See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233415474

How Should Microrobots Swim?

Article  in  The International Journal of Robotics Research · October 2009

DOI: 10.1177/0278364909341658

CITATIONS READS
596 1,919

7 authors, including:

Brad Nelson Jake Abbott

ETH Zurich University of Utah
875 PUBLICATIONS   36,192 CITATIONS    112 PUBLICATIONS   8,803 CITATIONS   

SEE PROFILE SEE PROFILE

Kathrin E Peyer Li Zhang

Sewtech Automation ETH Zurich
36 PUBLICATIONS   3,875 CITATIONS    324 PUBLICATIONS   16,114 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Magnetic Guidance for Improved Cochlear-Implant Insertion View project

Force Sensing Cathether for the Treatment of Cardiac Arrhythmia View project

All content following this page was uploaded by Kathrin E Peyer on 22 May 2014.

The user has requested enhancement of the downloaded file.

 The International Journal of Robotics
Research
http://ijr.sagepub.com

How Should Microrobots Swim?

Jake J. Abbott, Kathrin E. Peyer, Marco Cosentino Lagomarsino, Li Zhang, Lixin Dong, Ioannis K. Kaliakatsos and
Bradley J. Nelson
The International Journal of Robotics Research 2009; 28; 1434 originally published online Jul 21, 2009;
DOI: 10.1177/0278364909341658

The online version of this article can be found at:

http://ijr.sagepub.com/cgi/content/abstract/28/11-12/1434

Published by:

http://www.sagepublications.com

On behalf of:

Multimedia Archives

Additional services and information for The International Journal of Robotics Research can be found at:

Email Alerts: http://ijr.sagepub.com/cgi/alerts

Subscriptions: http://ijr.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.co.uk/journalsPermissions.nav

Citations http://ijr.sagepub.com/cgi/content/refs/28/11-12/1434

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

Jake J. Abbott
Institute of Robotics and Intelligent Systems,
How Should Microrobots
ETH Zurich, 8092 Zurich, Switzerland
and
Swim?
Department of Mechanical Engineering,
University of Utah, Salt Lake City, UT 84112, USA
jake.abbott@utah.edu

Kathrin E. Peyer
Institute of Robotics and Intelligent Systems,
ETH Zurich, 8092 Zurich, Switzerland

Marco Cosentino Lagomarsino

Department of Physics, University of Milan,
20133 Milan, Italy

Li Zhang
Institute of Robotics and Intelligent Systems,
ETH Zurich, 8092 Zurich, Switzerland

Lixin Dong
Institute of Robotics and Intelligent Systems,
ETH Zurich, 8092 Zurich, Switzerland
and
Department of Electrical and Computer Engineering,
Michigan State University, East Lansing, MI 48824, USA

Ioannis K. Kaliakatsos
Institute of Robotics and Intelligent Systems,
ETH Zurich, 8092 Zurich, Switzerland

Bradley J. Nelson
Institute of Robotics and Intelligent Systems,
ETH Zurich, 8092 Zurich, Switzerland

Abstract diagnosis and therapy. Researchers have proposed numerous micro-

robotic swimming methods, with the vast majority utilizing magnetic
Microrobots have the potential to dramatically change many aspects fields to wirelessly power and control the microrobot. In this paper,
of medicine by navigating through bodily fluids to perform targeted we compare three promising methods of microrobot swimming (us-
ing magnetic fields to rotate helical propellers that mimic bacterial
flagella, using magnetic fields to oscillate a magnetic head with a
The International Journal of Robotics Research
Vol. 28, No. 11–12, November/December 2009, pp. 1434–1447
rigidly attached elastic tail, and pulling directly with magnetic field
DOI: 10.1177/0278364909341658 gradients) considering practical hardware limitations in the genera-
1
c The Author(s), 2009. Reprints and permissions: tion of magnetic fields. We find that helical propellers and elastic tails
http://www.sagepub.co.uk/journalsPermissions.nav have very comparable performance, and they generally become more
Figures 1, 3, 5–10 appear in color online: http://ijr.sagepub.com

1434

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

 Abbott et al. / How Should Microrobots Swim? 1435

desirable than gradient pulling as size decreases and as distance from

the magnetic-field-generation source increases. We provide a discus-
sion of why helical propellers are likely the best overall choice for in
vivo applications.

KEY WORDS—microrobot, magnetic, wireless, untethered,

medical, in vivo

1. Introduction
Microrobots have the potential to dramatically change many
Fig. 1. Swimming microrobots with a 30:1 size differ-
aspects of medicine by navigating through bodily fluids to
ence. Magnetic fields are used to: (a) pull a 900-1m-
perform targeted diagnosis and therapy. Microrobots, like mi-
long assembled-MEMS microrobot (Yesin et al. 2006)1 and
croorganisms, swim in a low-Reynolds-number regime, requir-
(b) rotate a 30-1m-long artificial-bacterial-flagella microrobot
ing swimming methods that differ from macroscale swimmers.
(Zhang et al. 2009).
Microrobots can also vary in size by orders of magnitude. Re-
searchers have proposed numerous microrobotic swimming
methods, many biomimetic, with the vast majority utilizing
magnetic fields to wirelessly power and control the microro- To understand this phenomenon, we turn to the Navier–Stokes
bot. At ETH Zurich we have developed two different magnetic equations, which, when combined with boundary conditions,
microrobots that utilize very different propulsion schemes: one completely define a fluid flow. For a fluid with constant den-
is a submillimeter-sized device that is pulled with magnetic sity 2 and constant viscosity 3, the Navier–Stokes equations
field gradients (see Figure 1(a) and Yesin et al. (2006))1 the are given by a single vector equation, which can be non-
other has a helical propeller that mimics a bacterial flagellum dimensionalized in space and time by the magnitude of some
in both form and scale and is rotated with a magnetic field characteristic velocity U and some characteristic length 4:
(see Figure 1(b) and Zhang et al. (2009)). Another propul-
1 2 2
2U 4 2 78 Re 3 2U 4 5
sion scheme proposed in the literature utilizes an elastic tail dV
that is wiggled behind a magnetic head (Sudo et al. 20061 3 45 p2 6 5 2 V (1)
3 dt 3
Guo et al. 2008). It remains unclear which propulsion method
is optimal, and a comparison that considers the microrobot’s Here V 2 is the velocity vector field and p2 is the hydro-
size coupled with the practical limitations in generating mag- dynamic pressure scalar field, which have both been non-
netic fields is needed. In this paper, we compare these three dimensionalized as in White (1991). From this equation we
propulsion methods, which are representative of the majority discover the Reynolds number, the dimensionless quantity that
of active research in wireless microrobots. We show that met- embodies the interaction between a fluid’s inertia and viscos-
rics of efficiency, which are often used to characterize low- ity. At low Re, we are in a world that is very viscous, very
Reynolds-number swimming, can be quite misleading, and slow, or very small. Low-Re flow around a body is referred
that practical limitations in magnetic control have a major im- to as creeping flow or Stokes flow. We no longer see a tran-
pact on which method is best for a given application. We show sition to turbulence, even behind bluff bodies. At low Re, the
that helical propellers and elastic tails have very comparable role of time becomes negligible in (1)1 the flow pattern does
performance, and they generally become more desirable than not change appreciably whether it is slow or fast, and the flow
gradient pulling as size decreases and as distance from the is effectively reversible. Consequently, reciprocal motion (i.e.
magnetic-field-generation source increases. We also find that body motion that simply goes back and forth between two
limitations in the hardware used to generate the magnetic fields configurations) results in negligible net movement.
can influence which swimming method is best. In the end, we Microorganisms are able to swim at low Re using a variety
provide a discussion of why helical propellers are likely the of techniques (Brennen and Winet 19771 Vogel 20031 Lauga
best overall choice for in vivo applications. and Powers 2009), none of which look like macroscale swim-
mers (Figure 2). Cilia are active organelles that are held per-
pendicular to the flow during the power stroke and parallel to
2. Swimming at Low Reynolds Number the flow during the recovery stroke. Many cilia are used simul-
taneously. Eukaryotic flagella are active organelles that deform
It has long been known that swimming at the microscale re- to create paddling motions, such as traveling waves or circular
quires techniques that are very different from those used by translating movements. Bacterial (prokaryotic) flagella work
macroscale swimmers such as fish and humans (Purcell 1977). differently by using a molecular motor to turn the base of a

