You are on page 1of 6

Self-assembly of Milli-scale Robotic Manipulators:

A Path to Highly Adaptive, Robust Automation


Systems
Ron Pelrine, Annjoe Wong-Foy, Allen Hsu, Brian McCoy
Robotics Laboratory
SRI International
Menlo Park, CA USA
ron.pelrine@sri.com

Abstract—Today, trends in manufacturing automation favor experiments reported here, the driver functions similarly to a
high levels of adaptability and flexibility for rapid product multi-degree-of-freedom (DOF) linear stepper motor.
change, customization, and new process integration. This paper
looks at a new technology, diamagnetic micro manipulation
(DM3), for taking adaptability of automation to a higher level.
Experiments reported here show a path to automation systems
that can self-assemble their own robotic manipulators, fabricate
tools-on-demand to execute new processes, optimize existing
processes, and/or replace or repair tools. These printed circuit
board (PCB) systems can control very large numbers of milli-
scale robots. Hence, as described in this paper and in contrast to
conventional robotic systems, only a small percentage of DM3
system resources, approximately 5-10% of system area and robot
numbers, are needed to achieve on-board self-assembly and tool-
on-demand functions. This paper describes the key step of self-
assembly of self-levitating magnet arrays using 1 mm x 1 mm x
0.4 mm magnets, with self-assembly times of less than 10 s. We
also report experiments showing how milli-robots can build
simple end effector tools (a straight probe tip, a forked tip, and a
hook tip) on other milli-robots for on-board, tool-on-demand
functions.

Keywords—levitated robots, adaptive systems, diamagnetism,


self-assembly, self-repair, robot replication
Fig.1. Structure of a DM3 system. A robot consisting of a magnet array and
I. INTRODUCTION end effector self-levitates over diamagnetic graphite (top left). Driver PCB uses
two layers per DOF with quadrature drive and an eddy current damping layer
Highly adaptive, automated manufacturing systems that can (bottom left). Operation on a flex circuit which provides 3D and interconnect
implement rapid product and process changes and adjust for capability, enabling the systems to be modular (top right).
wear and other damage while maintaining optimal, or at least
acceptable, performance are highly desirable. One approach to
Milli- and micro-robots using permanent magnets have
achieving both goals is a mechanical system that can make and
long been an area of investigation, and drives using printed
implement its own mechanical parts within the same system.
circuits have been previously reported as well [1-5]. However,
This paper presents recent results of in-system replication a new aspect of DM3 systems is that robot precision, control,
of manipulators using a new milli (1-10 mm) robot technology and reliability is greatly enhanced by the diamagnetic layer.
known as diamagnetic micro manipulation (DM3). In this The increase in reliability enables large numbers of small robot
paper, manipulators refer to all moving parts of the robot, manipulators (> 1000) and large areas (> 0.1 m2) to be operated
including any on-board sensors and electronics. DM3 systems as a single integrated system [6]. This observation is relevant
were originally developed for manufacturing, using the bio- because it sets the typical system size for considering tool-on-
inspired approach of ant-size, controllable robots in large demand, self-repair, and related replication functions. While
numbers for massively parallel manufacturing applications [1]. these functions are obviously of system value, they are rarely
The name DM3 derives from the use of diamagnetic graphite possible with macro automation because providing the
to reduce or eliminate friction and wear between the moving functions would overwhelm the system (analogous to including
manipulator and the workspace shown in Fig. 1. In the an entire macro robot-making factory with each robot system).
In contrast, based on the experiments described here, typical
DM3 systems existing today can include self-repair and tool- [1], and performing other automated process functions [25]
on-demand functions using as little as 5-10% of system establishes the potential for DM3 systems to perform both the
resources in area, power, and number of robots. self-replication steps in Fig. 2, from input of single magnets to
self-construction, and the implementation of highly complex
Simple magnet array patterns can be magnetically self- and capable manipulators.
