U Kei Cheang⁎; Dejan Milutinović†; Jongeun Choi‡; MinJun Kim§ ⁎Drexel University, Philadelphia, PA, United States
†University of California, Santa Cruz, Santa Cruz, CA, United States
‡Yonsei University, Seoul, South Korea
§Southern Methodist University, Dallas, TX, United States
Miniaturized robotic swimmers are useful for navigating and performing tasks at micro- and nanoscales. Their realization can revolutionize minimally invasive procedures, for instance, micro- and nanoswimmers can navigate the human body to unclog arteries or delivery drug to solid tumors. The achiral microswimmers, in particular, can be used synergistically with particulate drug delivery systems due their self-assembling fabrication using simple magnetic particles. The achiral microswimmers can swim in bulk fluid and are controlled wirelessly via magnetic fields. One of the cores in developing micro- and nanoswimmers is to use automatous control to improve navigation capability and functionality. To demonstrate effective control over microswimmers, feedback control of the three-bead achiral microswimmers was demonstrated in experiment. This chapter will examine the properties and swimming characteristics of the achiral microswimmer, including the physical properties, hydrodynamics, and kinematics. Constraints and uncertainties will also be closely examined in order to develop a stochastic kinematic model and a nonlinear feedback controller. Uncertainties due to environmental factors such as Brownian motion and unsteady flow conditions are prominent issues to consider for velocity compensation. Finally, this chapter will discuss the successful navigation of achiral microswimmers from any initial conditions to a target position using the feedback controller.
Microrobotics; Nanorobotics; Magnetic control; Low Reynolds number; Chirality; Feedback control
The authors acknowledge Dr. Hoyeon Kim for his invaluable assistance in system setup. The authors would like to thank Prof. Henry Fu and Farshad Meshkati for their contributions and insightful discussions.
The key to utilizing micro- and nanorobotics for drug delivery, minimally invasive surgery, and other biomedical applications is the control and manipulation of micro- or nanorobots in physiological environments. The concept of using micro- and nanorobots to achieve surgical precision on the order of micro- and nanometer has been well explored [1–14]. Each potential biomedical application deals with different physiological environments; thus, each application is followed by a set of unique constraints and challenges. For this reason, the vagueness of mentioning biomedical applications can only serve as a general justification for micro- and nanorobotics. As this field advances, research has branched into the development of specialized robots targeting specific applications, including on-surface transportation [15–17], tissue incision [18], puncture of retinal veins [3], and cell scaffolding [5]. The microrobot reviewed in this chapter is the particle based achiral microswimmer, which was inspired by both particulate drug delivery systems (DDSs) and low Reynolds number hydrodynamics. The achiral microswimmer consists of three firmly connected magnetic particles, which are analogous to drug carriers in particulate DDSs, forming the minimal geometry for low Reynolds number propulsion with the advantage of increased controllability and penetration power. The advantages of using nanoparticles for targeted drug administration include the benefits of nontoxicity, biocompatibility, injectability, and high-level accumulation in the target tissue or organ, to name a few [19]. This serves as a strong motivation due to the harsh side-effects for patients during chemotherapy. Using mechanisms for particulate DDSs and active propulsion systems of micro- and nanorobotics, there is a potential to create new technology to improve therapeutic and clinical values in cancer treatments.
Microswimmers are subject to the low Reynolds number condition. This means that microswimmers in motion ignore inertial effects due to an extremely small ratio of inertial to viscous forces [20]. From the perspective of motion control, a microswimmer can instantly move at the desired velocity and instantly stop. At low Reynolds number, microswimmers must use nonreciprocal motions for locomotion. Following the scallop theorem, most existing microswimmers utilize either chirality or flexibility to generate propulsion [20]. The helical swimmers, inspired by flagellated bacteria, are the common examples of using chirality for microscale propulsion [21–27]. The flexible swimmers, including the sperm-like swimmers with flexible DNA linkage [28] and the nanowire robots [29], are less common than the chiral swimmer due to the need for complicated fabrication and flexible materials. Other example microrobots capable of low Reynolds number locomotion include the electrically- and optically-controlled bacterial microrobots [15], magnetically steered swimming cells [30], optically-deformed 3-bead systems [31], chemically-driven phoretic swimmers [32–34], and biflagellate micro-objects [35].
Aside from motion constraints due to low Reynolds number, environmental factors at the microscale also require important considerations. The movement of microswimmers, both natural and artificial, can be described as a combination of active propulsion, Brownian motion from thermal diffusion, and background fluid [36]. As evidenced in recent studies on control of magnetic micro- and nanoswimmers, the effects of Brownian motion on their motion can be very pronounced to a degree where active propulsion of nanoswimmers is almost suppressed by Brownian motion [37,38]. Unstable fluidic environments are also detrimental if the background flow velocity is similar or higher than the propulsion velocity. Therefore, the motivation of this work is largely inspired by the need for autonomous control systems to effectively control magnetic microswimmers by compensating for the various environmental factors.
