Proceedings of 6th
ESA Workshop on Advanced Space Technologies for Robotics and Automation - ASTRA 2000, ESTEC,
Noordwijk, the Netherlands, December 2000.
 Computer Control of Hazardous Payloads,” MA2-97-083, Space Shuttle Program Integration, 1997.
 M. Matsueda, S. Naoki, S. Takahisa, Y. Hisatome, D. Shinobu and F. Kuwao, “Safety Approach of Japanese
Experiment Module Remote Manipulator System”, 5th International Symposium on Artificial Intelligence,
Robotics and Automation in Space, ESTEC, Noordwijk, The Netherlands, June 1999.
 S. Roderick, P. Churchill, G. Gefke and D. Akin, “Design and Certification of a Total Computer-Based Hazard
Control System for a Space Shuttle Payload”, IEEE International Conference on Robotics and Automation, San
Francisco, California, USA, April 2000.
 A. Olivieri, T. Grasso, A. Rusconi and M. Colomba, “EUROPA Robot Controller: advances in design”,
Proceedings of the 7th International Symposium on Artificial Intelligence, Robotics and Automation in Space,
Nara, Japan, May 2003. 8 ...
View Full Document
This note was uploaded on 02/24/2011 for the course AEROSPACE aero520 taught by Professor Mok during the Spring '10 term at Istanbul Technical University.
- Spring '10