(a)

(b) (c)

Figure 1. (a) Schematic of the micro HTS system. (b) One layer of micro

fabricated stator. (c) Rotation of the levitated rotor with micro fabricated

stator above HTS [6].

Abstract— This paper describes the progress of small-scale

robots at nano, micro, milli meter scale and beyond since the

MEMS technology emerged 30 years ago. It shows some

example systems at each scale to shed light on how to harness

the advantages of small-scale robots. Then, it discusses the

emergence of soft robots. It presents small-scale soft robot

systems to show their promise and limitations to suggest a

future direction. Finally, it proposes a few flowing liquid robot

systems utilizing interesting phenomena of the switchable

surface tension of liquid metal.

It concludes that to realize the full potential of nano, micro,

milli robots, more efforts should be directed toward the

integration of the multi-scale systems together. Furthermore,

the intricate interplay of physics and chemistry in small scales

suggests that robots should be developed not only in various

scales, but also in all three states of matter; solids, liquids, and

gases.

I. INTRODUCTION

Following the predictions of the seminal paper by

Feynman in 1960, Micro Electro Mechanical Systems

(MEMS) technology triggered the birth of tiny micro-robots

in the late 1980s [1],[2]. About a decade later when many

MEMS sensors and actuators are in the market, the

discovery of the bulky ball ushered in the era of

Nanotechnology. Over the next decade, our understanding of

biological, physical systems at nano scale advanced at a

rapid pace, producing a variety of nano systems in use, such

as diagnosis and drug delivery. Yet another decade later,

small soft robots emerged to make robots work inside the

human body and with other human and environments

individually or with other units. One such system is a

dynamic system comprising millimeter to centimeter scale

gears that self-assemble into a simple machine at a fluid/air

interface [3]. They are driven by magnetic interactions.

Soft robots use both its morphology and material

property for locomotion, integration with other modules, and

manipulation to perform tasks. We can develop soft robots,

which are functionalized, first, by equipping them with a set

of devices necessary for a particular task, second, by

inducing a certain function, for example, increased surface

friction or affinity to water or surface tension, through

material reassembly at nano/micro scales.

We propose to develop techniques not only to actuate

the motion of the surface or body to create configuration

changes at the macro scale but also to change the topology

through chemical/physical interaction of materials at

nano/micro scales. In particular, we aim to exploit

interesting phenomena of the switchable surface tension of

E. Lee is with CROASAEN (Create Robots and Save Energy), Inc,

Cambridge, MA 02138 USA (281-757-6954; e-mail:eunjeonglee@ieee.org.

liquid metal [4]. Recently, it has been shown that the

interfacial tension of a liquid alloy of gallium can be

controlled via surface oxidation using a very low voltage. It

enables shape-reconfigurable metallic components

reversibly. This newly identified electrohydrodynamic

phenomenon can push the frontiers of soft robotics by

making robots evolve and flow to new shapes and locations.

It will usher us into the new era of flowing liquid robots. We

propose a few liquid robot systems: (1) slime mold robots,

(2) variable-stiffness robots, grippers, actuators, and (3)

further applications [5].

II. AN INTEGRATED MICRO

HIGH-TEMPERATURE SUPERCONDUCTING SYSTEM

FOR ENERGY STORAGE AND ATTITUDE CONTROL OF THREE-

AXIS STABILIZED NANOSATELLITES

This section presents an integrated micro high-

temperature superconductor system for energy storage and

attitude control of three-axis stabilized nano satellites in

figure 1 [6]. It has been developed using micro

electromechanical systems (MEMS) fabrication technology.

It is an integrated high-temperature superconductor (HTS) -

magnet bearing system with a motor/generator for energy

storage, power generation and attitude control for spacecraft

such as nano/pico satellites. This system is based on passive

magnetic levitation and the flux pinning effect of HTS. This

new system constitutes the first attempt to utilize these two

properties simultaneously in MEMS, while the capacity of

Putting Them All Together and in All States

E. Lee, Member, IEEE

high-temperature superconductors to carry currents without

resistance has been widely used in micro electronics and

imaging devices. The rationale lies in the unique capability

of the HTS to adapt to low temperatures, radiation, and

vacuum environments in space, and to enhance system

stability passively without power consumption.

The micro HTS system consists of a flywheel/rotor,

motor/generator, motor electronics, and a cooling system.

The flywheel/rotor has been fabricated by using sintered

NdFeB, and the stator for motor/generator has been

fabricated by micro fabrication technology. An alternative

stator has been fabricated by cutting a 50 micron-thick

copper film for comparison. A servo amplifier to drive the

DC brushless motor of the integrated HTS system has been

developed and successfully tested. A cooling system has

been developed to test the system.

The flywheel/rotor is made of aluminum and has a mass

of 12.76 g with an angular momentum capacity of 2.22 mJ

sec at a constant speed of 12 000 rpm using the micro

fabricated multi-layer stator. It far exceeds that of the other

stator made without using MEMS fabrication technology.

Experiments also show that the micro fabricated stator can

significantly reduce the motor/generator losses. The small

rotating HTS flywheels of about 10 grams can be used for

energy storage and attitude control of far larger systems,

nanosatellites of around 10kg.

In the future, the integrated micro HTS system for

energy storage and attitude control can be an enabling

technology for some yet undefined NASA missions and

commercial satellites.

III. CARBON NANOTUBE-TIPPED MICROCANTILEVER

ARRAYS FOR IMAGING, SENSING, AND 3D

NANOMANIPULATION

Since their discovery, carbon nanotubes ( CNT ) have

been recognized as important building blocks for nanoscopic

systems owing to their excellent electromechanical

properties [7],[8]. Particularly, their potential as probe tip for

high-resolution atomic force microscopy has been

emphasized for reasons, such as high wear resistance, high

length-to-diameter ratio, and high bending flexibility that

leads to minimal damage to samples [9],[10]. Also, the

unique potential of carbon nanotube tips for biological

imaging, sensing, and manipulation at the nanometer scale

has drawn the attention of life science community [11-15].

Figure 2 shows the schematic of the massively parallel

microcantilever arrays with multi-walled carbon nanotube

tips [16]. The integrated system can make a powerful tool

for high-speed imaging, sensing and 3D nanomanipulation

of nanoparticles and biological samples. The microcantilever

has a multi-walled carbon nanotube tip and four additional

carbon nanotubes for 3D fine manipulation by electrostatic

forces. The reflected light from the deflected microcantilever

is collected by a position sensitive photodetector, and fast

readout is achieved by a time-multiplexing scheme. A

distributed parameter system model has been developed to

study its dynamic behavior. Simulations have been

performed for the carbon nanotube tipped microcantilevers

of three different dimensions to investigate their open-loop

and closed-loop performances. It is shown via simulations

that with carefully selected dimensions they can demonstrate

an excellent capability for nanomanipulation of samples and

tapping mode operation for imaging under a simple PID

controller.

The proposed system will provide exciting tools for

studying a wide variety of phenomena of nanoparticles and

biological systems at the micro/nano scale, such as elasticity,

surface charge, biomolecular binding, adhesion, wear,

friction, spatial orientation, and single molecule

conformation [17]. By functionalizing the CNT tips, they

can also be used for chemical sensing [11].

This is an example system to show how the small-scale

robot systems integrated with larger systems work together

to produce outputs. The nano system, carbon nanotube tip,

is connected to the micro system, microcantilever, which is

then attached to the centimeter XYZ stage. The micro

piezoactuators in the positioning stage actuate the

Figure 2. (a) Schematic of the microcantilever with multi-walled carbon nanotubes (MWNT), (b) microcantilever arrays with an electrostatically

deflected MWNT tip, and (c) configuration of 5 MWNTs [16].

Figure 1. (a) Schematic of the microcantilever with multi-walled carbon nanotubes (MWNT), (b) microcantilever arrays with an electrostatically

deflected MWNT tip, and (c) configuration of 5 MWNTs.

microcantilever. The micro sensor arrays send signals back

to the controller in the main computer.

IV. SOFT AND AUTONOMOUS BIOMIMETIC MICRO-

ROBOTIC FISH

A flexible micro robotic fish shown in figure 3 is

designed to take advantage of the energy-saving feature of

fish [18]. The body of the fish is made of soft material with

embedded electromagnets. It has a flexure link made of

polyimide. Carbon-fiber reinforced composite material

sandwiches the flexible link at four locations, allowing the

four rigid links to be connected by a flexure joint. Nd-Fe-B

magnets and planar coils are secured to the top of the

composite material. The entire structure, including Li-ion

battery and electronics, is embedded within

polydimethylsiloxane (PDMS) sheets. The propagating

muscle activity pattern is achieved by activating the

electromagnets that correspond to each muscle group. The

fish has polyvinylidene-fluoride-trifluoroethylene (PVDF-

TrFE) piezoelectric polymer sensors on its surface to

measure the local flow pressure. When it senses vortices,

the fish automatically turns off the actuator so that its body

wave can be passively generated by interaction with the

oncoming flow of water. The power generated for positive

work can be stored as elastic energy and then used for

negative work, thus saving energy. The system is driven at

its natural frequency to minimize energy consumption.

Future generations of this micro robotic fish could be

deployed for underwater structure and ship inspection and

environmental monitoring.

V. ORIGAMI WORM ROBOT: FABRICATION AND ACTUATION

OF A SOFT ROBOT WITH DRAMATIC MORPHING

Previous biological studies, as well as previous work on compliant robotic systems, have demonstrated that the mechanical properties of the outer surfaces of mobile systems can contribute to effective locomotion and configuration changes. This section presents an origami worm robot with one of such biologically-inspired shapes using soft materials shown in Figure 4 (a) [19].

Analysis on folding-based expansion has resulted in a

generalized algorithm for producing a large expansion. The

topological model for shape-changing and movement yields a

negative Poisson ratio tube. The radial expansion is coupled

to the longitudinal elongation by the crossed creases and

scales with the number N of folds along the circumference.