The particle based achiral microswimmers consist of three magnetic particles that are firmly connected via chemical conjugation and magnetic attraction. When actuated via a rotating magnetic field, the microswimmers convert rotational motion into translational motion. The achiral microswimmers have a length scale on the order of 10 μm. Due to their achirality, the microswimmers possess an invisible handedness which inherently leads to motion control uncertainties [39]. In this chapter, we will discuss fabrication, properties, and the motion constraints of the microswimmers [39,40]. Then, we will examine the imaging system and the magnetic control system. Next, we will introduce a quantitative kinematic model and a nonlinear controller for feedback control [39]. Finally, we will discuss the use of an Integral (I) controller to compensate for velocity error due to microfluidic environmental factors.
The achiral microswimmers are fabricated by connecting three ferromagnetic beads (4.35 μm, Spherobeads) forming an achiral structure with two planes of symmetry. The beads are connected via avidin–biotin chemistry and magnetic force. The magnetic particles with amino coating (Spherotech) are separated into two batches: the first batch was mixed with 1 mg/mL of purified avidin protein, while the second batch with 25 mg/mL N-hydroxysuccinimide-biotin. Both batches are incubated for 2 h on a shaker at room temperature. Then, the two batches are combined in a reaction mix to yield microswimmers. By adjusting the concentration of the two batches, the aggregation of the particles can be controlled, thus, allowing for a control over the number of beads for the microswimmers. For experiments, samples were qualitatively observed to yield structures with one to three beads. Due to the strong avidin–biotin connection [41–43], microswimmers were able to maintain their original structures during experiment under a time varying magnetic field. A schematic of the fabrication process is shown in Fig. 7.1.
The achiral microswimmers have two important properties, achirality and rigidity. The geometry of the three-particle achiral microswimmers is achiral due to having only two planes of symmetry. The achiral shape with two planes of symmetry is the minimal geometry for low Reynolds propulsion, according to Happel and Brenner [44]. The achiral microswimmer is rigid due to the firm avidin–biotin connection. A rigidity test was performed and reported by Cheang et al. through a series of quantitative measurements of the changes in distance between the three connected beads during swimming [40]. The measurements were done using a 3D tracking algorithm that can track the position of each individual bead of any given swimmer. It was shown that the microswimmers experienced less than 5% changes in the distances during six cycles of rotations.
Though it is not possible to visually distinguish the geometrical handedness of achiral shapes, the magnetic microswimmers are handed as a result of their magnetic dipole. Fig. 7.2(A) illustrates the handedness of the achiral microswimmers which can be defined based on the direction of their magnetic dipole relative to the orientation of the microswimmers. Much like for the helices, the handedness plays an important role in determining the direction of motion; specifically, whether they will swim forward or backwards due to clockwise or counterclockwise rotation. Unlike for helices, the magnetic handedness of the achiral microswimmers can lead to uncertainties in the swimming direction. For instance, a reversal of rotation direction does not lead to a reversal in swimming direction for some of the microswimmers. This is attributed to geometrical variations due to randomness from the manufacturing process leading to different swimming behavior among microswimmers. As a result, the motion of the achiral microswimmers can be classified into two types of motions, primary motion and secondary motion (Fig. 7.2(B)), where the red arrow indicates the direction of the dipole, the blue arrow is the swimming velocity, and the purple circular arrow is the rotation direction.
The achiral microswimmers are actuated via a rotating magnetic field; once actuated the microswimmers convert the rotational motion created by the externally applied torque into translational motion (Fig. 7.1(C)). The translation motion is perpendicular to the magnetic field and is a consequence of the rotation of the microswimmer, not of an externally applied force. The use of magnetic fields includes the benefit of having the ability to permeate over long ranges with minimal health effects and to wirelessly transmit large amounts of power for propulsion, motion control, and localization [10,45].
The exact motion of the achiral microswimmer can be captured by tracking the movement of the individual beads using image processing. The individual beads of an achiral microswimmer have their own distinct helical path. The helical motion of each bead can be expressed as
where t is time, and a, b, c, and d are the parameters for the helical trajectory for each bead. These parameters can be determined through curve fitting of experiment data obtained from tracking the individual beads. The equation for is the same as the forward swimming velocity of the microswimmer and should have the same values for all three beads.