assembled on fast (< 10 s) time scales using a DM3 system,
and, as indicated by Fig. 2, these magnetic arrays are robot
manipulators in the DM3 system. Although robotic self- II. PRINCIPLE OF OPERATION
assembly processes, including magnetic self-assembly, have Current DM3 systems use arrays of millimeter-size
been demonstrated using motorized components [7-19], magnets to form the mobile unit, which is part or all of the
examining replication processes has revealed that motorized robot manipulator depending on whether tools and other
components tend to have longer and more difficult replication devices are added. The magnet array is driven magnetically by
process sequences (Fig. 2) than DM3 systems. DM3s belong to a printed circuit board (PCB) or a flex circuit (Fig. 1). The
a larger class of robots that might be called open-bearing PCB acts as fixed stator. Electrical currents, typically on the
(unconstrained in multiple degrees of freedom (DOF), wave- order of 0.1 – 1.0 A, are driven through the traces in the circuit
driven robots. Other examples of open bearing, wave-driven to generate local magnetic fields and field gradients. The
robots include acoustic and photonic (laser) levitation of rigid resulting magnetic forces and torques accelerate the magnet
bodies with multi-DOF control [20, 21]; electrostatic drives of array, typically in the range of 0.1 – 10 g’s (1-100 m/s2).
rigid bodies on air bearings [22]; and others [23]. Open
bearing, wave-driven machines enable relatively unstructured Serpentine traces in the PCB or flex circuit are used to
rigid bodies of single materials to operate as robotic generate a repeating magnetic wave pattern that pulls or pushes
manipulators. For example, a hypothetical manipulator the permanent magnetic array in multiple DOF. As previously
replication system could use acoustic waves to manipulate an reported [1, 25], four layers of serpentine copper traces in a
unstructured particle to cut or chip off another unstructured PCB are used to make a two-DOF (x,y) or three-DOF (x,y,z)
particle from a block of material. If the second unstructured controllable zone in a typical configuration, although other
particle could also be manipulated in three programmable patterns for rotation and tilt can be used. In essence, the magnet
DOF, it would then qualify as a robotic manipulator in this array forms the moving part of a multi-DOF linear motor.
hypothetical system. In the actual experiments reported here, Although such motors are not new, the use of a PCB for the
the input is not as simple as a block of uniform material, but it stator and the size scale are not common, primarily due to the
is still fairly simple, using small, 1 x 1 x 0.4 mm magnets. difficulties of reliably controlling friction, surface adhesion,
and wear on smaller scales [26-29].
Becomes a
robot here DM3 systems address the surface contact problems on
smaller scales using a layer of diamagnetic material such as
Motor Multi-DOF pyrolytic graphite [30, 31]. The diamagnetic repulsive force
Ore Magnetic Unmagnetized Single Magnet Bearings & Motor Electrical
Processing Material Magnets Magnets Arrays Shaft
Motor
Coils Connections Assembly
between the diamagnetic layer and the magnet array pushes up
Complex Manipulators
Ore (sensors, on the array, reducing both contact pressure and area [32, 33].
Self-assembly Building with
communication, etc.)
For diamagnetically enhanced sliding and micro-hopping
Robots
Ore Magnetic Unmagnetized Single Magnet systems, the diamagnetic force reduces wear and surface
Processing Material Magnets Magnets Arrays Complex Manipulators adhesion and somewhat improves open-loop precision. On
Ore (sensors,
communication, etc.) small scales, however, the diamagnetic force is strong enough
Becomes a
robot here
that magnet arrays will self-levitate [34] with zero surface
contact above the graphite (Fig. 1), completely eliminating
Fig. 2. Conceptual comparison of replication process sequences for surface contact problems of the manipulator drive. Micro-robot
conventional motor-driven robot manipulators that carry their own motors operation in this regime has zero wear, zero surface adhesion,
(top), and for DM3 manipulators (bottom). Conventional, motor-driven and orders of magnitude improved open-loop precision at
manipulators have longer, more complex process sequences to reach robotic submicron levels.
manipulator capabilities.