We can obtain any value of expansion ratio we want by

adjusting N. This concept is turned into a Solidworks model

and printed on a 3D printer to make molds for the origami

structure. They are then used for silicone molding of the

origami skin.

To develop a soft-bodied robot capable of locomotion

and shape changing, one of the most valuable capabilities is

the spatial control of variable compliance and topologies in a

soft material. We developed a method by which we can

pattern a soft (elastomeric) material to selectively rigidify

portions while simultaneously controlling the 3D surface

topology. Photoactive polymerization selectively embeds

rigid plastic in elastomers. A monomer and photoactive

polymerization chemistry are diffused into the elastomer,

then exposed to UV through a template. Then the unexposed

chemistry is flushed out, leaving the embedded plastic

element. It rigidifies only the exposed flat faces and leaves

the unexposed edges soft. This method can even be used to

create transitions via a gradient. By creating a light gradient,

material properties can also be made into a gradient. The

patterned material facilitates folding and gives structural

rigidity. In this way, we can create repeated fold patterns for

a class of origami-based robots. This novel technique is used

to selectively embed a more rigid polymer inside of a

molded Silicone rubber skin. We then actuate motion of the

surface to create configuration changes, using inverted SMA

wires. This remarkable self-assembly of origami can make it

possible to be used it as artificial muscle and material

carrier/dispenser. Soft-bodied robots, capable of untethered

control and the ability to deform or morph to fit through

small apertures, can be advantageous for rescue missions and

structural health and environmental monitoring. Independent

agents, capable of unobtrusively penetrating even the

smallest opening to gather information or dispense small

payloads, will be invaluable in medical diagnosis and

(a) (b)

Figure 4. (a) Origami tube with SMA wire spring actuators (The wires are placed along the circumference to obtain the highest expansion ratio and at every other row to achieve efficient folding. We insert passive elastic material inside the tube to produce a restoring force that causes the tube to recover its original diameter after actuation. Passive unfolding requires fine tuning of actuation current and period, and elasticity of the passive elastic material.) [19]. (b) Connection capability of the Slime Mold robot: Serial linking of three modules, lifted off the ground. Only the middle module is actuated [20].

Figure 3. Schematic design of microrobotic fish ( top view ). It has

four rigid links connected by a compliant flexure. Planar coils and permanent magnets actuate the fish. Li-ion battery and electronics

are located on the opposite side of the planar coils. The entire

system is encapsulated by PDMS [18].

(1) (2) (3)

(4) (5) (6)

Fig. 5. Snap shots of the locomotion sequences of the Slime Mold robot: (1) Start, (2) Top front SMA wire spring actuates bending the front body

section upwards, while the bottom center SMA wire spring actuates bending the center section downward. (6) End [20].

.

treatment, human-robot interaction, and distributed

asymmetric warfare scenarios.

VI. CELLULAR SLIME MOLD ROBOT I : POLYMER

Cellular Slime Mold, Dictyostelium discoideum, has a

unique series of developmental events. The cells begin to

associate, forming streams of migrating cells which merge in

an aggregate of a number of cells. We have developed the

first prototype of biologically-inspired cellular Slime Mold

robot to imitate this behavior [20]. Slime Mold robot is a

modular robot with remarkable flexibility, shape change, and

integration capability. It is the first soft modular robot to the

best of the author’s knowledge. Their body is made of

flexible polymer with shape memory alloy wire springs

embedded inside as muscles as shown in figure 4 (b). It is a

rectangular shape silicone rubber with three links or

sections. The muscles of the Slime Mold robot consist of

inverted shape memory alloy (SMA) wire springs. By

actuating different sections of the springs, we can create

several different shapes from the same module, and control

the size of the individual Slime Mold modules. These basic

unit shapes are similar to DNA origami [21]. Hence, it can

enhance versatile assembly with other Slime Mold modules.

The Slime Mold modules are capable of latching onto other

modules. Figure 4 (b) shows the serial linking of three

modules, lifted off the ground. Note that only one module is

actuated to make a connection. The flexibility of the body

enables the versatile grasping, which allows various modes

of connection and group shape.

It has dual modes of locomotion: crawling and swimming.

Experiments have shown their crawling locomotion (figure

5) and grasping capability successfully. The slime mold

robots can function individually, collectively or

collaboratively.

VII. FLOWING ROBOTS WITH SWITCHABLE SURFACE

ACTIVITY OF LIQUID METAL

A. Slime Mold Robots II : Liquid Metal

Dictyostelium discoideum changes surface tension to aggregate and migrate. The first slime mold robot made of polymer has limitations. We propose to develop continuum modular robots by changing surface tension to better imitate the behavior of cellular slime mold. Liquid metal eutectic gallium indium (EGaIn) initially in a spherical shape in electrolyte flattens and spreads without bound upon application of an oxidative potential [4]. We can make EGaIn continuously flow and disperse at a low voltage. As it disperses, it loses contact with electrode and surface tension becomes high again. Then, it quickly forms a spherical shape. When an oxidation potential is applied, the liquid metal flows again and spreads to touch adjacent liquid metal drops, causing them flow together [4]. By moving the electrode touching the flowing liquid metal, we can make the aggregated drops flow/travel together. This phenomenon can be used to transport particles borne in liquid metal, to pass through narrow openings, to camouflage by wrapping an object, and to clean by carrying/moving debris. Also, by exciting a different electrode node at capillary channels, liquid metal flows into a different path.

B. Variable-Stiffness Robots, Gripper, Actuator

By changing the surface tension of EGaIn, we can change

stiffness. One method I propose is to interlace solid layers

with a soft capillary channel. When it is not filled with

EGaIn, it is a stack of solid sheets on top of one another with

a thin channel between them. The channel connects the

cathode at the bottom of the stack and the anode at the top of

the stack. Upon application of oxidation potential, EGaIn

flows into the capillary channel inflating the channel layer

by layer. The height of the stack can be controlled by the

amount of EGaIn in the channel and determines its stiffness.

Another way to control stiffness is by changing the electric

potential applied to the liquid metal inside a flexible tube.

The stiffness changes with surface tension. We can use the

unique rheological behavior of EGaIn to achieve variable

stiffness in many different ways.

B. Further Applications

Because it is observed that disturbance induced distortion of EGaIn can be restored at a certain potential, we can use well-controlled liquid metal drop as a micro gripper to grasp a variety of objects by varying the potential.

We can also use it to pump or rotate gears in water by oscillating the liquid metal inside a capillary channel.

We can also use the flowing liquid metal to manufacture wires through a complex shape.

VIII. CONCLUSION

Since Micro electro mechanical systems (MEMS)

fabrication technology enabled tiny micro robots in the late

1980s, many MEMS sensors and actuators have come into

the market. About a decade later, the discovery of the bulky

ball captured our imagination and ushered in the era of

Nanotechnology. Yet another decade later, small soft robots

emerged to make robots work well with human and with

environments individually or with other units.

We have shown some example systems at each scale to

shed light on how to harness the advantages of small-scale

robots. Then, this paper discusses the emergence of soft

robots. It presents small-scale soft robot systems to show

their promise and limitations to suggest a future direction.

Finally, it proposes a few flowing liquid robot systems

utilizing interesting phenomena of the switchable surface

tension of liquid metal.

The sample systems indicate that as the scale of robots

shrinks, it is inevitable for them to get connected to larger-

scale systems to work. The next decades will accelerate the

niche use of the small-scale robots if researchers focus more

on integrating them together as the robot technology at nano,

micro, milli meter scale mature.

We should develop soft robots to achieve morphology

change triggered by sensor signals eventually. Finally, the

intricate interplay of physics and chemistry in small scales

points that robotics community can richly benefit from

exploiting all three states of matter; solids, liquids, and

gases. We have to pursue all of these ideas and more to draw

out the full potential of small-scale robots.

REFERENCES

[1] Feynman, Richard P. (1960) There's Plenty of Room at the Bottom.

Engineering and Science, 23 (5). pp. 22-36.

[2] Lee, Eunjeong, “Design and Analysis of an Electrostatic Micro Crawler,” Northwestern University, Dec. 1989.

[3] J. M K Ng, M. J. Fuerstman, B. A. Grzybowski, H. A. Stone, G. M. Whitesides, “Self-Assembly of Gears at a Fluid/Air Interface,

“Journal of the American Chemical Society, 125(26):7948-58, August

2003.

[4] M. R. Khan, C. B. Eaker, E. F. Bowden, and M. D. Dickey, “Giant

and switchable surface activity of liquid metal via surface oxidation,” PNAS, vol. 111 no. 39 14047-14051, September 30, 2014.

[5] Lee, Eunjeong, "Evolutionary Flowing Robots with Switchable Surface Activity of Liquid Metal", the Proceedings of the workshop

on Self-healing, Growing and Evolving Soft Robots, the IEEE Int.

Conference on Robotics and Automation, Brisbane, Australia, May 2018.

[6] Lee, Eunjeong, Kim B., Ko J., Song C. Y., Kim S., Jeong S., and Lee

S., “An Integrated Micro HTS System for Energy Storage and Attitude Control for Three Axis Stabilized Nanosat,” the IEEE

Transactions on Applied Superconductivity, vol. 15, no. 2, pp.2324-2327, June 2005 ( also the 2004 IEEE Applied Superconductivity

Conference ).

[7] H. Dai, J. H. Hafner, A. G. Rinzler, D. T. Colbert, and R. E. Smalley, “Nanotubes as nanoprobes in scanning probe microscopy,” Nature,

vol. 384, pp. 147-150, 5 1996.

[8] R. Vajtai, B. Wei, Y. J. Jung, A. Cao, S. K. Biswas, G. Ramanath, and P. M. Ajayan, “Building and testing organized architectures of carbon

nanotubes,” IEEE Trans. Nanotechnology, vol. 2, no. 4, pp. 355-361, 2003.