The achiral microswimmer, as the name suggests, is a rigid achiral structure that can be actuated via rotation. When rotated, the microswimmer can swim forward in bulk fluid. Previous studies examined the mechanical properties of the microswimmer; most importantly, the rigidity of the microswimmers was investigated by examining the structural deformation of the microswimmers under stress by the hydrodynamic forces during swimming [40]. Such an examination was done by observing relative positions of the three beads to one another. The examination was done quantitatively using a 3D tracking algorithm that can track the position of each individual bead of any given swimmer. It was confirmed that the deformation of the microswimmers was less than 5% over multiple rotation cycles [40].
The microswimmers' ability to swim can be determined using a symmetry analysis. According to Purcell, a shape that is asymmetrical can swim forward when rotated. This is a simplified rule. A deeper investigation requires an examination of the number of planes of symmetry. According to the paraphrased statement by Purcell, it is intuitive to consider an achiral object to be incapability of swimming due to the existence of planes of symmetry. However, according to Happel and Brenner, an achiral object with two planes of symmetry has been shown to be capable of propulsion [44]. Considering the three-bead achiral microswimmer which has 2 planes of symmetry, the translational velocity v and angular velocity Ω of the swimmer are related to the applied force F and torque N by
where the submatrices K, M, and C are the translational, rotational, and coupling resistance tensors, respectively. The resistance tensor depends on the geometry of the swimmer. To actuate the microswimmers, we applied a uniform magnetic field, which means that a torque was applied, not a force. To clarify, only a magnetic field gradient generates a magnetic force that can result in pushing and pulling; a uniform magnetic field will only result in a reorientation of a magnet. This simplifies the relationship to
The torque N is equal to the cross-product of the magnetic dipole m and the magnetic field H. The translational and angular velocity become
and
respectively. Essentially, if the geometry of the swimmer can produce nonzero rotational and coupling resistance tensors (M and C) then it is possible to yield a nonzero value for the translational velocity v. For the achiral microswimmer with 2 planes of symmetry, the M and C tensors are nonzero and can produce a nonzero translational velocity v if a nonzero torque N is applied [40]. The direction of N will be reflected by rotation axis observed during swimming. The numerical simulations were performed based on the analysis performed by Cheang et al. [40] and Meshkati et al. [46]. The details served as a validation of the swimming capability of the achiral microswimmers and showed how the microswimmers swim.
The swimming speed and direction of the achiral microswimmers are controlled using rotating magnetic fields generated via electromagnetic coils. The motion control system consists of three pairs of electromagnetic coils arranged in an approximate Helmholtz configuration, three power supplies (Kepco), a National Instrument (NI) data acquisition (DAQ) controller, a computer, an inverted microscope (Leica DM IRB), and a camera (Point Grey), as shown in Fig. 7.3. Through the use of the DAQ controller, the power supplies can be programmed to generate sinusoidal outputs to the electromagnetic coils in order to create rotating magnetic fields. The camera provides visual feedback and records raw videos. The computer is used for a LabVIEW control interface.
The approximate Helmholtz coil system is based on the Helmholtz configuration and, likewise, can generate a near-uniform magnetic field. The purpose of the near-uniform field is to exert a constant torque on the microswimmer without introducing translational force. A pair of true Helmholtz coil requires a distance between two coils of the same size to be the radius of the coils. Given the space constraint of the microscope, the distances between the coils were modified to be equal to the combined dimension of the outer diameter and the thickness for the coil, creating a cube-like configuration (Fig. 7.3). While the approximate Helmholtz coils can generate near-uniform magnetic fields, the field strength is significantly weaker. The strength of the magnetic field generated from a pair of approximate Helmholtz coils is calculated using a modified version of the Biot–Savart law [47]
where is the permeability, n is the number of turns of wires per coil, I is the electrical current passing through the wires, R is the effective radius of the coil, d is the distance between the center of a coil pair, and x is the coil distance to a point. With 1 A applied to the coils, experimental measurements yield a field strength of 5 mT while the calculated value using Eq. (7.6) yields 5.06 mT. The field profile of the coils validates the claim for a near-uniform magnetic field with a 2 mm region at the center. Given the size of the microswimmers, the 2 mm region provides sufficient space for experiments.
For the implementation of a vision-based feedback controller, a real time imaging processing program was created using LabVIEW. Microswimmers from the camera live feedback images were tracked. Through tracking, the x and y positions of microswimmers were obtained. The x and y positions were then used to calculate velocity and heading angle which were then used for feedback control.
Tracking involved four main steps: (i) image binarization using grayscale thresholding, (ii) size thresholding, (iii) structure definition, and (iv) geometrical centroid calculation. First, in image binarization, each individual frame was converted to a binary black/white image, and the area of objects was defined by setting a grayscale threshold. Then, by removing unwanted objects such as debris using size thresholding, we deleted objects with too few or too many pixels. Next, the structure was defined by filling the interior of the microswimmers and labeling the microswimmers. Finally, the geometrical centroid () was calculated. With the centroid as the position of the microswimmer, the displacement, instantaneous velocity, and heading angle could be calculated.