A key aspect of these DM3 systems is that they can drive
Although the open-bearing, wave-driven approach enables and manipulate virtually any alternate polarity (checkerboard-
simple input materials to reach robotic motion capabilities, type) magnetic array of the appropriate pitch placed on the
much of the interest in machine self-replication for system system’s surface. The systems are typically divided into
adaptability is in the self-replication of complex machines that independently controlled areas or zones so that one
can perform specific tasks, rather than just simple replicating manipulator can be controlled independently of another
particles. Making more complex machines typically requires a manipulator in another area. Control zones can also control
variety of differentiated tools (end effectors) on the different numbers of DOF with different trace patterns, and
manipulator to automate other processes. A second experiment some zones can fully levitate manipulators while others with
shows a milli-robot building three simple, but different tools thinner graphite are just diamagnetically enhanced to reduce
(straight probe, hook, and fork) on three different DM3 contact pressure but not eliminate it entirely. In most systems,
manipulators using a carbon suspension material. These results the manipulators are mobile and can move freely between
along with other data showing DM3s building integrated different zones.
electronic and mechanical structures [6], testing [24], weighing
The PCB used for the self-assembly experiments described magnets are driven to the incubator, then down a square tube
below used 1-mm center-to-center trace spacing. The system with one open and one closed end formed by the roof of the
uses a quadrature drive in each of two DOF for the self- incubator, the PCB floor, and side spacers (black rods visible
assembly process. For simplicity, the current in each trace is in the upper photos of Fig. 3).
always fully on, and only the current direction is switched to
move the robot. If +1 denotes positive current polarity for the
trace, and -1 denotes negative polarity, then the control pattern
to move in the positive x-direction is (1,1), (1,-1), (-1,-1), (-1,1)
where (x1, x2) denotes the current polarity in the two traces, x1
and x2, that control x-direction motion. A similar drive is used
to control two traces that move the manipulator in the y-
direction. With simple stepper control, the system is a multi-
DOF linear stepper motor. However, because the top layers of
the PCB are closer to the manipulator than deeper layers, the
absolute value of the current for each layer is different. For the
self-assembly system described here, the current amplitudes
were set at 0.27 A, 0.33 A, 0.5 A, and 0.75 A for layers 1 (top),
2, 3, and 4, respectively.
As noted earlier, DM3 systems use diamagnetic graphite to
reduce or eliminate adverse effects from surface contact.
Panasonic graphite film (EYG-A121807M) measuring roughly
70 µm thick with adhesive backing was used in the magnet
self-assembly experiment. The end-effector process described
below was an earlier experiment and used the same PCB
control but with a thicker, 0.5-mm pyrolytic graphite. DM3
systems can also use conductive metal layers to increase eddy
current damping and improve control. The system for Fig. 3. Dashed line shows motion of single 1.4 x 1.4 x 0.4 mm magnet into an
fabricating the three end effectors described below used an incubator structure (upper left). Additional magnets are driven into the
incubator to make an Nx1 array, which can then exit the incubator (upper
18-µm thick aluminum layer on top; the experimental self- right). Freestyle self-assembly using cylindrical magnets with only the top of
assembly system omitted the conductive layer and used sliding the incubator used as a glass confining structure (lower left). More complex
friction for damping. Although the two systems were arrays can be made with square magnets using this technique, e.g., five self-
configured slightly differently, this difference is not considered assembled manipulators (lower right). See video [35].
significant, and both systems can execute both processes.
The top of the tube prevents the magnets from flipping over
III. EXPERIMENTS on top of each other, a lower magnetic energy state. Instead,
because the side-by-side checkerboard pattern of polarities is
A. Self-Assembly of Magnet Array Manipulators magnetically stable (just not at the global lowest energy state),
as individual magnets are fed in, they form a 1 x N
Individual magnets were placed on the PCB in the desired checkerboard magnet array, where N is the number of
magnetic orientation. Three types of magnets were tested alternating polarity magnets fed into the system. N is
separately with similar results, except as noted below. One set theoretically limited only by the length of the incubator
of magnets used 1.4 x 1.4 x 0.4 mm neodymium-iron magnets structure. We tested up to N=4.