[9] C. V. Nguyen, K.-J. Chao, R. M. D. Stevens, L. Delzeit, A. Cassell, J. Han, and M. Meyyappan, “Carbon nanotube tip probes: stability and

lateral resolution in scanning probe microscopy and application to

surface science in semiconductors,” Nanotechnology, vol. 12, pp. 363-367, 2001.

[10] S. I. Lee, S. W. Howell, A. Raman, R. Reifenberger, C. V. Nguyen,

and M. Meyyappan, “Nonlinear tapping dynamics of multi-walled carbon nanotube tipped atomic force microcantilevers,”

Nanotechnology, vol. 15, pp. 416-421, 2004.

[11] J.H. Hafner, C.L. Cheung, A.T. Woolley and C.M. Lieber, “Structural

and functional imaging with carbon nanotube AFM probes,” Prog.

Biophys. Mol. Biol., 77, p. 73, 2001.

[12] T. Ando, N. Kodera, E. Takai, D. Maruyama, K. Saito, and A. Toda,

“A High-speed atomic force microscope for studying biological macromolecules,” Proc. of National. Academy of Science, vol.98, no.

22, pp. 12468-12472, 2001.

[13] T. Ando, N. Kodera, Y. Naito, T. Kinoshita, K. Furuta, and Y. Toyoshima, “A High-speed atomic force microscope for studying

biological macromolecules in action,” A European Jr. of Chemical Physics and Physical Chemistry, 4, pp. 1196-1202, 2003.

[14] P.J. de Pablo, I.A.T. Schaap, C.F. Schmidt, Nanotechnology, 14,

pp.143-146, 2003.

[15] A.N. Round and M.J. Miles, Nanotechnology, 15, p. S176-S183,

2004.

[16] Lee, Eunjeong, “An Integrated System of Microcantilever Arrays with

Carbon Nanotube Tips for Imaging, Sensing, and 3D Nanomanipulation: Design and Control,” the Journal of Sensors and

Actuators A: Physical, Vol. 134, Issue 1, pp. 286-295, Feb. 2007.

[17] C. Ortiz, “Atomic force microscopy and high-resolution force spectroscopy for studies of the extracellular matrix of musculoskeletal

tissues,” Proc.of the 50th Annual Meeting of the Orthopedic Research Society, San Francisco, CA, March 2004.

[18] Lee, Eunjeong, “Design of a Soft and Autonomous Biomimetic

Micro-Robotic Fish,” the Proceedings of the 5th IEEE Conference on Industrial Electronics and Applications (ICIEA2010), June 2010.

[19] Lee, Eunjeong, Chen, X., Onal, C., Choi, W., Wood, R., Rus, D., and Whitesides, G., M., ‘Fabrication and Actuation of a Soft Robot with

Dramatic Morphing,” Conferences on Foundations of Nanoscience

(FANO), April 2010.

[20] Lee, Eunjeong, Yang, J., Jolda, M., and Wood, R., "Cellular Slime

Mold Robot," the Proceedings of the workshop on Modular Robots:

State of the Art, the IEEE Int. Conference on Robotics and Automation, Anchorage, Alaska, May 2010.

[21] S.M. Douglas, A. H. Marblestone, S. Teerapittayanon, A. Vazquez, G.M. Church, and W. M. Shih, “Rapid prototyping of 3D DNA-

origami shapes with caDNAno, “ Nucleic Acids Res. 37(15):5001-6,

August 2009.

Toward 3D Dexterous Micro-Manipulation1st Jean-Antoine Seon

Depart. of Electrical Engineering and AutomationAalto University

Finlandjean-antoine.seon@aalto.fi

2nd Pardeep Kumar, 3rd Redwan Dahmouche, 4th Michal GauthierDepart. of Automation and Micro-Mechatronic Systems-AS2M

FEMTO-ST Institut, Univ. Bourgogne Franche-ComteFrance

firstname.lastname@femto-st.fr

Abstract—One of the most effective ways to carry out complexmanipulation tasks is through contact manipulation. This paperpresents and validates through simulation and experimentationa dexterous micro-manipulation approach using a 6 Degreesof Freedom (DoF) hand. The developed algorithm associatedwith the dexterous micro-manipulation hand is able to generateoptimal trajectories to translate and rotate arbitrary shapedplanar objects in presence of adhesion forces. Simulation resultsshow that the presence of adhesive contacts can be exploited toenhance the manipulation performances. In addition, the pre-sented experiments show that large rotations can be performedin open loop with less than 5% error.

Index Terms—Dexterous Micro-Manipulation, Grasping, Ad-hesion Forces.

I. INTRODUCTION

For more than three decades, Dexterous Manipulation hasbeen a very active field of research [1] in macro-scale robotics.Indeed, the grasping and motion planning problems have beenformalized previously [2], and various approaches to performdexterous manipulation have been studied such as rolling [3],sliding [4] and finger gaiting [5]. However, it has not yet beenlargely investigated in micro-manipulation techniques whichare often limited to basic manipulation operations for simpleobjects [6]–[8].

Using a robot, there are two ways to obtain the rotations;the first one is the most commonly used in industries, consistsof using basic gripper mounted on a robot, which rotates thecarried gripper. In these systems, accuracy is limited by themechanical defects such as backlashes and eccentricity of therevolute joints.

The second method consists in using a dexterous handto perform in-hand rotations [9]–[11]. This requires moreelaborated grippers, and is very versatile as a single hand isable to manipulate various objects. Recently, Brazey et al. havedeveloped a micro-manipulation system that overcomes theproblem of large rotations of micro-objects [10]. For dexterousmicro-manipulations, the micro-hand fingers can be mountedon simple translation actuators with high repeatability. In thiscase, the accuracy depends on the contact between the fingersand object surface.

This work was supported by ACTION, the French ANR Labex no. ”ANR-11-LABX-01-01” (http://www.labex-action.fr), by the EquipexROBOTEXproject (contract ”ANR-10-EQPX-44-01”) and by the ConseilRegional de Franche-Comte

Moreover, at the micro-scale the dexterous manipulation isdifferent from the macro-scale because of the surface forcessuch as van der Waals, electrostatic and capillary forces thatare dominant over gravitational and inertial forces [12]. Dueto these dominating attractive forces also known as adhesionforces, the manipulation paradigm is changed as the objectsticks to both the manipulating fingers and the substrate.

To take into account these adhesion forces, two approachesmight be considered. The first one is to modify the contactproperties to reduce adhesion forces by varying environmentalparameters such as humidity and temperature [13] and control-ling the surface roughness [14]. While the second one consistsin developing new strategies to exploit the adhesion forces.Indeed, it has been shown that taking advantage of adhesionforces can enhance the gripping strategies [15].

Fig. 1. Illustration of an automatic 180°rotation of 200µm2 object performedusing three fingers on our manipulation setup.

Fig.1 represents the dexterous micro-manipulation of asquare shaped object performed on one of the systems with 6DoF developed by Brazey et al. [10]. This setup is composed

of three cylindrical fingers having two degrees of freedomeach, mounted on a support and translated by the actuationsystem to perform the manipulation. It is able to carry outlarge rotations.

In this paper, we propose to exploit these adhesion forces, toenhance the stability and the micro-manipulation performance.Original fingers’ trajectories are proposed and the benefit ofexploiting adhesion forces is demonstrated. Moreover, thiswork is not only limited to square shaped objects as simulationand experimental results of an arbitrary shaped object arepresented. Furthermore, the rolling constraint of the fingersduring the object rotation along with the sticky forces thatmay exist in the micro-scale are taken into account in thefingers path planning.

II. IN-HAND MANIPULATION PLANNING

This section is divided into two subsections; the first oneformalizes the grasping and in-hand manipulation problem inmicro-scale and emphasizes the contact modeling differenceswith the macro-scale, while the second subsection presents themanipulation strategy.

A. Contact Modeling

Under Coulomb friction, the grasping force−→f is con-

strained to lie in a friction cone centered about the internalsurface normal at the contact point X with half-angle α.Nonetheless, in the case of micro-manipulation, an additionalforce, called pull-off force (

−→fpo), is observed [16] (Fig. 2a).

This force modifies the Coulomb friction cone and the contactlaw is rewritten as follow:

ft < µ(fn + fpo) (1)

where ft is the tangential component of the force, fn is thenormal component and µ is the friction coefficient.

Because of the pull-off force−→fpo, it is possible to apply

negative forces on the contact point (pull the object) as longas the force lies in the modified friction cone.

Fig. 2. The modified Coulomb friction cone in micro-manipulation in whichit is possible to observe the pull-off force

−→fpo.

B. Grasp and Equilibrium

When n ngers grasp an object, the grasping wrench wext

must exactly compensate the external wrenches in order toreach the static equilibrium.

n∑w=1

wi + wext = 0 (2)

The external wrench is composed of all the external forces(apart from the finger contact forces) and the weight of themanipulated object wp.

C. Manipulation Planning

To perform the task of manipulation, it is assumed that thegeometry of object to be manipulated is well defined typicallyby using the CAD model. The fingers used in manipulationsystem are cylindrical having the same geometrical and me-chanical properties. After that we propose to generate thefinger trajectory by using two step method. The first step is tocomputing the set of all the stable grasps and the possiblefinger gaiting configurations. The second step consists innavigating between these configurations to define a path froman initial configuration to desired one. The former step can beachieved off-line and performed only once for a given object,while the later step can be carried out by a suitable searchalgorithm. In this paper, to generate the fingers’ trajectories,the A* algorithm is used.

During the manipulation process, it should be assured thatthe object remains in static equilibrium. To do so we haveused [18]. Manipulation of the object has been carried outusing rolling contacts and finger gaiting strategies proposed in[10].