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

1436 THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH / November/December 2009

Fig. 2. Locomotion of microorganisms. The cells are drawn

identically to highlight the differences in the propulsion meth- Fig. 3. Magnetic swimming methods. In each case, the micro-
ods, but there is a wide range of shapes and sizes of microor- robot is swimming from right to left. In each case, the entire
ganisms. (a) Cilia move across the flow during the power microrobot body is depicted as a magnet, but it is also possible
stroke, and fold near the body during the recovery stroke. for the magnet to be rigidly embedded in a larger body. (a) Gra-
(b) Eukaryotic flagella create patterns such as traveling waves. dients in the magnetic field pull the microrobot directly. (b) A
(c) A molecular motor spins a passive bacterial flagella. rotating uniform magnetic field is transduced into forward mo-
tion using a helical propeller. (c) An oscillating uniform mag-
netic field is transduced into forward motion using an elastic
tail.
passive flagellum. Some bacteria have multiple flagella that
bundle during swimming. All of the swimming methods uti-
lized by microorganisms capitalize on the difference in drag through natural selection, might outperform biomimetic meth-
on a slender body normal and parallel to the slender direction ods.
as it is pulled through fluid. All of the methods are fairly in-
efficient (efficiency is discussed in detail in Section 7).
A number of robotic swimming methods have been shown 3. Magnetic Power and Control
to work at relatively small scales, but will have reduced effec-
tiveness as size decreases to the microscale because they make If we want to apply controlled forces and torques to a body
use of reciprocating configurations. Other biomimetic swim- with average magnetization M using a controlled magnetic
ming methods utilize physics that scale well to the microscale, field H (both quantities are vectors in amps per meter), the
but require mechatronic components that present challenges in governing control equations are as follows (Jiles 1991). The
microfabrication and wireless power and control (Behkem and magnetic torque tends to align the magnetization of the body
Sitti 20061 Kósa et al. 2007, 2008). A number of microrobots with the applied field:
are inspired by nature and also utilize techniques that facilitate
microfabrication and wireless power and control. Nearly every T 3 10 6M 9 H7 (2)
one utilizes magnetic fields (see Figure 3). No other actuation
where 6 is the volume of the body in cubic meters and 10 3
principle offers the ability to transfer such large amounts of
48 9 1047 T
m A41 is the permeability of free space (here the
power wirelessly. For example, a rotating magnetic field is an
non-bold T represents the unit Tesla). The magnetic force on
obvious choice to rotate a helical propeller (Honda et al. 19961
the body is
Zhang et al. 2009), eliminating the need to replicate a mole-
F 3 10 6 9M
5
H5 (3)
cular motor in a microrobot. However, even with all of their
positive attributes, magnetic fields impose strict limitations on We can also express the applied magnetic field as an applied
the design of wireless microrobots. magnetic flux density B in Tesla. Here B is related to H simply
A controllable external pulling source is not available to as B 3 10 H, since air and biological materials are effectively
microorganisms, but engineers can utilize gradients in mag- non-magnetic. Both (2) and (3) are based on the assumption
netic fields to apply forces and torques to untethered microro- that the magnetic body is small compared with spatial changes
bots (Mathieu et al. 20061 Yesin et al. 2006). This greatly sim- in the applied field, such that the applied field is fairly uniform
plifies fabrication since no microactuator or special structure across the body, and H is the value of the applied field at the
is needed for propulsion. Biomimetic methods will clearly be center of mass of the body. We have verified in prior work that
effective, but it is also reasonable to wonder whether this form this assumption gives an accurate prediction of magnetic force
of direct-pulling propulsion, which could not have evolved and torque (Abbott et al. 20071 Nagy et al. 2008).