grade 50M. The second test used 1 x 1 x 0.4 mm neodymium-
iron magnets grade 52, and a third test used cylindrical grade For N=3, the current microfactory system can build further
52 magnets with 1-mm diameter and 0.5-mm thickness. With magnets onto the array in a freestyle fashion. That is, a 1 x 3
sufficiently thick graphite and the proper magnetic design, all array is held magnetically by the DM3 system underneath a
three types can diamagnetically self-levitate, and self-levitation roof with no other confining structure. In our experiment, the
was confirmed using self-assembled 2x2 arrays of 1 x 1 x 0.4 roof was a glass plate held 0.5 mm above the PCB driver board
mm magnets. by spacers, giving a nominal 0.1 mm clearance using 0.4 mm
thick magnets. The system can then bring in other magnets to
As shown in Fig. 3, the PCB traces drive the magnets into a magnetically lock in desired locations on the array by holding
mechanical structure (incubator) for self-assembly. The the array fixed with one set of PCB traces, while another set of
incubator is a fixed part of the system structure and, from the nearby, independent traces brings in other magnets. Because
robot’s point of view, part of the environment. Magnets in the magnetic lock can be quite violent mechanically with high
these experiments were manually fed into the system and strength magnets, the roof prevents magnet flipping, and the
placed 2.5-10 cm away from the incubator on the system’s 1x3 array starts the process with greater inertia and holding
surface. A future design for adaptive manufacturing envisions power than a single magnet can provide at current system
an automated feed from a magnet reservoir or source. power levels.
The incubator provides a series of mechanical stops so that Freestyle building with the cylindrical magnets was more
the magnet(s) do not have to be precisely controlled. The limited in that the magnets would often rotate away from a
desired location to form 1-dimensional chains as seen in the shown in Fig. 4 (lower right photo), may use acetone or other
lower left photo of Fig. 3. A variety of array geometries can be aggressive solvents for end-effector cleaning. This part used
self-assembled using these techniques, such as the V-shaped roughly 1000 droplets of UV adhesive, and further validates
arrays in Fig. 3, lower left. Regular arrays, such as 2x2 and the feasibility of making complex parts by micro robotic 3D
3x3, can also be made with an incubator structure by printing. The end-effector building experiment in Fig. 4 used
combining 2x1 and 3x1 arrays, respectively. Most experiments roughly 50 cm2 of PCB (robot driver) area.
used 2.4 cm spacing between feed magnets, with speeds of
approximately 3-6 cm/s. The 5-magnet V-shaped manipulators
in Fig. 3 took 6 s to assemble. Hence, depending on the size of
the array, manipulators can be assembled in less than 10 s. As
with any checkerboard magnet array of suitable pitch, the
arrays can exit the self-assembly area and be driven throughout
the DM3 system, such as to locations where end effectors and
other features (electronics, sensors, etc.) can be added.
Performance of the self-assembled arrays as robotic
manipulators has not yet been tested extensively, but
qualitatively they seem to perform similar to hand-assembled
arrays, with 5-10 cm/s operating speeds in 3 DOF
demonstrated. Most self-assembly processes we tested used
small areas, about 25 cm2, including areas from magnet input
to self-assembled manipulator exit.
Fig. 4. DM3 robot (left) building three different end effectors (upper right).
B. End-Effector Experiments: Robots Building Robots Example of a plastic part built by the robot manipulator shown with drop
deposition (lower right). See video [35]
As noted earlier, much of the interest in robot replication is
motivated by the need to replicate complex and/or highly The structures in Fig. 4 are crude (low precision) structures
capable or adaptable machines. This is shown as the last because these early experiments didn’t use feedback other than
conceptual process step in Fig. 1. The complexity may require manually fixing an occasional motion error. However, with
structural complexity to perform some task, but may also fixed sensors providing feedback (fixed and therefore not part
include electronics, sensors, and other components. Robots are of the manipulator replication), it should be possible to make
well-suited for making complex structures because of their structures with resolutions approximating the motion resolution
programmable, multi-DOF motion. However, in addition to of the DM3 systems, currently on the order of 25-50 µm for a
controllable motion, robot manipulators need various tools for diamagnetically enhanced (sliding) system, and on the order of
different types of building tasks. Ordinarily, tools are added to 0.1-0.5 µm for a diamagnetically levitated system [1, 25] .