We assume that the geometry of the object is known(typically a CAD model of the micro-part) and we propose togenerate the finger trajectories based on a two steps method.Given an object shape and the number of fingers of themanipulation system, the first step of the trajectory generationconsists in computing the set of all stable grasps and admissi-ble finger gaiting configurations. This step can be achieved off-line and is done only once for a given object. The second stepconsists in navigating between these configurations to definea path from the initial configuration to the desired one.

III. SIMULATION RESULTS

The methodology presented in the previous section has beenimplemented and tested to generate trajectories for planarmanipulation. In fact, in micro-assembly, most of the objectsmade using micro-fabrication techniques are planar objects.The results presented in this section are thus applicable inmicro-assembly.

The results below present trajectories for an arbitrary shapedobject with sticky fingers. Note that every initial grasp is takenrandomly in the set of admissible initial grasp.

One can note that, the translation of the object is simplyperformed by the translation of all the fingers which does not

affect the stability of the grasp. Thus, we will focus in thisstudy on the rotation.

Simulations have been carried out to illustrate the generatedtrajectories when considering sticky fingers. We consider themanipulation with a three fingers hand where each finger cantranslate in the plan.

Fig.3 illustrates the computed trajectory for a 206°rotationusing fingers with a radius of 500µm. By choosing a suitablesubstrate material the pull-off force between the object and thesubstrate can be neglected. The pull-off force between fingersand the object and the friction coefficient are respectivelyestimated at 1.5mN and 0.26 and the weight of the objectis neglected.

a) b)

c) d)

e) f)

Fig. 3. Images sequence describing the trajectory generated by the plannerfor a 206°rotation with sticky fingers: a) represents the initial graspingconfiguration while b) to f) represent the rotation steps.

IV. EXPERIMENTAL RESULTS

In order to validate these new manipulation strategies, theproposed trajectories exploiting adhesion forces have beentested experimentally at the millimeter scale.

A. Experimental Set-up

The experimental set-up (Fig. 4) consists of three cylindricalfingers of 1mm diameter made of steel, each one mountedon a 2 Degrees of Freedom translation table. Each table isactuated using 2 Smaract SLC-1730 piezoelectric actuatorswith a repeatability of 1.5µm. The scene is observed usingtwo high resolution (2560 × 2018 pixels) cameras from JAI.

The manipulated objects are made in acrylic and have thefollowing dimensions: 6mm long and 4mm tall.

As this experimental setup is at millimeter scale, so adhesionforces are not predominant. Thus to exhibit the behavior that

Fig. 4. Experimental set-up composed of three fingers moving in a planthanks to six micro-positioning stages.

is encountered in micro-scale, a polymer was deposited onthe fingers to enhance the adhesion forces. It generates abehaviour which is not identical to the micro-scale behaviorbut almost similar. In previous paper [18], we highlightedthat the adhesion strength has no impact on the stable graspsand re-grasping configurations. Thus, even if this polymerdoes not accurately mimic the micro-world, it is sufficient toemulate micro-scale specificities where adhesion forces arepredominant over other forces.

The trajectories being generated in the object frame and exe-cuted without feedback, the manipulation system must be wellcalibrated. Thus the homography between the camera frame,the object frame and the actuators frame are computed. Thanksto these homography matrices the computed trajectory for eachfinger is converted in the actuators frame and executed.

B. Experimental Validation

The trajectories presented previously have been tested onthe experimental set-up. Fig.5 represents an images sequencewhich shows the achievement of the 206°rotation for thearbitrary shaped object. It can be seen that this manipulationprocess is successfully executed. The reference trajectory andthe real one obtained using the ESM visual tracking algorithm[17] are shown in Fig.6. It can be noticed that the angle of themanipulated object reaches 199.25°which represents an errorof 3.26%. We can see from the same figure the evolution of theposition of the manipulated object during the manipulation.

A larger error can be observed on the object’s position.Indeed, on x−axis and y−axis there is an offset of 500µmand 2mm respectively at the final position of the object relativeto the reference trajectory. These errors can be explained by thefact that the characteristics of both the fingers and the object(dimension, stiffness, etc.) are not identical between simulationand experimentation. Nevertheless, the error in position can beeasily corrected by performing a visual servoing.

V. CONCLUSION

This paper presented a dexterous in-hand micro-manipulation method of an arbitrary shaped object usinga 6 DoF hand with cylindrical fingers. The manipulationapproach is based on the A* graph search algorithm in order

a) b)

c) d)

e) f)

1mm

Fig. 5. Images sequence captured during the experimental validation for a206°rotation: a) represents the initial grasping configuration while b) to f)represent the rotation steps. Fingers are represented by the red circles.

Fig. 6. Representation of the position and orientation of the manipulatedobject during the experimental validation of the 206°rotation. Left graphicscompare the reference trajectory with the executed one. Right graphics showthe various errors during the process.

to obtain optimal and complete trajectories. This algorithmexploits the adhesion forces which are usually observedbetween the fingers and the manipulated object at the smallscales. The analysis of the simulation results on the generatedtrajectories considering sticky fingers show that the existenceof adhesion forces can significantly enhance the performanceof the micro-manipulation operations when they are properlytaking into account. Simulation results have been confirmedand validated experimentally using open loop control. Largerotations of arbitrary shaped planar objects have then beendemonstrated with low rotation errors.

REFERENCES

[1] A. M. Okamura, N. Smaby and M. R. Cutkosky, “An overview ofdexterous manipulation,” in International Conference on Robotics andAutomation, pp. 255–262, 2000.

[2] Z. Li, J. Canny and S. Sastry, “On motion planning for dexterousmanipulation. i) the problem formulation,” in International Conferenceon Robotics and Automation, pp. 775–780, 1989.

[3] A. Bicchi and R. Sorrentino, “Dexterous manipulation through rolling,”in International Conference on Robotics and Automation, pp. 452–457,1995.

[4] D. L. Brock, “Enhancing the dexterity of robot hand using controlledslip,” in International Conference on Robotics and Automation, pp. 249–251, 1988.

[5] L. Han and J. C. Trinkle, “Dexterous manipulation by rolling and fingergaiting,” in International Conference on Robotics and Automation, pp.730–735, 1998.

[6] H. Xie and S. Regnier, “Three-dimensional automated micromanipula-tion using a nanotip gripper with multi-feedback,” Journal of Microme-chanics and Microengineering, vol. 19, no. 7, 2009.

[7] H. K. Chu, J. K. Mills and W. L. Cleghorn, “Dual-arm micromanipu-lation and handling of objects through visual images,” in InternationalConference on Mechatronics and Automation, pp. 813–818, 2012.

[8] D. J. Cappelleri, P. Cheng, J. Fink, B. Gavrea and V. Kumar, “Automatedassembly for mesoscale parts,” in Transactions on Automation Scienceand Engineering, vol. 8, no. 3, pp. 598–613, 2011.

[9] Q. Zhou, P. Korhonen, J. Laitinen and S. Sjovall, “Automatic dextrousmicrohandling based on a 6-dof microgripper,” Journal of Micromecha-tronics, vol. 3, no. 3, pp. 359–387, 2006.

[10] B. Brazey, R. Dahmouche, J. A. Seon and M. Gauthier, “Experimentalvalidation of in-hand planar orientation and translation at microscale,”Intelligent Service Robotics, Special Issue on Multi-scale ManipulationToward Robotic Manufacturing Technologies, vol. 9, pp. 101–112, 2015.

[11] J. Thompson and R. S. Fearing, “Automating microassembly with ortho-tweezers and force sensing,” in International Conference on IntelligentRobots and Systems, pp. 1327–1337, 2001.

[12] R. S. Fearing, “Survey of sticking effects for micro parts handling,” inInternational Conference on Intelligent Robots and Systems, pp. 212–217, 1995.

[13] M. Savia and H. N. Koivo, “Contact micromanipulation survey ofstrategies,” Tronsaction on Mechatronics, vol. 14, no. 4, pp. 504–514,2009.

[14] F. Arai, T. Fukuda, H. Iwata and K. Itoigawa, “Integrated micro endeffector for dexterous micromanipulation,” in Seventh InternationalSymposium Micro Machine and Human Science, pp. 149–156, 1996.

[15] J. Corney, P. Lambert and A. Delchambre, “A study of capillary forces asa gripping principle,” Assembly automation, vol. 25, no. 4, pp. 275–283,2005.

[16] J. A. Seon, R. Dahmouche, B. Brazey and M. Gauthier, ”FingerTrajectory Generation for Planar Dexterous Micro-Manipulation,” IEEEInt. Conf. on Robotics and Automation (ICRA), pp. 392-398, 2016.

[17] S. Benhimane and E. Malis, “Homography-based 2d visual tracking andservoing,” in The International Journal of Robotics Research, vol. 26,no. 7, pp. 661–676, 2007.

[18] J. A. Seon, R. Dahmouche, and M. Gauthier, “Enhance in-Hand Dex-terous Micro-Manipulation by Exploiting Adhesion Forces,” in IEEETransactions on Robotics, 2017.

A Compiler for Programming Molecular Robots∗

Inbal Wiesel, Noa Agmon and Gal A. KaminkaComputer Science Department

Bar Ilan University, IsraelCorresponding author: galk@cs.biu.ac.il

1. Introduction. Molecular robotics are a promising approach for biomedical applications. Molecularrobots (nanobots) can operate inside a living body [2, 3, 1, 11], carrying out molecular actions, such asreleasing a molecular payload only under some environmental conditions or shielding the body from toxicpayloads [4]. If used as a platform for drug delivery, a nanobot can, in principle, overcome many of the safetyissues, as drugs are released only in the presence of their targets. Nano-scale devices—including but notlimited to nano-robots—may also be used to carry out computations [13], but our focus here is on robotictasks (such as drug delivery) rather than arbitrary computations.

Currently, every nanobot must be designed by experts, for the specific task: medical expertise mustmeet nanobot design expertise. As procedures grow in complexity, the challenge is exacerbated: nanobotdevelopers mix different types of nanobots—each type specifically tailored—in nanobot cocktails (swarms),such that the medical outcome emerges out of the interactions of the various nanobots [10].