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

 Abbott et al. / How Should Microrobots Swim? 1437

If the body of interest is a permanent magnet, the aver-

age magnetization M is effectively constant with respect to
the body with a magnitude governed by the remanent mag-
netization of the material. We can increase torque by increas-
ing the angle between H and M, up to 90 , or by increasing
the strength of H. We can increase force by increasing the
gradients in the applied field. If the body of interest is made Fig. 4. Definition of parameters for a microrobot with a helical
of a soft-magnetic material, the magnetization is a non-linear propeller with right-handed chirality. The force f and torque
function of the applied field, with a magnitude bounded from represent the sum of all non-fluidic loads.
above by the saturation magnetization value for the material,
and can rotate with respect to the body1 the governing equa-
tions for control are significantly more complex (Abbott et al.
20071 Nagy et al. 2008). We sometimes refer to the “magnetic F 3   V5 (4)
moment” or “magnetic dipole moment”, which represent the If we consider a spherical bead of diameter d, the translational
total strength of a magnet (permanent or soft). The magnetic drag coefficient is described in Stokes flow (see White (1991))
moment is simply the product of the volume 6 and the average as
magnetization M.   3 383d5 (5)
Every magnetic body has geometric effects that make its
magnetic properties differ from those of the bulk material and It is clear that velocity is inversely proportional to fluid viscos-
different along different directions within the body. This is ity with all other parameters held constant.
known as shape anisotropy. Demagnetizing fields are created The magnetic force of (3) also simplifies in the case of a
that tend to weaken magnetization, and demagnetizing fields soft-magnetic spherical bead to
are largest along short directions of the body. A long direction
in a body is referred to as an easy axis, since it is a relatively F 3 10 6M 95H
5 (6)
easy direction in which to magnetize the material. Other types
This equation also applies to a permanent magnet under the
of anisotropy exist, such as crystalline anisotropy, but these
assumption that the magnetization M is always aligned with
are typically negligible compared with shape effects, even at
the applied field H.
the scale of microrobots.
Controlled magnetic fields can be generated by stationary
current-controlled electromagnets (Meeker et al. 19961 Zhang
et al. 2009), such as the system shown in Figure 1(b), by elec-
5. Swimming with a Helical Propeller
tromagnets that are position and current controlled (Grady et
The next method of wireless magnetic swimming that we con-
al. 19901 Yesin et al. 2006), such as the system shown in Fig-
sider is swimming with a helical propeller, which is depicted
ure 1(a), by position-controlled permanent magnets, such as
in Figure 3(b). Consider the helical propeller with parameters
with the Stereotaxis Niobe Magnetic Navigation System, or
defined as shown in Figure 4. Swimming along the axis of a
even by a commercial magnetic resonance imaging (MRI) sys-
helical propeller is described by a symmetric propulsion ma-
tem (Mathieu et al. 2006). In all cases, the rapid decay of mag-
trix (Purcell 1977, 1997) relating the four principle (scalar)
netic field strength with distance from its source creates a ma-
quantities, forward velocity , angular velocity , non-fluidic
jor challenge for magnetic control.
applied torque , and non-fluidic applied force f , of the helical
propeller: 3 5 3 53 5
4. Pulling Through Fluid with Field Gradients f a b 
4 634 64 67 (7)
b c 
The first method of wireless magnetic swimming that we con-
sider is simply pulling a rigid object through fluid using mag- where the matrix parameters are computed as
netic field gradients, which is depicted in Figure 3(a). Al- 7 8
though this is a valid method to actively move through fluid  cos2  6   sin2 
a 3 28n 7 (8)
that we would like to consider, it is not technically “swim- sin 
ming” since it does not use the fluid to assist in propulsion.
When pulling a magnetic object through Newtonian fluid at 9

b 3 28n 2  4   cos  7 (9)

low Re, the object nearly instantaneously reaches its terminal
7 8
velocity V where the viscous drag force, which is linearly re-   cos2  6  sin2 
lated to velocity through a drag coefficient   , exactly bal- c 3 28n 3
7 (10)
sin 
ances the applied magnetic force F:

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

1438 THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH / November/December 2009

where n is the number of turns of the helix (such that the helix
length is n), and the constants   and  are the viscous drag
coefficients for a thin cylindrical element of a helical propeller
for flow locally normal to the cylinder’s axis and along the
length of the cylinder’s axis, respectively, given by Lighthill
(1976) as
483
 3 9 05368 
(11)
ln r sin 
6 055

283
 3 9 05368 
(12)
ln r sin 

The effect of the rigidly attached head is to modify the propul-

sion matrix of the helical propeller of (7) to
3 5 3 53 5
f a 6  b 
4 634 64 67 (13)
b c 6  
Fig. 5. Qualitative behavior of helical-propeller swimming,
where   and   are translational and rotational drag co- with rotational frequency as the control variable, which is par-
efficients for the head, respectively. If we consider a spherical ticularly useful when considering helical propellers driven by
head, the rotational drag coefficient is described in Stokes flow applied magnetic fields. Parameters are defined in (15) and
(see White (1991)) as Figure 4.

  3 83d 3 (14)

and the translational drag coefficient is given in (5). et al. 2009). The step-out frequency is a function of max , the
In the case of magnetically applied torques, the rotational maximum magnetic torque that can be generated as described
frequency of the magnetic field is the fundamental control in- in (2), as depicted in Figure 5.
put. The microrobot rotates synchronously with the applied It is notable that the fluid viscosity 3 enters into (8)–(10)
field, nearly instantaneously reaching an equilibrium phase linearly. Consequently, 3 enters into  linearly, but does not
shift such that the magnetic torque perfectly counterbalances enter into . Taking this into consideration, from Figure 5 we
the fluidic drag torque. It is more instructive in this case to re- find that the effect of doubling viscosity with all other parame-
arrange the linear equations with the non-fluidic applied force ters held constant would be to reduce the step-out frequency
f and angular velocity  as the input variables: by half and, consequently, to reduce the maximum velocity by
3 5 3 53 5 half. That is, the maximum velocity is inversely proportional
   f to fluid viscosity.
4 634 64 67 (15)
4  
6. Swimming with an Elastic Tail
where (13) and (15) are related by
1 4b The final method of wireless magnetic swimming that we con-
3 7 3 7 sider is swimming with an elastic tail (sometimes referred to as
a 6  a 6 
a “flexible oar”), which is depicted in Figure 3(c). A magnetic
b2 field is oscillated in time, and a magnetic torque is applied
 3 c 6  4 5 (16) to the magnetic head as it attempts to align with the applied
a 6 
field. There have been a few experimental demonstrations of
Figure 5 shows the behavior observed with this type of swim- this type of microrobot swimming (Sudo et al. 20061 Guo et al.
ming. The forward velocity grows linearly with frequency un- 2008), but the deformable nature of the elastic tail makes even
til a step-out frequency is reached. Beyond this step-out fre- basic analysis of this type of swimming significantly more
quency, the available magnetic torque is no longer sufficient complicated than that of the rigid helical propeller, and there
to keep the microrobot rotating in sync with the applied field, currently exists no analytical model for this method. The clos-
and a drastic and non-deterministic decrease in the swimming est is the free-swimming elastic tail being driven by a pure
velocity is observed. This behavior has been demonstrated ex- torque at one end considered analytically for small deforma-
perimentally at a number of scales (Honda et al. 19961 Zhang tions in Wiggins and Goldstein (1998) and numerically for

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

 Abbott et al. / How Should Microrobots Swim? 1439

large deformations in Cosentino Lagomarsino et al. (2003),

with the differences to the case of Figure 3(c) being that in
their case a constraint on a fixed angular deflection was con-
sidered, and that the application of the torque had no associ-
ated additional viscous effects (i.e. no increased drag due to
an increased magnetic bead size associated with an increase in
magnetic torque). The same authors also considered the case
where the end of an elastic filament is driven with an oscillat-
ing lateral movement of fixed amplitude (Wiggins and Gold-
stein 19981 Wiggins et al. 19981 Cosentino Lagomarsino et al.
2003).
Let us consider the case of the elastic tail driven by a pure
oscillating torque described by 3 0 sin9t
. The elastic
tail has a length l, a bending stiffness , and a perpendicular
viscous drag coefficient of   . We are interested in the net
swimming velocity . From dimensional analysis, we find that
the dimensionless velocity