robots based on the desired task, but for self-replication, adding
a tool changes the task because now the added tool, which End effectors, not much more complex than the robot-made
becomes part of the robot’s manipulator, must be replicated as ones in Fig. 4, using flat wetting pads have been demonstrated
well. building with electronics, carbon rods, fiberglass, and other
materials using DM3 magnet arrays [6]. Typically, surface
The problem is how to go from simple to complex without tension from water and other liquids on a wet structure (e.g., a
making the tools too difficult to replicate. One attractive option loop or pad) can be used to pick and place small parts,
is to use the micro robots to execute a 3D printing process. A eliminating the need to replicate complex, active gripping
robot with a simple rod or a rod with a loop at the end picks up mechanisms. If these end effectors for building with other
a droplet of curable or self-curing liquid, such as evaporating materials were made from a 3D micro robotic printing process
suspensions, UV curable adhesive, or meltable materials. The on self-assembled magnet arrays, the system could then,
robot controls the rod to touch the liquid onto a building theoretically, bootstrap to self-replicating very complex
surface, such as for building end effector(s) on magnet arrays manipulators, starting with a few simple materials.
or other structures, as seen in Fig. 4. Repeated liquid deposition
and curing from different directions builds up a 3D part. The IV. ANALYSIS
building robot(s) can be a “seed” robot made using other
methods (by hand in this case), or, in a future version of the The self-assembly process for in-system fabrication of
technology, it could be a robot made earlier by other robots. magnet arrays works well in testing to date. Based on
preliminary experiments, it may be possible to eliminate part or
For example, the three different end effectors, probe, hook, all of the incubator structure, particularly using higher power
and Y-fork, in Fig. 4 were made by another micro robot using and/or system design to maintain control of the magnet-to-
Wire Gluetm, a carbon-based suspension that cures or dries to a magnet impact. However, the incubator is a simple, relatively
conductive solid. Because curable liquids typically tend to compact structure and may offer more robust control over a
adversely build up on end effectors and degrade or even stop variety of conditions.
the building process, an important part of these processes is a
cleaning reservoir for the robot to access (dip into) periodically The 3D micro robotic printing process demonstrated by the
for cleaning. For Wire Gluetm, we used a warm water cleaning fabrication of straight, hook, and fork end effectors (and other
reservoir, but other builds, such as the plastic UV-cured part parts such as the one in Fig. 3) is a promising path to in-system
fabrication of complex manipulators for highly adaptive systems [2] S. Chowdhury, W. Jing, D. Cappelleri, “Towards independent control of
and tool-on-demand. These structures are still fairly crude and multiple magnetic microrobots,” Micromachines, Vol. 7, Issue 1, Article
3, 2016.
built open loop. It may be possible to build with precision
[3] S. Floyd, C. Pawashe, and M. Sitti, Two-dimensional contact and non-
using some purely open-loop processes, but higher precision contact micro-manipulation in liquid using an untethered mobile
might be more easily obtained using feedback from micro magnetic micro-robot, IEEE Transactions on Robotics, Vol. 25,
robot-compatible sensors, such as photo-interrupters for robot pp. 1332-1342, 2009.
tactile sensing, to detect the part surface as it is being built. [4] E. B. Steager, M. S. Sakar, C. Magee, M. Kennedy, A. Cowley, and
V. Kumar, “Automated biomanipulation of single cells using magnetic
microrobots,” International Journal of Robotics Research, Vol. 32, No.
3, p346–359, 2013.
[5] R. Pelrine, “Maglev microrobotics: an approach toward highly
integrated small scale manufacturing systems,” Proceedings of the
Electronic Manufacturing Technology Symposium, Seventh
IEEE/CHMT, pp. 273–276, 1989.
[6] SRI International, “Magnetically actuated micro-robots for advanced
manufacturing applications,” 2015.
https://www.youtube.com/watch?v=Bxb3-bT8uxk; see also,
“Magnetically actuated micro-robots for advanced manipulation
applications,” 2014. https://www.youtube.com/watch?v=uL6e3co4Qqc
[7] K. Lee and G. S. Chirikjian, “Robotic self-replication,” IEEE Robot.