Recent advances have begun to explore generic nanobot arch-types, which can be “programmed” (spe-cialized) in specific ways. Recently developed nano-particles [11] serve as an example. These nanobots arebuilt from a nanometer-scale gold bead, to which various DNA strands can be attached, e.g., to bind withspecific biomarkers. The DNA-based clamshell [4] and walker [8] nanobots are other examples.

The development of programmable nanobot arch-types offers an opportunity to consider tools for pro-gramming nanobot cocktails. Inspired by modern software development environments, which separates high-level programming languages from specific CPU details, we aim to allow medical professionals to directlyprogram treatments in a Athelas, a medication programming language. the design of Athelas is motivatedby the success of rule-based systems at capturing expert knowledge [6, 7]. A compiler (Bilbo) translatesAthelas programs to nanobot specifications, which implement the program. The compiler relies on a libraryof generic nanobot arch-types, and specializes them to create the specific roles needed for the swarm.

We believe this separation between medical expertise and nanobot design expertise can significantlyaccelerate the development of new medical treatments relying on nanobot technology: Medical experts willprogram treatments. Molecular roboticists will develop generic nanobots. And compilers will synthesizeswarms of nanobots that carry out the programs, with performance and safety guarantees.

2. The Athelas Language. We consider nanobot tasks such as moving compounds between locations inthe body, picking compounds (by molecular binding), or exposing (and sometimes releasing) them in diseasedareas. To specify tasks, we adopt a rule-based programming paradigm, in which programs are specified bysets of rules that are continuously considered in parallel, against changing conditions [5]).

Each rule has four clauses. The Initialize clause specifies the set of payloads to be built into the drugwhen it is injected (i.e., before any action is taken). The When and Until clauses are each composed of a setof tests, e.g., pH level or concentration of a specific chemical in specific location. The When tests must holdin order for the drug to become activated (here, when the concentration of Y in the vicinity of T is above5mol/m3). The Until terminate the activity of the drug. The Actions clause contains the actions to beexecuted when the drug is active, e.g., pick, drop, protect, expose, disable and other actions.

∗This research was supported by ISF Grant 1511/12. Thanks to K. Ushi.

1

3. The Bilbo Compiler. The Bilbo compiler takes two inputs: an Athelas program, and a libraryof generic robot types (with defined ways of parameterizing them, including parameterizable preparationprotocols). It then synthesizes a specification for a heterogeneous swarm of specialized nanobots, whichwould carry out the program, once deployed. The output specification for each specialized robot includes aspecialized preparation protocol.

The compilation process is done in two phases. A front-end phase consists of the lexical and syntaxanalyzers, generates finite state machines (FSMs) representing the rules. The back-end phase then transformssuch FSMs into a final nanobot swarm specification (recipe). This is done by a graph-rewriting approach,with specific operators for merging, expanding FSM transitions and states, and rejecting incorrect paths. Atthe end of this process, an AND/OR graph emerges, which represents all possible nano-swarms (cocktails)that can carry out the program. The compiler uses the AO* [9] algorithm to determine an optimal AND/ORpath in the graph, which corresponds to a specific heterogeneous swarm, made of specialized nanobot arch-types and their preparation protocols.

4. Validation Elsewhere [12], we reported on experiments where the compiler was used to generatecocktails with 2–5 nanobot types, made of clamshell and nano-particle arch-types. While we continue ourwork on this front, we want to report here on other validation work we are carrying out.

In particular, since the generated nanobot swarms are intended to one day serve in biomedical appli-cations, we also consider safety and performance guarantees. We have proven that the main compilationalgorithms are complete, meaning that all possible cocktail alternatives are found. We have additionallyproven that these algorithms are sound, meaning that any cocktail found by the system is indeed a valid one(in the sense of matching the Athelas program).

References

[1] Y. Amir, E. Ben-Ishay, D. Levner, S. Ittah, A. Abu-Horowitz, and I. Bachelet. Universal computing byDNA origami robots in a living animal. Nature Nanotechnology, 9(5):353–357, May 2014.

[2] A. Cavalcanti, B. Shirinzadeh, T. Fukuda, and S. Ikeda. Nanorobot for brain aneurysm. InternationalJournal of Robotics Research, 28(4):558–570, 2009.

[3] L. Dong and B. Nelson. Tutorial - robotics in the small part ii: Nanorobotics. Robotics AutomationMagazine, IEEE, 14(3):111–121, Sept 2007.

[4] S. M. Douglas, I. Bachelet, and G. M. Church. A logic-gated nanorobot for targeted transport ofmolecular payloads. Science, 335(6070):831–834, Feb 2012.

[5] A. Gupta, C. Forgy, A. Newell, and R. Wedig. Parallel algorithms and architectures for rule-basedsystems. SIGARCH Computer Architecture News, 14(2):28–37, May 1986.

[6] F. Hayes-Roth. Rule-based systems. Communications of the ACM, 28(9):921–932, Sep 1985.

[7] A. A. Hopgood. Intelligent Systems for Engineers and Scientists. CRC Press, 2001.

[8] K. Lund, A. J. Manzo, N. Dabby, N. Michelotti, A. Johnson-Buck, J. Nangreave, S. Taylor, R. Pei,M. N. Stojanovic, N. G. Walter, E. Winfree, and H. Yan. Molecular robots guided by prescriptivelandscapes. Nature, 465(7295):206–210, 2010.

[9] N. Nilsson. Principles of Artificial Intelligence. Tioga Publishing, Palo Alto, CA, 1980.

[10] E. Ruoslahti, S. N. Bhatia, and M. J. Sailor. Targeting of drugs and nanoparticles to tumors. Journalof Cell Biology, 188(6):759–768, 2010.

[11] G. Y. Tonga, Y. Jeong, B. Duncan, T. Mizuhara, R. Mout, R. Das, S. T. Kim, Y.-C. Yeh, B. Yan,S. Hou, et al. Supramolecular regulation of bioorthogonal catalysis in cells using nanoparticle-embeddedtransition metal catalysts. Nature chemistry, 7(7):597–603, 2015.

[12] I. Wiesel, G. A. Kaminka, G. Hachmon, N. Agmon, and I. Bachelet. Late-breaking: First steps towardsautomated implementation of molecular robot tasks. In DNA Computing (DNA-21), 2015.

[13] E. Winfree, X. Yang, and N. C. Seeman. Universal computation via self-assembly of dna: Some theoryand experiments. In DNA based computers 2, volume 2, pages 191–213. Amer Mathematical Society,1999.

Toward Fully Functional Micro Robots: Actuation, Control, and Sensing

Elizabeth E. Hunter, Denise Wong, Edward B. Steager, Vijay KumarGRASP Laboratory, University of Pennsylvania

Philadelphia, PA 19104 USA

Over the past several decades, small-scale robotic systemshave emerged as a promising technology to enable a wide-range of applications such as targeted delivery [1], [2], cellmanipulation [1], [3], environmental monitoring [4], andmicroassembly [5]. Exploiting these capabilities, microscalerobots have the potential to aid in the study of small-scalephysical phenomena, to address significant challenges inprecision medicine, and to spur unprecedented innovations.Previously, studies have advanced progress in small-scalerobot design by developing micromanufacturing techniques[3], miniature sensors [4], and controllable actuators [6], [7].

Previously, we have made significant progress in theareas of actuation, control, and sensing for micro bio robots(MBRs) that span from millimeter to micrometer scales. Wehave primarily focused on creating robots that are well-suitedfor use in a variety of chemical and biological environmentssuch that they can assist in microbiological experimentation.Consequently, our robotic platforms have been engineered toincorporate or interact with cellular components.

Magnetic Actuation Many successful demonstrations ofsynthetic microstructure propulsion utilize external magneticfields [1]-[5]. We have previously fabricated and actuatedsmall-scale magnetic robots in order to collect and transporta variety of payloads. Fig. 1 shows magnetic microrobots,made of SU8 and iron oxide nanoparticles (IONPs), suc-cessfully transporting fluorescently labeled microbeads totargeted neurons enabled by closed-loop visual feedback

Fig. 1. Demonstration of delivery of a fluorescently labeled microbeadto individual neurons [1]; (a) Initial workspace showing neuron cellbodies, branching dendrites, microbeads, and a magnetic microrobot; (b)Autonomous delivery of a microbead to a targeted neuron.

(a) (b) (c)

Fig. 2. Magnetic microrobots facilitate binding of proteins to antibodies[2]; (a) Robots transport fluorescent microbeads coated with antibodies to aprotein coated region on a glass slide; (b) Initial configuration of beads priorto transport; (c) A microrobot is teleoperated to test the interactions betweenantibodies coated on the bead surface and the target protein (PECAM).

[1]. This robotic platform was also used to investigate thepicoscale forces involved in protein to antibody binding,which has significant implications in delivering therapeuticnanocarriers to specific tissues (Fig. 2) [2]. Furthermore,we have demonstrated manipulation with magnetic robotsacross length scales by controlling millimeter-scale robots atoil-water interfaces in order to exploit curvature effects toassemble micro-scale passive particles (Fig. 3) [5].

Biological Actuation In addition to magnetic actuation,we have engineered hybrid MBRs in which synthetic mi-crostructures are powered by swimming bacteria [7], [8].We have investigated two types of propulsion mechanismsarising from interactions between flagellated bacteria andsynthetic microstructures. In the first, MBRs are propelled bybacteria adhered to planar SU8 microstructures and releasedin a motility medium causing the microstructure to movewithout external stimuli (Fig. 4). These MBRs rotate in a

(a)

(b)

Fig. 3. Magnetic robots can be used assemble passive microstructures usingcapillary effects at fluid interfaces [5]; (a) Triangular robots can selectivelyassemble passive structures along their perimeter by exploiting curvatureof the fluid interface; (b) Demonstration of a teleoperated robot assemblypassive particles.