13 (17)
l
is a function of two dimensionless parameters: one is com-
monly referred to as the Sperm number
1 2
  14
Sp 3 l 7 (18)

which encapsulates the waveform, or “floppiness”, of the tail,
and the other is the dimensionless torque magnitude
0l
2 3 5 (19)

Fig. 6. Dimensionless results for numerical simulations of an
Figure 6(a) shows the results of numerical simulations us- elastic tail being driven by a pure sinusoidal torque at one end.
ing the same methods as presented in Cosentino Lagomarsino (a) Dimensionless velocity versus the Sperm number for vari-
et al. (2003). It shows how 1 varies with Sp for various values ous value of constant dimensionless torque. (b) Dimensionless
of constant 2 . We find that 2 affects 1 linearly, which allows velocity normalized by dimensionless torque versus the Sperm
us to plot a single dimensionless quantity 12 as a function number, showing that dimensionless torque affects dimension-
of Sp in Figure 6(b). The results indicate that a peak value in less velocity linearly.
dimensionless velocity occurs at Sp  453, regardless of the
value of 2 .
For a simple model of bending stiffness, we can consider
that of a cylindrical beam of radius r and Young’s modulus E: hardware that generates the magnetic field. In practice, there
will be limits to how close the field sources can be placed
8r 4 E to the microrobot. Let us compare the control of a microro-
3 5 (20)
4 bot with two current-carrying coils, as shown in Figure 7. Our
field sources are assumed to be located a distance L from our
From Vogel (2003), an appropriate value for   in this case is
microrobot. The limitations imposed by this two-coil system
48 3 are representative of other magnetic control systems, such as
 3 9l
5 (21) those as described in Section 3.
ln r 6 05193

7. A Critical Comparison 7.1. Helical Propeller versus Gradient Pulling

In order to make a fair comparison between microrobots that We begin by comparing swimming with a helical propeller to
utilize magnetic fields in different ways, we must consider the simply pulling with field gradients. We assume a simple mag-

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

1440 THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH / November/December 2009

Fig. 7. Two electromagnetic coils as field-generation hardware

for microrobot propulsion. (a) Uniform magnetic fields in the
center of the workspace are generated by current flowing in
the same direction. The field is calculated accurately with a
current-loop model along the axis of the coils (Jiles 1991).
(b) Gradient fields can be generated by running the current
in opposite directions.

netic bead of diameter d for our microrobot. We have two op-

tions for control: we can attach a helical propeller to our bead
and then apply current in the same direction in the two loops,
creating a uniform field that is used to rotate and thus propel
the bead1 or we can apply the current in opposite directions,
creating a field gradient that is used to pull on the bead directly.
From the equations that describe the magnetic field and field Fig. 8. Comparison of (a) normalized no-load ( f 3 0) veloc-
gradient at the midpoint of the current loops, shown in Fig- ity and (b) normalized stall ( 3 0) force vs. bead radius for
ure 7, we see that the field that we can generate changes with helical propulsion and field-gradient pulling.
the gap between the two loops as  L 43 , while the gradient
that we can generate changes as  L 44 , as expected. We also
find that there are optimal values for the coil radii fora given
 the respective quantities: R 3 2L for
coil gap to maximize imum force varies as  d 2 for helical propulsion and as  d 3
case (a) and R 3 23L for case (b). for pulling. This indicates that there will always exist a micro-
In Figure 8 we plot the maximum no-load ( f 3 0) veloc- robot size below which using helical propulsion is desirable
ity and the maximum stall ( 3 0) force versus bead size. over pulling with field gradients. In addition, even though the
For the helical parameters, we choose  3 d2, r 3 d200, helical parameters used here were not optimized, Figure 8 in-
and  3 45 . For the coil gap, we assume L 3 052 m, which dicates that helical propulsion is desirable (by a large margin)
is the approximate value that would be needed to control a mi- over pulling for any microrobot size that would reasonably be
crorobot somewhere inside a human head. To avoid biasing the considered as microscale. In addition, increasing the distance
results, we use the optimal radii R for the respective cases. The between the magnetic field sources and the microrobot (i.e.
results for two helices are shown: one for n 3 3 and one for L), which will likely be necessary for in vivo applications, will
n 3 10. The magnetization of the bead and the current through always tend to improve helical propulsion relative to pulling,
the loops enter into the maximum velocity and force linearly, although the effectiveness of both will be reduced.
so we can normalize our results to these quantities. The recip- Although helical propulsion of a body is described as
rocal of viscosity enters into the maximum velocity linearly, being quite inefficient compared with simply pulling the
so we can normalize to this quantity as well1 the viscosity has body through the fluid (Purcell 1977, 1997), this is mis-
no bearing on the maximum force. leading in the context of magnetic control. Purcell uses a
We find that the maximum velocity varies as  d for heli- definition of swimming efficiency that compares the power
cal propulsion and as  d 2 for pulling. We find that the max- used to propel a body at a given velocity and the power re-