Autom. Mag., Vol. 14, 2007.
Fig. 5. Example configuration using a flex circuit or other DM3 3D motion
[8] V. Zykov, E. Mytilinaios, B. Adams, and H. Lipson, “Robotics: Self-
capabilities to allow complete access for building 3D structures (end
reproducing machines,” Nature, Vol. 435, pp. 163–164, 2005.
effectors) on magnet arrays. An alternate configuration is for the building
robot to come from above the magnet array, an orientation and process similar [9] M. Yim, et al., “Modular self-reconfigurable robot systems [Grand
to some conventional drop-deposition 3D printers. Challenges of Robotics],” IEEE Robot. Autom. Mag., Vol. 14, pp. 43–
52, 2007.
Another limitation of the experimental end effector [10] D. Rus, “Self-reconfiguring robots,” IEEE Intell. Syst. Their Appl.,
Vol. 13, pp. 2–4, 1998.
fabrication method reported here is that it can only build in one
plane. Fortunately, DM3 systems can access the part being [11] D. Rus and M. Vona, “Crystalline Robots: Self-reconfiguration with
compressible unit modules,” Auton. Robots, Vol. 10, pp. 107–124,
built in 3D using a variety of configurations, such as the one 2001.
shown in Fig. 5. [12] V. Zykov, E. Mytilinaios, M. Desnoyer, and H. Lipson, “Evolved and
designed self-reproducing modular robotics,” IEEE Trans. Robot., Vol.
Although a first step with significant limitations, these 23, pp. 308–319, 2007.
experiments illustrate that DM3 adaptive capabilities using
[13] S. Griffith, D. Goldwater, and J. M. Jacobson, “Robotics: Self-
replication can be implemented using only a small portion of replication from random parts,” Nature, Vol. 437, pp. 636–636, 2005.
system resources. The end-effector building experiment used [14] M. Sipper, “Fifty years of research on self-replication: an overview,”
roughly 50 cm2 of PCB surface. The incubator itself uses only Artif. Life, Vol. 4, pp. 237–257, 1998.
about 10 cm2 of PCB surface area, and even allowing [15] R. Gross, E. Tuci, M. Dorigo, M. Bonani, and F. Mondada, “Object
additional area for automatic feed, the self-assembly could be transport by modular robots that self-assemble,” Proceedings 2006 IEEE
approximately 50 cm2 or less. By comparison, DM3 systems International Conference on Robotics and Automation, ICRA 2006,
having 1000 cm2 or more usable surface area are fabricated pp. 2558–2564, 2006. doi:10.1109/ROBOT.2006.1642087
today using commercial PCB technology. Hence, it is easy to [16] K. Lee, M. Moses, and G. S. Chirikjian, “Robotic self-replication in
structured environments: physical demonstrations and complexity
imagine DM3 systems using replication and in-system measures,” Int. J. Robot. Res., Vol. 27, pp. 387–401, 2008.
fabrication functions at a minor system cost compared to the
[17] R. Pfeifer, M. Lungarella, and F. Iida, “Self-organization, embodiment,
potential benefits in adaptability and robustness. and biologically inspired robotics,” Science, Vol. 318, pp. 1088–1093,
2007.
V. SUMMARY AND CONCLUSION [18] J. Romanishin, K. Gilpin, and D. Rus, “M-blocks: Momentum-driven,
magnetic modular robots,” in 2013 IEEE/RSJ International Conference
Although DM3 systems are new, and their ultimate on Intelligent Robots and Systems (IROS), pp. 4288–4295, 2013.
capabilities are presently unknown, these early experiments in doi:10.1109/IROS.2013.6696971
self-assembly and robot-building-robots indicate that DM3 [19] S. M. Felton, M. T. Tolley, C. D. Onal, D. Rus, and R. J. Wood, “Robot
systems are capable of functions such as tool-on-demand and self-assembly by folding: A printed inchworm robot,” in 2013 IEEE
robust self-repair. And, unlike conventional macro automation, International Conference on Robotics and Automation (ICRA) pp. 277–
282, 2013. doi:10.1109/ICRA.2013.6630588
such functions in DM3 systems can be implemented with an
[20] Y. Ochiai, T. Hoshi, and J. Rekimoto, “Three-dimensional mid-air
attractive tradeoff in system cost (size, power, etc.) versus the acoustic manipulation by ultrasonic phased arrays,” PLoS ONE, Vol. 9,
potential benefits. e97590, 2014.