(d)(a)

(c)(b)

Fig. 4. Microstructures can be propelled by adherent motile bacteria [8];(a) MBRs with adherent bacteria demonstrate a predominantly clockwiserotation caused by the normal force (Fnni) generated from the forwardthrust of the bacterium. A clockwise moment is applied to the MBR bythe tangential force (Ft ti) generated from the hydrodynamic interactionbetween the flagellum and glass substrate; (b) Cells swim by rotating aflagellar bundle to produce a forward thrust; (c) When cells are exposed toblue light, they are immobilized; (d) Immobilizing cells on specific regionsof the MBR enables the study of forces that contribute to MBR motion.

40 µm 40 µm 40 µm

t = 50 s t = 105 s t = 305 s(c) (d) (e)

(a) (b)

Fig. 5. Microstructures having asymmetric features are powered viastochastic collisions between swimming bacteria and microstructure con-tours [7]; (a) Micro-gears exhibit rotation when immersed in bacterialbaths whereas micro-chevrons predominantly translate; (b) Asymmetricmicro-gears in a bacterial bath exhibit a biased rotation direction whichcorresponds to tooth chirality; (c) Prior to exposure with blue light, themicro-chevron translates with a slight clockwise rotational bias due touneven collisions on either side of the structure; (d) Automated blue lightexposure of a half-plane which selectively deactivates bacterial motors; (e)Post-exposure, the micro-chevron is still translating but its heading hasshifted in the clockwise direction.

predominantly clockwise direction due to an applied momentfrom bacteria located along the microstructure edge [8].Secondly, we designed and fabricated asymmetric micro-gears and micro-chevrons, which are propelled by stochasticcollisions between swimming bacteria and rigid microstruc-ture boundaries (Fig. 5a). Micro-gears exhibit biased rotationdirection corresponding to tooth chirality and micro-chevronsprimarily translate along their major axis (Fig. 5b) [7].

(c)

40μm

Fig. 6. Biosensing microrobots [4]; (a) Mag-netic microrobots transport sensor payloadswithin a region of interest; (b) Cells attachedto sensor payloads readout green fluorescentprotein (GFP) when exposed to UV light.

Control MBRspropelled by motilebacteria can beindependentlycontrolled byexposing the cellsto high intensityblue light. Motilebacteria normallyexert a propulsivethrust to swimthrough a fluidenvironment (Fig.

4b). However, when exposed to blue light, their flagellarbundle ceases to rotate and attenuates their forward thrust(Fig. 4c). We have shown that spatial light patterning canbe used to selectively deactivate bacterial motors in orderto achieve more precise MBR trajectories (Fig. 5c-e) andto investigate force contributions of bacteria interacting atdifferent positions along the MBR (Fig. 4d) [6]-[8].

Sensing In addition to actuators, cells can serve as sensorson-board microrobots. Leveraging tools from synthetic biol-ogy, we have created biological microrobots that can explore,record, and report on the state of their microenvironment [4].Fig. 6 shows a microrobot that deploys biosensor payloadsto monitor and report on environmental signals (UV light).

Recently, we created soft micro bio robots (SMBRs) fromnaturally-derived hydrogels embedded with IONPs that can

Chemical Storage and Release

Cellular Encapsulation

Magnetic Hydrogel Composite

(a)

pre-transport

post-transport

t = 20 s

t = 38 s

pre-transport

t = 20 s

t = 38 s

post-transport

(b)

t = 0 s t = 20 s

t = 38 s t = 60 s 10 mm

(a) (b)

5 µm

(b)

5 µm

(a)

(b)

(c)

(a)

Fig. 7. Soft micro bio robots (SMBRs) [3] (a) An idealized SMBR trans-ports cellular and chemical payloads; (b) Environmental scanning electronmicrograph of cells proliferating on the robot surface; (c) Demonstration ofcellular transport. Cells produce green fluorescent protein (GFP) on-board.

encapsulate and transport chemicals or living cells, whilesupporting cell growth and function on-board [3]. We demon-strated that helical SMBRs can be propelled using rotatingmagnetic fields through a wide-range of Reynolds numberregimes in order to transport live cellular cargo (Fig. 7).

Over the past decade, we have demonstrated significantprogress toward creating fully functional microscale robotsoutfitted with a rich suite of capabilities. We have workedtoward integrating actuators, controllers, and sensors intoa complete robotic platform that can perform precisionassembly and delivery of synthetic and cellular payloads.Magnetism has offered the most promise as an actuationmodality, but we have also had success using motile cells topropel man-made structures. Using tools from synthetic biol-ogy, we have used living cells as sensors in small-scale robotsand have been inspired to create a robot architecture thatfully supports cellular growth and function. These studiessuggest the great potential of using cells as programming andmanufacturing units on-board small-scale robotic systems.

REFERENCES

[1] E. B. Steager, M. Selman Sakar, C. Magee, M. Kennedy, A. Cowley,and V. Kumar, “Automated biomanipulation of single cells usingmagnetic microrobots,” Int. J. Rob. Res., vol. 32, no. 3, pp. 346–359,mar 2013.

[2] E. B. Steager, B. Zern, M. S. Sakar, V. Muzykantov, and V. Kumar,“Assessment of protein binding with magnetic microrobots in fluid,”in 2013 IEEE Int. Conf. Robot. Autom. IEEE, may 2013, pp.5502–5507.

[3] E. Hunter, E. Brink, E. Steager, and V. Kumar, “Toward Soft MicroBio Robots for Cellular and Chemical Delivery,” IEEE Robot. Autom.Lett., pp. 1–1, 2018.

[4] E. B. Steager, D. Wong, D. Mishra, R. Weiss, and V. Kumar, “Sensorsfor micro bio robots via synthetic biology,” in 2014 IEEE Int. Conf.Robot. Autom. IEEE, may 2014, pp. 3783–3788.

[5] D. Wong, I. B. Liu, E. B. Steager, K. J. Stebe, and V. Kumar,“Directed micro assembly of passive particles at fluid interfaces usingmagnetic robots,” in 2016 Int. Conf. Manip. Autom. Robot. SmallScales. IEEE, jul 2016, pp. 1–6.

[6] E. B. Steager, D. Wong, N. Chodosh, and V. Kumar, “Opticallyaddressing microscopic bioactuators for real-time control,” in 2015IEEE Int. Conf. Robot. Autom. IEEE, may 2015, pp. 3519–3524.

[7] E. E. Hunter, N. Chodosh, E. B. Steager, and V. Kumar, “Control ofmicrostructures propelled via bacterial baths,” in 2016 IEEE Int. Conf.Robot. Autom. IEEE, may 2016, pp. 1693–1700.

[8] D. Wong, E. E. Beattie, E. B. Steager, and V. Kumar, “Effect ofsurface interactions and geometry on the motion of micro bio robots,”Appl. Phys. Lett., vol. 103, no. 15, p. 153707, 2013.

Microrobotics from Science to the Public: an Embedded Mobile

Microrobotic Kit for Education

Ioan Alexandru IVAN12, Florin DRAGOMIR1, Ioana Daniela DULAMA3, Ion Valentin GURGU3,

Nicolae Gabriel RADULESCU13, Ioan Alin BUCURICA3

1 Valahia University of Targoviste, Electrical Engineering, Electronics and Information Technology Faculty,Automation , Computer Science and Electrical Engineering Department, 130004 Targoviste, Romania

2 Universite de Lyon, ENISE, LTDS, UMR 5513 CNRS, 42023 Saint-Etienne Cedex 2, France 3 Valahia University of Targoviste, Institute of Multidisciplinary Research for Science and Technology, 130004 Targoviste,

Romania

e-mail: ivan@enise .fr, ioan.alexandru.ivan@gmail. com

This mobile microrobotic system was designed in a twofold purpose: designing a very compact platform for scientific

research and bridging the gap between science and general public. The microrobot kit has been designed in two versions:

an advanced one (Fig.1) designed for university students and researchers interested in microrobotics and nonlinear

control as well as basic version (Fig.2) for the general public dissemination. The microrobotic system consists of a sub-

millimeter magnetic element (or microrobotic agent) placed with other passive objects to be manipulated (mechanical or

biological) in a microfluidic arena. The system is fully embedded with camera, computer and electronics. The actuation is

manual or automatic under continuous magnetic field gradients generated by a set of four specific coils. The smooth and

precise control is quite challenging due to the complex influence of shape, size, fluid viscosity, friction coefficients of the

surface, field gradients nonlinearities etc.

The operation of the platform may be manual or automatic. The manual operation consists in attaching a joystick to the

device, which allows the user to control the movement of the robotic agent in two dimensions. Even though normally

this type of operation is supposed to be easily achieved, the human reaction in comparison with the computer control is

much slower and inaccurate. Therefore, to assist the imprecision of the joystick manipulation, we designed a specific

HMI (Human to Machine Interface) and attached a series of buttons for the adjustment of displacement speeds and

travel directions. This operating mode has been showcased in various scientific events (Fig.3) where top level scientists

as well as regular public tested this microscale system. In the regular demonstration, the 200µm microrobot may

assemble a planar group of microscale components of 200x350 µm in a confined region. This task simulates anticipated

medical applications such as inserting objects in a blood vessel.

In the automatic mode the system uses computer vision algorithms to feedback the microrobot’s position and

orientation. The control system is able to move this small robotic agent through different given paths with less than

50µm precision. The user intervention is required only in the first stage of the process, when the arena calibration and

reference trajectory are set.

Applications of the advanced kit is for the Microrobotics and Science dissemination from the academic research

environment towards the general public. It has been also successfully used for teaching in University lab classes such as

explaining the micro mechatronic concept, computer vision, tuning the PID-controller, study of the electronic board

amplifiers and signal converters etc. Not at least, it is an advanced example fields of Software development for the IoTs

(the device can be controlled remotely via internet).