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

 Abbott et al. / How Should Microrobots Swim? 1441

quired to simply pull the body through the fluid at the same the field-generation hardware, and is governed by (2). Neither
velocity: swimming method relies on field gradients, which means that
both methods will improve relative to gradient pulling as the
power required to simply pull body
3 5 (22) distance to the field-generation hardware is increased.
power consumed during propulsion Let us consider an elastic tail driven by a pure sinusoidal
Under this definition of efficiency, pulling with field gradi- torque at one end, as detailed in Section 6. We assume that the
ents is 100% efficient (i.e.  3 1), and every other swimming tail is being driven optimally, and that we are making use of
method will always perform less efficiently. The problem with the available magnetic torque such that 0 3 max . Although
this definition in this context is that the amount of power that the peak in dimensionless velocity in Figure 6 occurs at Sp 
can be harvested from the applied magnetic field is different in 453, this is not the location of peak velocity, due to the way
each of our cases, so a comparison of the efficient use of that that  affects both 1 and Sp. We find that velocity is actually
power does not give meaningful results. optimized when Sp4 1 is maximized, which we find occurs at
Theoretical models of microrobot swimming should be im- Sp  552, and which corresponds to
proved, but the discrepancy between the models at the mi- 
croscale observed in Figure 8 might be too large to be drasti- 3 154 9 1043 5 (23)
 0l 2
cally affected by improved models. However, there is another
practical limitation in the magnetic control of helical swim- After some substitutions, we can express the maximum swim-
mers that we must consider. Our actual control variable is ming velocity as
the rotation frequency  of the magnetic field. As described 1 2
150
in Section 5, the microrobot rotates in sync with the field,  max 3 max 5 (24)
 l 2
with the field leading the magnetization such that the magnetic
torque from (2) balances the viscous torque. As  is increased, With an understanding that the maximum magnetic torque will
the magnetic torque eventually reaches its step-out frequency, scale with the volume of material, and with all other parame-
above which the microrobot can no longer track the rotating ters held constant, we find that  max scales with microrobot
field. The maximum velocity and force curves of Figure 8 as- size as  l. We also find that  max is inversely proportional to
sume that the microrobot is turning at its step-out frequency, fluid viscosity. We can also consider the case of a helical pro-
which also increases linearly with the quantity Mi3. For peller being driven by a pure torque. For the helical propeller
any field-generation system, there are practical saturation lim- of (7), we compute
its in generating high  due to, for example, induction, eddy 1 2
currents, motor speeds, or sampling rates. If this -saturation  max 3
b
max 5 (25)
is reached, the values shown in Figure 8 will over-estimate b2 4 ac
the capabilities of helical propulsion. We must also recall that
We again find that  max scales with microrobot size as  l, and
the propulsion model implicitly assumes low-Re flow, which
that  max is inversely proportional to fluid viscosity.
could be violated for very high . Thus, although it is always
The performance of both helical propellers and elastic tails
desirable to increase M and i for improved control using ei-
scale in the same way. However, we have not yet considered
ther swimming method, it could reduce the performance of he-
how the driving frequency  changes with scale. To main-
lical propulsion relative to pulling. It should be noted, however,
tain the elastic tail swimmers running at peak performance,
that the step-out frequency is scale invariant (discussed more
we must achieve 1 2
in the next section), and published experimental results for he-
 552 4
lical swimmers have typically had step-out frequencies below 3 5 (26)
60 Hz, indicating that this high- problem is unlikely to sig-  l
nificantly impact on our conclusions. Assuming that the stiffness  varies as  l 4 (see (20)), we
find that the peak value for  does not change as we scale
the elastic-tail microrobot. If, however, the length of the tail
7.2. Helical Propeller versus Elastic Tail is changed without changing the cross section, we find that 
varies as  l 44 , which would lead to a significant increase in
To compare helical-propeller swimming with elastic-tail optimal driving frequency if we shorten the length of the tail.
swimming in the context of wireless magnetic control, we For the helical-propeller swimmer, for optimal performance
make use of the fact that both methods transduce mechani- we must run the swimmer at the step-out frequency, which is
cal power from the applied magnetic field in the same way: the maximum frequency max that can be achieved for a given
through magnetic torque. So, for a given magnetic bead driven max , and which is described in free swimming by
by either type of propeller, we can simply assume some maxi-
1 2
mum available torque max available to the microrobot, which a
is a function of the bead geometry and material as well as max 3 max 5 (27)
ac 4 b2

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

1442 THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH / November/December 2009

For the helical-propeller, we find that the driving frequency outperform field-gradient-pulled microrobots once we con-
does not change as we scale the microrobot size. sider the practical limitations involved in generating magnetic
We have shown that the performance of these two methods fields.
of microrobot swimming scale identically to one another, as- From a fabrication standpoint, constructing an elastic-tail
suming the cross section of the elastic tail scales proportionally microswimmer seems feasible, considering the relatively sim-
to all other parameters. This means that if one type of propul- ple design. However, this has not been done to date (prior
sion outperforms the other, it will always outperform, regard- experiments have considered up-scaled models), and it is not
less of scale. However, we would still like to know which, if clear which materials would be the best choice for the tail. It
either, swimming method is superior. To do this, let us con- has already been shown that it is possible with currently avail-
sider an example. If we construct a 2-mm-long helical swim- able technology to microfabricate helical-propeller magnetic
mer with a pitch of  3 45 , four turns, and a cross sec- microrobots (Zhang et al. 2009). This has been accomplished
tion with r 3 50 1m, we compute a maximum velocity of with nanocoils, which are rolled-up pre-stressed multilayer
 max 3 9757 9 104
max 3, with a corresponding angular ve- strips. This technique has better control over helical geome-
locity of max 3 9753 9 109
max 3. Note that both max and try than grown helical carbon nanotubes or ZnO nanobelts.
the reciprocal of fluid viscosity enter into these terms linearly, The radius of the coil is determined by the thicknesses of
and can therefore be factored out. For an elastic-tail swimmer the films, the Young’s moduli of the materials, and the lat-
with 2-mm length and a cross section with r 3 50 1m, we tice mismatches of the layers. Figure 9(a) shows the process
again calculate  max 3 9757 9 104
max 3. It is a coincidence sequence to fabricate nanocoil microrobots, which consist of
that these two peak-velocity values are identical, and we find a 42-nm-thick ribbon that, upon wet etch release, self-forms
that by varying parameters we can make either type of mi- into a 3-1m-diameter coil with a length of 30–40 1m. A
croswimmer outperform the other. However, this example does 455 1m 9 455 1m 9 052 1m Ni plate is formed on one end that
give us an indication that we should expect similar peak per- serves as a “head”. The geometrical relation of the nanocoil
formance from these two types of microswimmers, assuming parameters is shown in Figure 9(b). The width of the stripe is
the same length and the same maximum magnetic torque. We given by its initial pattern design, and the depth is controlled
have compared our numerical models with prior experimental during fabrication. Figure 9(c) shows a scanning electron mi-
results (Sudo et al. 20061 Zhang et al. 2009) and find excellent croscopy (SEM) image of an as-fabricated nanocoil with a
agreement. These prior experiments also indicate that optimal Ni plate on one end. For a first propulsion experiment, indi-
driving frequencies are comparable between the two types of vidual magnetic nanocoils were immersed in water, actuated
microswimmers, and also feasible from a hardware standpoint with a rotating magnetic field (see hardware in Figure 1(b)),
(e.g. 30–60 Hz). and their motion was captured on video through a micro-
It may be the case that comparing microswimmers of the scope. One sequence for a 40-1m-long nanocoil is shown in
same length biases the comparison in favor of the helical pro- Figure 9(d).
peller. In some cases it may be acceptable to shorten the length Microrobots that swim using helical propellers have a num-
of the elastic tail to increase the forward velocity, as indicated ber of additional potential benefits for use as in vivo medical
in (24). However, we know that this shortening may drastically devices that we believe makes them the most promising as a
increase the driving frequency, as indicated in (26). technology worth pursuing. They include the following.
In our comparison of peak velocity, we found that max
could be factored out for both microswimmers. However, we 1. Reversing direction is simple with a rigid helical pro-
find that the optimal  for helical propulsion is linearly af- peller. It simply entails reversing the rotation direction
fected by the non-scale-related contribution of max (i.e. the of the magnetic field. This could be particularly useful
field strength and the magnetization of the material), whereas in retracing a path already taken. Microrobots that are
the optimal  for elastic tails is completely independent of pulled with field gradients are also easily reversed. How-
max . This indicates that, as we reduce the strength of the ap- ever, to reverse the swimming direction of an elastic-
plied magnetic field, the optimal driving frequency of the he- tail microrobot, the microrobot must turn completely
lical swimmer will reduce relative to that of the elastic-tail around, which is less efficient and potentially more
swimmer, which may be desirable from a hardware and control difficult to accomplish.
standpoint.
2. A rigid helical propeller can be functionalized (e.g.
coated) without significantly changing its fluid-dynamic
8. Discussion properties. This is not likely true of an elastic tail, whose
bending properties will change due to functionalization.
It appears that, under some reasonable simplifying assump- Adding extra fluid drag will always reduce performance
tions, helical-propeller microrobots and elastic-tail microro- with gradient-pulling methods. It has also been sug-
bots have very comparable peak performance, and both will gested that the microrobot’s payload, such as a strip of