[21] A. Ashkin, “Acceleration and trapping of particles by radiation
REFERENCES pressure,” Phys. Rev. Lett., Vol. 24, pp. 156–159, 1970.
[22] R. S. Fearing, “A planar milli-robot system on an air bearing,” Robotics
[1] R. Pelrine, A. Wong-Foy, A., B. McCoy, D. Holeman, R. Mahoney, G. Research (eds. Giralt, G. & Hirzinger, G.), pp. 570–581 (Springer
Myers, J. Herson, and T. Low, “Diamagnetically levitated robots: An London, 1996. http://link.springer.com/10.1007/978-1-4471-0765-1_64
approach to massively parallel robotic systems with unusual motion
properties,” in Proceedings IEEE International Conference on Robotics [23] E. Diller, C. Pawashe, S. Floyd, and M. Sitti, Assembly and disassembly
and Automation (ICRA), pp. 739-744, 2012. of magnetic mobile micro-robots towards deterministic 2-D
reconfigurable micro-systems, Int. J. Robot. Res., 0278364911416140, 1997.
2011. doi:10.1177/0278364911416140 http://scitation.aip.org/content/avs/journal/jvstb/15/1/10.1116/1.589247
[24] D. L. Christensen, E. W. A. Hawkes, A. Wong-Foy, R. Pelrine, and M. [30] D. B. Fischbach, “Diamagnetic susceptibility of pyrolytic graphite,”
R. Cutkosky, Incremental inspection for microrobotic quality assurance, Phys. Rev., Vol. 123, pp. 1613–1614, 1961.
V001T09A030, 2013. doi:10.1115/DETC2013-12790 [31] R. D. Waldron, “Diamagnetic levitation using pyrolytic graphite,” Rev.
[25] R. Pelrine, A. Wong-Foy, A., B. McCoy, D. Holeman, R. Mahoney, Sci. Instrum., Vol. 37, pp. 29–35, 1966.
“Micro robot manufacturing,” MFG 2011 Workshop, August 8-10, [32] W. Braunbeck, “Free suspension of bodies in electric and magnetic
2011, Napa, California, USA. http://www.transducer-research- fields,” Zeitschrift für Physik, Vol. 112, pp. 753–763, 1939.
foundation.org/archive/mfg2011/
[33] R. Pelrine, “Room temperature, open-loop levitation of microdevices
[26] N. Tas, T. Sonnenberg, H. Jansen, R. Legtenberg, and M. Elwenspoek, using diamagnetic materials,” in IEEE Micro Electro Mechanical
“Stiction in surface micromachining.” J. Micromechanics Systems, 1990, Proceedings, An investigation of micro structures,
Microengineering, Vol. 6, pp. 385, 1996. sensors, actuators, machines and robots, pp. 34–37, 1990.
[27] J. M. Bustillo, R. T. Howe, and R. S. Muller, “Surface micromachining doi:10.1109/MEMSYS.1990.110242
for microelectromechanical systems.” Proc. IEEE, Vol. 86, pp. 1552– [34] R. Pelrine, “Magnetic field levitation,” U.S. Patent 5,396,136, 1995.
1574, 1998.
[35] SRI International, “SRI Micro Robots Self Assesmbly and Tools Fab,”
[28] C. H. Mastrangelo, “Adhesion-related failure mechanisms in 2016.
micromechanical devices.” Tribol. Lett., Vol. 3, pp. 223–238, 1997. https://www.youtube.com/watch?v=0JNQW34P74o&feature=youtu.be
[29] R. Maboudian and R. T. Howe, “Critical Review: Adhesion in surface
micromechanical structures,” J. Vac. Sci. Technol. B, Vol. 15, pp. 1,

You might also like