Acknowlwdgements

This work was supported by the Romanian National Authority for Scientific Research and Innovation, CNCS/CCCDI - UEFISCDI,

project number PN-III-P2-2.1_PED-2016-1675, within PNCDI.

Figure 1. The main elements of the embedded mobile microrobotic kit.

Figure 2. A “low-cost” version of the kit designed with open-source and removable components for middle-schools.

Figure 3. Microrobotics from science to the general public. The micororobotic kit being manipulated by children in public

science events such as Researcher’s Night in Bucharest, 2016.

Integration of Coil Design, Path Planning, andImage Processing for an Autonomous Microrobotic

ManipulatorChristopher Lucasius, Lucas Botelho, You Dong

Abstract—In this poster presentation, we designed and manu-factured a 4 degree of freedom mobile microrobotic system thatcan autonomously or manually be actuated to manipulate cargoto a desired pose. We used a 3-axis Helmholtz coil system toactuate the microrobot in all three cardinal directions. A PIDcontroller uses these actuations to have the microrobot track adesired trajectory and reach a correct pose, using an A* pathplanner to choose the most efficient path. To inform the pathtracking controller, standard image processing algorithms areused to differentiate the microrobot and cargo and track theirpose in real time.

Index Terms—Microrobotics, Electromagnetics, Path Planning,Path Tracking Control, Computer Vision

I. INTRODUCTION

This abstract is written in support of the University ofToronto’s poster presentation for the Mobile MicroroboticsChallenge. It presents an overview of the competition tasks,an overview of our methods based on research over the past30 years, and technical details about the coil design, controls,and computer vision algorithms.

II. COMPETITION DESCRIPTION

The Mobile Microrobotics Challenge (MMC) is an annualcompetition hosted by the IEEE Robotics and AutomationSociety as part of the International Conference for Roboticsand Automation. This challenge aims to encourage researchersto accelerate the development of microrobots that can au-tonomously navigate and manipulate objects at the microscale.Currently, several methodologies have been explored in thewireless control of untethered magnetic microrobots [1], andtheir improvement for the use of this competition will be themain focus of this proposed research.

The MMCs competition tasks are designed to simulateanticipated applications in microassembly, including the ma-nipulation of blood vessels in biomedical projects as well asthe assembly of components in nanomanufacturing [2]. Themicrorobots are designed to be untethered through its controlvia external magnetic fields, enabling them to be minimallyinvasive and maximally dexterous in medical applications suchas intraocular drug deliveries [3].

III. OVERALL APPROACH

Computer vision is the first stage that enables the lat-ter control stages of the microrobot. The algorithm utilizesthe functions available within the OpenCV library based onC/C++. It runs on a separate thread in parallel with the

path tracking controller thread above. The algorithm consistsof an outer loop, which continues to run as long as livecamera data is fed into the algorithm. Inside the loop, itfirst performs preliminary image processing to differentiate themicrorobot from the cargo based on sizes and grayscale values.It then determines both of their position and orientation usingexisting OpenCV functions and custom-written functions. Asthey move through the field, they are tracked separately bythe algorithm in real time, and their pose information iscontinuously fed into the path planning and control algorithmfor their actuation.

Over the past few years, several groups have developedadvanced methods for planning trajectories for the intelligentcontrol of multiple robots [4], efficient path planners for quickmovements of microrobots with obstacles [5], and state-of-the art actuation techniques to overcome friction [6]. Thisposter presentation aims to combine these areas of researchby developing a path planner and a path tracking controllerin a microrobotic closed-loop system that can autonomouslynavigate towards a cargo and manipulate its pose.

Through a rigorous selection process, it was determined thatthe optimal settings for the path tracking controller wouldbe to use a 20 cSt silicone oil medium in a smooth castarena with a polyurethane robot and cargo. The path trackingcontroller used a combination of gradient pulling, and stickslip motion was implemented to allow the robot to followa desired position in the arena. The gradient pulling methodallows the robots position to be finely controlled with a PIDcontroller, hence leading to relatively smooth and accuratemotions. However, the robot can often become stuck to thefloor of the arena due to high static friction. To mitigate this, astick slip motion is implemented by using the vertical magneticfield the robot to have minimal contact with the floor. Pathtrackings input trajectory is also optimized by using the A*path planning algorithm to ensure that the robots controllertracks the best set of waypoints to follow.

After conducting an extensive literature review, it wasdiscerned that various projects have used tri-axial Helmholtzcoils, and each of the respective research groups have re-designed the coils for their own implementations. Therefore, arecent research paper was published in 2015 by J.J. Abbot inReview of Scientific Instruments, which discussed methods tostandardize the design of tri-axial Helmholtz coils to be usedas an electromagnetic actuation technique for the control ofmicrorobots [3].

The hardware used to actuate the microrobotic systemconsists of two main components: the coil system and servodrivers. The coil system was designed using the Biot-Savartlaw to approximate the strength of a magnetic field due tocurrent flowing through a wire. By understanding the strengthof the magnetic field, it was possible to determine the smallestpossible coil which provides an adequately uniform field. Oncethe inner coil was designed, the middle and outer coils weredesigned based on the geometry of the inner coils to createa 3-axis helmholtz system. Servo controllers are used as ameans of communication between the computer and the coilsystem. Once the path-planning code determines the directionthe microrobot must move in, a signal is sent to the servocontrollers. These then use this signal to provide a current tothe coils which creates the desired magnetic field. The servocontrollers selected interface directly to the computer via USBfor simplicity.

IV. IMPLEMENTATION DETAILS

In the following section, technical details are described thatmake up the proposed autonomous microrobotic manipulator.

A. Image Processing

Basic functions such as dilation and erosion are used forpreliminary image processing to prepare for finding contours.Then, depending on the actual shape of the part, a shapedetection function is used to determine the location of thepart. Since these functions only return angle values from 0to 90 degrees, custom functions are written to output the fullorientation of the part from 0 to 360 degrees. In addition, anoccupation matrix with the size of the frame is created to tellthe path planner which areas are occupied (see Fig. 1).

B. Path Tracking Controller and Planning Design

A PID controller is used to output the currents in theamplifier system’s electromagnetic coils in order to generate acontrollable gradient field. One problem that arose from onlyusing the gradient force was that the robot would be sub-jected to large static friction, causing unintended accelerations.Therefore, apart from the gradient force, stick-slip motion isalso utilized as the main actuation technique to enable therobot to move more easily across its medium.

At each point in time, the microrobot is given a singlewaypoint to follow, and the coils are controlled to have therobot move towards one point at a time until it is within anempirically found threshold. These waypoints are intelligentlyfound by using the A* path planner. Since the path therobot needs to follow is along a simple two-dimensionalconfiguration space, use the Manhattan Distance is used asan admissible heuristic to ensure that the generated path isalways optimal. Only the robot’s position is considered in pathplanning, assuming the robot can be oriented appropriately ateach endpoint of its path by using the uniform field to rotatethe robot to the required angular position.

Fig. 1. Occupation Matrix

(a) Coil System (b) Microrobotic Arena

Fig. 2. (a) 3-axis Helmholtz Coil System. (b) Microrobotic system performingthe Autonomous Manipulation Challenge.

C. Coil Design

The final coil system used to actuate the microrobot iscomprised of three pairs of orthogonal electromagnetic coilsto generate a magnetic field in any direction in 3-dimensionalspace. Each of the six coils is driven independently by a USBservo controller. The independent controls allows for controlover both the magnetic field and magnetic field gradient. Whilethe USB servo controllers allow for efficient communicationwith the path tracking controller.

V. RESULTS

In the latest experiment, the microrobotic system wasaccurately controlled to complete the MMC’s AutonomousManipulation task within 57 s (see Fig. 2).

REFERENCES

[1] ICRA, ”Mobile Microrobotics Challenge,”06 October 2017. [Online]. Available:https://sites.google.com/site/mobilemicrorboticschallenge/home.[Accessed 05 February 2018].

[2] J. E. Normey-Rico, I. Alcala and J. Gomez-Ortega, ”Mobile Robot PathTracking Using a Robust PID Controller,” Control Engineering Practice,vol. 9, no. 11, pp. 1209-1214, 2001.

[3] J. J. Abbott, Parametric design of tri-axial nested Helmholtz coils,Review of Scientific Instruments, vol. 86, no. 5, p. 054701, 2015.

[4] M. Salehizadeh and E. Diller, Two-Agent Formation Control of Mag-netic Microrobots in Two Dimensions, Journal of Micro-Bio Robotics,vol. 12, pp. 9-19, 2017.

[5] S. Chowdhury, W. Jing, P. Jaron and D. J. Cappelleri, Path Planning andControl for Autonomous Navigation of Single and Multiple MagneticMobile Microrobots, in ASME 2015 International Design EngineeringTechnical Conferences and Computers and Information in EngineeringConference, Boston, 2015.

[6] Pawashe, C., Floyd, S. and Sitti, M. (2009). Modeling and ExperimentalCharacterization of an Untethered Magnetic Micro-Robot. The Interna-tional Journal of Robotics Research, 28(8).

1

Motion Control, Planning and Manipulation ofMultiple Nanowires under Electric-Fields in Fluid Suspension for

Automated Characterization and NanoassemblyKaiyan Yu, Jingang Yi, and Jerry W. Shan

Abstract— We present an electric-field-based autonomoussystem to motion plan and control of multiple nanowires inliquid suspension with a simple, generic set of electrodes. Theproposed robust motion control has been proved to be stable forprecisely drive multiple various types of nanowires. The motionplanning algorithms significantly reduce the computationalcomplexity while maintain suboptimal performance in boththe travel time and distances. The performance of the motionplanning algorithms is guaranteed by analyses and design. Wefabricate the microfluidic devices and extensively demonstrateexperimental results to validate the analysis and the design ofnanowire motion control, planning and manipulation. Finally,We validate the automated manipulation scheme by assemblingfield-effect transistors (FETs) with silicon nanowires that arecharacterized, separated, and deposited according to theirelectrical conductivities. The scheme provides a key enablingtechnology for the scalable, automated sorting and assembly ofnanowires and nanotubes to build functional nanodevices.