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

 Abbott et al. / How Should Microrobots Swim? 1443

the urethra, the microrobot crawls toward the bladder

by pressing its helical propeller against the partially col-
lapsed walls of the urethra. The microrobot does not
swim in this case, but actually crawls forward, ideally
advancing by one pitch per rotation, but with the possi-
bility of some amount of slip. This type of helical lumen
crawling has been explored previously (Sendoh et al.
2003). Once the microrobot reaches the bladder, it tran-
sitions into the type of swimming discussed in this pa-
per. The microrobot could even swim toward the open-
ing of the ureter and transition back to crawling to ad-
vance toward the kidney. It is unclear whether an elas-
tic tail or gradient-pulling would be effective inside a
lumen whose size is comparable to that of the micro-
robot and potentially even collapsed. A driven elastic
tail makes lateral movements throughout its cycle (with
no net lateral movement) as it is propelled forward. A
wall constraint, such as a lumen environment, will affect
the swimming pattern and performance of an elastic-tail
swimmer compared with the model described in Sec-
tion 6.

4. Swimming with helical propulsion opens up the pos-

sibility of control in non-uniform magnetic fields. Al-
though the analysis presented in this paper assumes a
uniform (gradient-free) magnetic field, which in prac-
tice is developed in the center of two equally power-
ful electromagnets as depicted in Figure 7(a), it is also
possible to swim against magnetic field gradients pro-
vided that the propulsive force of the helical propeller is
Fig. 9. Nanocoils as artificial bacterial flagella (Zhang et al. large enough. We have already demonstrated this exper-
2009). (a) Microfabrication process sequence: an initially pla- imentally with the microrobots of Figure 9 by using the
nar epitaxial trilayer with a plate on one end consisting largely field of a single rotating permanent magnet. This possi-
of soft-magnetic Ni, all patterned through conventional mi- bility of using non-uniform fields allows us to be more
crofabrication techniques, self-forms into a three-dimensional creative in the design of field-generation hardware, and
structure during wet etch release. (b) Schematic drawing show- may prove particularly useful in the control of in vivo
ing the relationship between the stripe and etching direction in microrobots. Swimming against field gradients may also
nanocoil parameters. (c) SEM image of an as-fabricated micro- be possible with an elastic tail, however, the inability
robot. (d) Optical microscope image sequence of a microrobot to reverse direction easily (discussed above) makes the
swimming in water. Arrows indicate the forward direction and design of field-generation hardware more challenging.
corresponding rotation being commanded by the electromag- Swimming against a field gradient is never possible with
netic system. gradient-pulling methods.

5. The same magnetic control system that is designed

to control helical swimmers can also be used to con-
drug, can be twisted up into a helical propeller to actu- trol screw-type microrobots, which are microrobots that
ally contribute to propulsion (Li et al. 2006). An elastic consist of a magnet rigidly embedded in a microrobot
tail might also be made of drug, provided the elasticity shaped something like a wood screw (Ishiyama et al.
and stiffness of the drug can be fabricated correctly. 2003). These screw-type microrobots provide a promis-
ing method to move through soft tissues such as brain
3. Microrobots that swim using helical propellers have the and liver. With the ability to control two types of micro-
potential to smoothly transition from lumen to open robots with the same hardware, the probability of both
environments. For example, consider a microrobot de- of these two technologies taking hold in clinical prac-
signed for use in the urinary system. After insertion into tice increases, since costs, both in terms of finances and

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

1444 THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH / November/December 2009

space, will be distributed over multiple potential proce-

dures.

There are a number of modeling assumptions used in this

paper that could be improved to increase the accuracy of the
models1 the most salient are discussed below. However, it is
unlikely that these assumptions will significantly affect the
conclusions that we have reached.

1. The model in Section 5 assumes that the head and the

helical propeller do not affect each others’ flow fields,
which clearly cannot be true. The model in Section 6
does not account for any fluid drag on the magnetic
head.
2. The individual Stokes-flow models are all derived as-
suming an infinite extent of fluid. Figure 10 shows mag-
netic and hydrodynamic data for three steel beads, ob-
tained with a custom measurement system (Kummer et
al. 2007). We see that the magnetic models are quite ac-
curate, but the viscous drag coefficient   for Stokes
flow around a sphere is less accurate, due to the wall
effects of the container (i.e. fluid viscosity effectively
increases nearer to a solid boundary). Wall effects have
also been observed to affect helical propulsion (Behkem
and Sitti 2006), elastic-tail propulsion (Yu et al. 2006),
and are generally important in low-Re swimming (Vogel
20031 Lauga and Powers 2009). From Figure 10(b), the
speed of gradient-pulled beads is clearly reduced due to
wall effects. The cases of the other two swimmers that
we consider are more complicated. The efficiency of
both helical-propeller swimmers and elastic-tail swim-
mers are governed by the ratio of the fluid drag co- Fig. 10. Experimental results for three steel beads using the
efficients of the propeller:    . That is, swimming be- custom measurement system of Kummer et al. (2007). Data
comes more efficient in cases when the fluid drag for originally appeared in Ergeneman et al. (2008). (a) Mag-
flow perpendicular to an element of the propeller in- netic forces are accurately predicted using a theoretical model.
creases relative to fluid drag for flow along an element (b) Hydrodynamic predictions of the beads being pulled
of the propeller. It has been shown that wall effects tend through silicon oil (2 3 970 kg m43 , 3 3 0598 Pa
s) are less
to increase this drag-coefficient ratio, and thereby in- accurate, as Stokes-flow equations do not accurately account
crease efficiency (Brennen and Winet 1977). However, for wall effects of the 26-mm-diameter container.
this is another use of “efficiency” that can be misleading
in the case of magnetic control. This use of “efficiency”
relates to how much forward velocity results from a in swimming speed near solid boundaries for both the
given propeller movement, but does not account for the helical-propeller and elastic-tail microrobots.
power that is required to generate said movement. In the
case of constant-power swimming, the presence of walls 3. Our swimming models and critical comparison all con-
typically tends to decrease swimming speed in microor- sider one-degree-of-freedom motion. In practice, it will
ganisms (Lauga and Powers 2009). In our case of con- be necessary to steer the microrobots while creating
stant magnetic torque input, we can expect power input propulsion, and the coupled relationship between mag-
to decrease with increased fluid drag, since power in- netic steering and propulsion is not always trivial. In ad-
put is the product of torque and angular velocity, and dition, the helical swimmer and the elastic-tail swimmer
angular velocity will reduce with increased drag for a must use non-holonomic steering techniques, whereas
given torque. In addition, the magnetic head will expe- gradient pulling has the potential for control authority
rience an increase in fluid drag similar to the gradient- in every direction. Walls also affect steering. As viscous
pulled bead. Consequently, we should expect a decrease drag increases nearer to a wall, rotating bodies tend to

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

 Abbott et al. / How Should Microrobots Swim? 1445

Fig. 11. (a) Simple helical propulsion results in drift due to

gravity. (b) Swimming similar to helicopter flight can counter-
act gravity.