I. BACKGROUND

A wide variety of nanoscale materials with well-definedgeometry have been explored in recent years for their uniqueand often enhanced properties [1]. Automated manipula-tion of nanowires and nanotubes would enable the scalablemanufacturing of nanodevices for a variety of applications,including micro/nanoelectronics and biomedical applications.Precisely placement of nanostructures such as nanowires ornanotubes and automated scalable characterization, manipu-lation and assembly of nanostructures are among technolog-ical challenges to fabricate these nanodevices [2]–[4]. Theuse of electric fields to position and sort micro- and nano-particles provides one solution to address the aforementionedchallenges for nano-material characterization and assembly.Under precisely controlled electric fields in fluid suspension,a variety of phenomena, including electrophoresis (EP),electro-osmosis (EO), dielectrophoresis (DEP), and electro-orientation, can be used as a driving force to steer, character-ize and sort nanowires [5]–[10]. Particles in fluid suspensionexperience an EP force under DC fields, while the fluiditself experiences EO forces [11]. The EP force magnitudeis proportional to the effective electrokinetic potential (i.e.,the zeta-potential) and the field strength. The use of EP forceas an actuation source is convenient in that even nominallyelectrically neutral particles typically have a nonzero elec-trostatic potential at the slip plane in the interfacial doublelayer in some solvents or at some pH values. Thus, byactuating a set of electrodes, nanowires in fluid suspensioncan be manipulated and steered to desired locations under a

K. Yu is with Department of Mechanical Engineering, BinghamtonUniversity, Binghamton, NY 13902 USA. J. Yi and J. W. Shan arewith the Department of Mechanical and Aerospace Engineering, RutgersUniversity, Piscataway, NJ 08854 USA. (email: kyu@binghamton.edu;jgyi@rutgers.edu; jshan@jove.rutgers.edu).

A

B

LL

S1

T1

Ez

locations

Top electrode

Ey

Ex

Electrode arraysDeposited

S2

T2

NN

x

y z

Fig. 1. Schematic of microfluidic device with N � N electrodes inde-pendently actuated electrodes on the bottom substrate, with a common topelectrode.

precisely controlled electric field [10], [12]. An EP-basedmotion-planning and manipulation scheme was developedto drive single [8]–[10] and multiple [12]–[14] individualnanowires in fluid suspension from one location to theirtargeted positions.

II. METHOD

In order to scalable manipulate of nanostructures, a simpleand generic set of N �N lattice-shape distributed electrodearray is designed to actuate suspended nanowires underelectric-fields in fluid suspension. The precisely controlledelectric fields generated by electrode arrays can be used tocontrol the motion of nanowire. Fig. 1 shows the schematicof the microfluidic device to manipulate nanowires. Thenanowires were suspended in a droplet of fluid with thelattice assembly electrode array. Those electrodes can beindependently actuated with different DC voltages to gen-erate the electrical field to control the nanowires motion inthe horizontal plane. A top electrode is used to control thenanowires in the vertical direction. Once the nanowires reachtheir desires deposition locations, the bottom electrodes areturned off and the electrode on the top surface is turned on toalign and drive the nanowires vertically until they reach thebottom substrate. Finally, the electrode arrays are turned onagain to lay down and deposit the nanowire onto the devicesurface with the desired orientation.

To compensate for un-modeled disturbances such as EO-induced flow motions, a robust motion control strategy wasdeveloped to guarantee the path-following errors convergencefor individual nanowires [10], [12]. The path-tracking con-trol adjusts the desired velocity profile according to thedirection of the desired trajectory, penalizes the trackingdirection, and compensates for the predicted error. A self-

2

20 um

11’

2’

3’

3

2

(a)

20 um

1

1’

2’

3’3

2

4

4’

5

5’

(b)

1

1’

2’

3’

3

2

20 um

(c)

1

1’

2’

3’

3

2

20 um

(d)

10 um

(e)

Fig. 2. (a) - (d) Simultaneous control and manipulation of multiple nanowires to form different geometric patterns. The “�” and “�” marks indicate thenanowires’ initial and final positions, respectively. (a) Overlaid nanowire trajectories to form a straight-line, (b) a circle with one nanowire at the center, and(c) an equilateral triangle pattern. (d) Overlaid nanowire trajectories of simultaneous tracking control using three nanowires to track a circular trajectory.(e) Deposition of six nanowires on the device substrate to form a straight-line pattern.

50 um

Starting location of NWsDeposit location of NWs

Characterization region

HC FET

electrodes

LC FET

electrodes

Fig. 3. Overlaid trajectories of two nanowires separated by their differentelectrical conductivities. The low-conductivity nanowire (σLC � 0.045S/m) is placed between the right pair of FET electrodes, while the high-conductivity nanowire (σHC ¥ 0.46 S/m) is placed between the left pairof FET electrodes. The red dash square shows the characterization region.The star and ellipse marks represent the starting and target locations of eachnanowire. The figure shows overlaid center-of-mass trajectories of the twosimultaneously characterized and separated nanowires.

placing time-suspension technique is also used to adjustthe desired rate of the progression and then the velocityprofile online within the planning framework at each timestep. Heuristic- [10], network flow- [12], and sampling [14]-based motion planning algorithms are utilized to achievethe shortest-distance and minimum-time objectives. Theseplanning algorithms addressed combinatorial optimizationwith coupled actuation to control multiple nanowires andwere validated to simultaneously steer multiple nanowires tofollow desired trajectories under controlled EP forces [8]–[10], [12]–[14]. Fig. 2 shows the motion control, manipula-tion, and deposition of multiple nanowires. The nanowirescan be positioned with spatial accuracy of 2µm.

III. RESULTS

We combine solution-based electro-orientation spec-troscopy (EOS) [6], [7] and the proposed motion-control,planning and manipulation strategies to simultaneously char-acterize and manipulate multiple individual nanowires. Thesenanowires can be selected according to their electrical char-acteristics and precisely positioned at different locations toform functional nanodevice with desired electrical properties.

Fig. 3 shows the automated manipulation and assembly ofnanowires into field-effect transistors (FETs). The siliconnanowires are characterized, separated, and deposited bytheir electrical conductivities. More measurement results ofthe nanowire-assembly devices will be presented in the ICRA2018: 30 Years of Small-Scale Robotics Workshop.

REFERENCES

[1] D. L. Fan, F. Q. Zhu, R. C. Cammarata, and C. L. Chien, “Electrictweezers,” Nano Today, vol. 6, pp. 339–354, 2011.

[2] Q. Cao and J. A. Rogers, “Ultrathin films of single-walled carbonnanotubes for electronics and sensors: A review of fundamental andapplied aspects,” Adv. Mater., vol. 21, no. 1, pp. 29–53, 2009.

[3] D. Fan, Z. Yin, R. Cheong, F. Q. Zhu, R. C. Cammarata, C. L. Chien,and A. Levchenko, “Subcellularresolution delivery of a cytokinethrough precisely manipulated nanowires,” Nat. Nanotech, vol. 5, pp.545–551, 2010.

[4] A. J. Hilmer and M. S. Strano, “Nanobiotechnology: Nanowires havecells in their sights,” Nat. Nanotech, vol. 5, no. 7, pp. 481–482, 2010.

[5] R. Probst, Z. Cummins, C. Ropp, E. Eaks, and B. Shapiro, “Flowcontrol of small objects on chip: Manipulating live cells, quantumdots, and nanowires ,” IEEE Control Syst. Mag., vol. 32, no. 2, pp.26–53, 2012.

[6] C. Akin, J. Yi, L. C. Feldman, C. Durand, A.-P. Li, M. A. Filler, andJ. W. Shan, “Contactless determination of electrical conductivity ofone-dimensional nanomaterials by solution-based electro-orientationspectroscopy,” ACS Nano, vol. 9, no. 5, pp. 5405–5412, 2015.

[7] C. Akin, L. C. Feldman, C. Durand, S. M. Hus, A.-P. Li, H. Y.Hui, M. A. Filler, J. Yi, and J. W. Shan., “High-throughput electricalmeasurement and microfluidic sorting of semiconductor nanowires,”Lab Chip, vol. 16, pp. 2126–2134, 2016.

[8] K. Yu, X. Lu, J. Yi, and J. Shan, “Electrophoresis-based motionplanning and control of nanowires in suspended fluids,” in Proc. IEEEConf. Automat. Sci. Eng., Madison, WI, 2013, pp. 831–836.

[9] K. Yu, J. Yi, and J. Shan, “Motion control and manipulationof nanowires under electric-fields in fluid suspension,” in Proc.IEEE/ASME Int. Conf. Adv. Intelli. Mechatronics, Besancon, France,2014, pp. 366–371.

[10] ——, “Motion control, planning and manipulation of nanowires underelectric-fields in fluid suspension,” IEEE Trans. Automat. Sci. Eng.,vol. 12, no. 1, pp. 37–49, 2015.

[11] R. Hunter, Foundations of Colloid Science. Oxford, UK: OxfordUniversity Press, 1989.

[12] K. Yu, J. Yi, and J. Shan, “Simultaneous multiple-nanowire motioncontrol, planning, and manipulation under electric fields in fluidsuspension,” IEEE Trans. Automat. Sci. Eng., vol. 15, no. 1, pp. 80–91,2018.

[13] ——, “Motion planning and manipulation of multiple nanowiressimultaneously under electric-fields in fluid suspension,” in Proc. IEEEConf. Automat. Sci. Eng., Gothenburg, Sweden, 2015, pp. 489–494.

[14] ——, “Time-optimal simultaneous motion planning and manipulationof multiple nanowires under electric-fields in fluid suspension,” inProc. IEEE Conf. Automat. Sci. Eng., Dallas, TX, 2016, pp. 954–959.