Fig. 12. A nanocoil attached to the chip, similar to that shown

roll along walls, even when there is no direct contact. It in Figure 9(c), is subjected to a magnetic torque and character-
is known that some bacteria tend to swim in circles near ized as a spring in Bell et al. (2007).
solid boundaries, which is due to the counter-rotation
of the bacterium and its flagellum, each tending to roll
along the wall in opposite directions, resulting in a net
torque (Lauga et al. 2006). We have observed an anal- a nanocoil still attached to the chip, like that seen in Fig-
ogous behavior in the magnetic helical microswimmers ure 9(c), subjected to a magnetic field and behaving like
of Figure 9: they tend to drift sideways along a solid a spring1 the nanocoil is clearly not rigid. The charac-
boundary as they move forward, rather than swim in cir- terization of the stiffness of the helical propellers (Bell
cles, because the magnetic head and the helical propeller et al. 2006) can be used for improved helical-propulsion
do not counter-rotate. models.
4. There is another factor that must be considered in micro-
Finally, it would be interesting to compare another type
robot swimming: gravity. It is often claimed that inertia
of magnetic propulsion scheme with the three methods pre-
is negligible at low Re, but “inertia” is used differently
here than “mass”. A microrobot will reach its terminal sented in this paper. Dreyfus et al. (2005) assembled a chain
velocity in a fluid nearly instantaneously, and the termi- of paramagnetic beads, interconnected by DNA, which, when
attached to a body, can be used for propulsion. The microrobot
nal velocity decreases with size. However, even at low
is propelled by an oscillating magnetic field, and the hardware
Re, a microrobot will still fall downward under its own
weight. The propulsion models developed in Sections 5 requirements for magnetic field generation are the same as that
and 6 assume that flow is along the axis of the pro- for elastic-tail propulsion. The oscillating field induces some-
thing like a wave in the chain that travels from the distal end
peller. However, this only provides for propulsion along
toward the body, resulting in a propeller that pulls the body
the axis, which will be insufficient to counteract gravity
when swimming horizontally, as shown in Figure 11(a). behind it, as opposed to the elastic tail, which pushes the mag-
It may be necessary to swim with an angle of attack, netic body. One desirable property of this type of propulsion
is that the attached body can be any payload, and does not it-
as shown in Figure 11(b), and we have, in fact, experi-
self need to be magnetic in any way (in Dreyfus et al. (2005)
mentally found this to be the case (Zhang et al. 2009).
Microorganisms that swim using flagella have a density they pull a red blood cell). An undesirable property of this type
similar to water, and are thus nearly neutrally buoyant, of propulsion is that the chain of paramagnetic beads theoret-
ically requires an attached body for any propulsion to occur.
making this effect of gravity on propulsion unique to mi-
Subsequent work has shown that defects in the chain can cre-
crorobots that are constructed of denser-than-water ma-
terials. ate the asymmetry needed to break the “scallop theorem” and
result in some net movement without an attached body (Roper
5. It is possible to combine the benefits of field gradients et al. 2006). However, it is not clear how such a defect would
with those of helical propulsion or elastic-tail propul- affect the swimming efficiency with an attached body. It is also
sion. That is, we can pull as we rotate. This hybrid not clear how well the fabrication of these microswimmers,
method has the potential for improved performance, and which was successfully performed at very small scales under
may be particularly useful in counteracting the effect of a microscope, scales up to the type of microrobot dimensions
gravity. However, it will require a more complicated ac- that might be of more practical use for in vivo medical appli-
tuation scheme than that considered in Figure 7(a). cations. A critical comparison of the Dreyfus microswimmer
with other common microswimming methods is an interesting
6. Helical-propulsion models typically assume a perfectly topic for future consideration, and a recent work (Roper et al.
rigid helix. In Figure 12 we show experimental data for 2008) may form the basis for such a comparison.

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

1446 THE INTERNATIONAL JOURNAL OF ROBOTICS RESEARCH / November/December 2009

9. Conclusions Brennen, C. and Winet, H. (1977). Fluid mechanics of propul-

sion by cilia and flagella. Annual Review of Fluid Mechan-
Magnetic fields provide an unequaled means of wireless power ics, 9: 339–398.
and control for microrobots. However, the strength of magnetic Cosentino Lagomarsino, M., Capuani, F. and Lowe, C. P.
fields and field gradients decreases rapidly with distance from (2003). A simulation study of the dynamics of a driven
the source, which has a profound impact on the best way to filament in an Aristotelian fluid. Journal of Theoretical Bi-
make use of these fields for microrobot propulsion. Although ology, 224: 215–224.
it has been previously observed that a swimming microrobot Dreyfus, R., Baudry, J., Roper, M. L., Fermigier, M., Stone,
with a helical propeller is far less efficient than simply pulling H. A. and Bibette, J. (2005). Microscopic artificial swim-
the microrobot through the fluid, we find that a helical pro- mers. Nature, 437(6): 862–865.
peller is far superior to pulling if we consider the limitations Ergeneman, O., Dogangil, G., Kummer, M. P., Abbott, J. J.,
of magnetic field sources. We find that optimal performance Nazeeruddin, M. K. and Nelson, B. J. (2008). A magneti-
of helical propellers and elastic tails are very comparable. We cally controlled wireless optical oxygen sensor for intraoc-
find that both generally become preferable to pulling with field ular measurements. IEEE Sensors Journal, 8(1): 29–37.
gradients as microrobot size decreases or as the distance from Grady, M. S., Howard, M. A., III, Molloy, J. A., Ritter, R. C.,
the magnetic field sources increases. Considering additional Quate, E. G. and Gillies, G. T. (1990). Nonlinear mag-
practical considerations, we made the case that helical propul- netic stereotaxis: three-dimensional, in vivo remote mag-
sion will likely be the best choice for in vivo applications. netic manipulation of a small object in canine brain. Med-
However, the design of hardware that realizes the theoretical ical Physics, 17(3): 405–415.
possibilities of any of the propulsion methods remains a chal- Guo, S., Pan, Q. and Khamesee, M. B. (2008). Development
lenging problem. of a novel type of microrobot for biomedical application.
Microsystem Technologies, 14: 307–314.
Honda, T., Arai, K. I. and Ishiyama, K. (1996). Micro swim-
ming mechanisms propelled by external magnetic fields.
Acknowledgments
IEEE Transactions on Magnetics, 32(5): 5085–5087.
Ishiyama, K., Arai, K. I., Sendoh, M. and Yamazaki, A. (2003).
This work was presented in part at the 2007 International Sym-
Spiral-type micro-machine for medical applications. Jour-
posium of Robotics Research. This work is supported in part
nal of Micromechatronics, 2(1): 77–86.
by the NCCR Co-Me of the Swiss National Science Foun-
Jiles, D. (1991). Introduction to Magnetism and Magnetic Ma-
dation. The authors would like to thank Haixin Zhang for
terials. London, Chapman and Hall.
technical help in the experiments shown in Figure 9(d), and
Kósa, G., Jakab, P., Jólesz, F. and Hata, N. (2008). Swimming
Michael Kummer for technical assistance in collecting the data
capsule endoscope using static and RF magnetic field of
in Figure 10.
MRI for propulstion. Proceedings of the IEEE International
Conference on Robotics and Automation, pp. 2922–2927.
Kósa, G., Shoham, M. and Zaaroor, M. (2007). Propulsion
References method for swimming microrobots. IEEE Transactions on
Robotics, 23(1): 137–150.
Abbott, J. J., Ergeneman, O., Kummer, M. P., Hirt, A. M. and Kummer, M. P., Abbott, J. J., Vollmers, K. and Nelson, B. J.
Nelson, B. J. (2007). Modeling magnetic torque and force (2007). Measuring the magnetic and hydrodynamic proper-
for controlled manipulation of soft-magnetic bodies. IEEE ties of assembled-MEMS microrobots. Proceedings of the
Transactions on Robotics, 23(6): 1247–1252. IEEE International Conference on Robotics and Automa-
Behkem, B. and Sitti, M. (2006). Design methodology for bio- tion, pp. 1122–1127.
mimetic propulsion of miniature swimming robots. ASME Lauga, E., DiLuzio, W. R., Whitesides, G. M. and Stone, H. A.
Journal of Dynamic Systems, Measurement, and Control, (2006). Swimming in circles: motion of bacteria near solid
128: 36–43. boundaries. Biophysical Journal, 90: 400–412.
Bell, D. J., Dong, L. X., Nelson, B. J., Golling, M., Zhang, Lauga, E. and Powers, T. R. (2009). The hydrodynamics of
L. and Grützmacher, D. (2006). Fabrication and character- swimming microorganisms. Reports on Progress in Physics
ization of three-dimensional of InGaAs/GaAs nanosprings. (accepted).
Nano Letters, 6: 725–729. Li, H., Tan, J. and Zhang, M. (2006). Dynamics modeling and
Bell, D. J., Leutenegger, S., Hammar, K. M., Dong, L. X. and analysis of a swimming microrobot for controlled drug de-
Nelson, B. J. (2007). Flagella-like propulsion for microro- livery. Proceedings of the IEEE International Conference
bots using a magnetic nanocoil and rotating electromag- on Robotics and Automation, pp. 1768–1773.
netic field. Proceedings of the IEEE International Confer- Lighthill, J. (1976). Flagellar hydrodynamics. SIAM Review,
ence on Robotics and Automation, pp. 1128–1133. 18(2): 161–230.

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

 Abbott et al. / How Should Microrobots Swim? 1447

Mathieu, J.-B., Beaudoin, G. and Martel, S. (2006). Method Sendoh, M., Ishiyama, K. and Arai, K.-I. (2003). Fabrication
of propulsion of a ferromagnetic core in the cardiovas- of magnetic actuator for use in a capsule endoscope. IEEE
cular system through magnetic gradients generated by an Transactions on Magnetics, 39(5): 3232–3234.
MRI system. IEEE Transactions on Biomedical Engineer- Sudo, S., Segawa, S. and Honda, T. (2006). Magnetic swim-
ing, 53(2): 292–299. ming mechanism in a viscous liquid. Journal of Intelligent
Meeker, D. C., Maslen, E. H., Ritter, R. C. and Creighton, Material Systems and Structures, 17: 729–736.
F. M. (1996). Optimal realization of arbitrary forces in a Vogel, S. (2003). Comparative Biomechanics: Life’s Physical
magnetic stereotaxis system. IEEE Transactions on Mag- World. Princeton, NJ, Princeton University Press.
netics, 32(2): 320–328. White, F. M. (1991). Viscous Fluid Flow (2nd edition). New
Nagy, Z., Ergeneman, O., Abbott, J. J., Hutter, M., Hirt, A. M. York, McGraw-Hill.
and Nelson, B. J. (2008). Modeling assembled-MEMS mi- Wiggins, C. H. and Goldstein, R. E. (1998). Flexive and
crorobots for wireless magnetic control. Proceedings of the propulsive dynamics of elastica at low Reynolds number.
IEEE International Conference on Robotics and Automa- Physical Review Letters, 80(17): 3879–3882.
tion, pp. 874–879. Wiggins, C. H., Riveline, D., Ott, A. and Goldstein, R. E.
Purcell, E. M. (1977). Life at low Reynolds number. American (1998). Trapping and wiggling: elastohydrodynamics of
Journal of Physics, 45(1): 3–11. driven microfilaments. Biophysical Journal, 74: 1043–
Purcell, E. M. (1997). The efficiency of propulsion by a ro- 1060.
tating flagellum. Proceedings of the National Academy of Yesin, K. B., Vollmers, K. and Nelson, B. J. (2006). Modeling
Sciences of the United States of America, 94: 11307–11311. and control of untethered biomicrorobots in a fluidic en-
Roper, M., Dreyfus, R., Baudry, J., Fermigier, M., Bibette, J. vironment using electromagnetic fields. The International
and Stone, H. A. (2006). On the dynamics of magnetically Journal of Robotics Research, 25(5–6): 527–536.
driven elastic filaments. Journal of Fluid Mechanics, 554: Yu, T. S., Lauga, E. and Hosoi, A. E. (2006). Experimental in-
167–190. vestigations of elastic tail propulsion at low Reynolds num-
Roper, M., Dreyfus, R., Baudry, J., Fermigier, M., Bibette, ber. Physics of Fluids, 18: 091701.
J. and Stone, H. A. (2008). Do magnetic micro-swimmers Zhang, L., Abbott, J. J., Dong, L. X., Kratochvil, B. E., Bell, D.
move like eukaryotic cells? Proceedings of the Royal Soci- and Nelson, B. J. (2009). Artificial bacterial flagella: fabri-
ety A, 464: 877–904. cation and magnetic control. Applied Physics Letters, 94:
064107.

Downloaded from http://ijr.sagepub.com at EIDGENOSSISCHE TECHNISCHE on October 26, 2009

View publication stats