At low Reynolds number, microswimmers are subjected to environmental disturbances such as Brownian motion and background flows. In cases where environments are highly stochastic, the microswimmer may not be able to reach the destination without compensating for all the environmentally dependent uncertainties. Thus, the external disturbances must be included in the kinematics of the microswimmers. The displacement due to random disturbances can be modeled with Brownian motion
where is the displacement, P is a normally distributed random number, D is the diffusion coefficient of the microswimmers, and t is time. The theoretical diffusion coefficient was calculated as 0.0546 μm2/s. To increase the stochasticity of the environments for the purpose of evaluating a controller's performance, the coefficient can be artificially increased.
The achiral microswimmer's kinematics is described by the nonlinear model
where v is the velocity of the microswimmer, is the heading angle,
is the rotation frequency, and
is the turning rate of the rotating magnetic field. The velocity v is obtained quantitatively from the linear velocity profile plotted against rotation frequency (i.e., 1,
, and 8 Hz) from 7 different microswimmers. Using measurement errors due to video resolution, 30 random samples for each microswimmer were generated. The estimated linear velocity v by curve fitting statistic data is written as
where the coefficient and the estimated covariance matrix is
Using the estimated coefficients, the following estimated deterministic model can be obtained:
where . A stochastic model is then created by adding the environmental factors and other uncertainties, represented as the estimated offset
,
where
models the uncontrolled flow and models random disturbances at time t
where
The motion uncertainties from the microswimmers' handedness can be eliminated by suppressing the secondary motion; this can be achieved using a combination of a static and rotation magnetic fields [39]. The addition of the static magnetic field serves to control the orientation the microswimmer. The achiral microswimmers can be controlled in a 2D plane using the resultant magnetic field generated from three pairs of approximate Helmholtz coils
where is the maximum amplitude of the rotating magnetic field,
is the magnitude of the static magnetic field, ω is the rotational frequency of the field,
is the direction of rotation, and t is time. The magnitudes of the rotating field
and the static field
can be calculated using Eq. (7.6). A schematic for resultant field from Eq. (7.16) is shown in Fig. 7.1(C). The swimming direction can be expressed as the vector perpendicular to the rotating field
In experiments, the speed of the microswimmers is controlled using ω from Eq. (7.16) and the swimming direction is controlled using from Eq. (7.17). For direction control, we expect the heading angle of the microswimmer θ and the direction of the rotating field
to be approximately equal.
Based on the kinematics (7.11), the nonlinear feedback controller can guide an achiral microswimmer from any initial condition towards a target set of the position and angle by calculating the control variables and
given by
which are the rotation frequency and turning rate, respectively.
The controller (7.18) was evaluated in both simulation and experiment. Using parameters ,
, and
, a simulation was performed to evaluate the performance of the controller. The microswimmer started from the initial position at (
,
) and was guided to reach the final position at (150, 125). The microswimmer's rotation frequency
reached a peak at 14 Hz and eventually arrived at the target position in 301 s (Fig. 7.4). The microswimmer's time to reach the target was too long and impractical for applications. The long duration is attributed to the uncertainty from the parameter
since this parameter assumes a constant relationship between the swimming speed and the rotation frequency and ignores environmental effects.
To compensate for the uncertainty from the parameter , a nested control loop was implemented to control the swimming speed of the microswimmers. The nonlinear control (7.18) computes the turning rate and relative velocity
while the inner loop compensates the velocity error using an Integral (I) controller. The inner loop controller is expressed as
where is the reference velocity,
is the measured velocity, and
is the integral gain. The I controller (7.19) has a single pole at zero which is the simplest pole for maintaining the reference velocity without introducing additional zeros and poles that can interfere with the unmodeled dynamics of the swimmer at high rotation frequencies. The value of
determines a balance between maintaining the reference velocity and the stability of the velocity feedback loop. For instance, a small
, close to 0, results in a slow increase of the frequency
until the actual velocity reaches the reference velocity. A higher
results in a faster increase of the frequency
, but may lead to oscillation.
A simulation was performed using using controller (7.19) and is shown in Fig. 7.5. The parameters γ, k, and h were set to 0.1, 0.2, and 1, respectively. Based on the simulation, the time for the microswimmer to reach the target is 44 s. This is around 7 times faster than the previous case with controller (7.18).
Experiments were performed to validate the simulation and to demonstrate the performance of the controller in realistic experimental conditions. In experiments, the microswimmers were subjected to environmental disturbances such as Brownian motion and adverse flow which causes motion uncertainties. In cases when environmental disturbances were too prominent, the use of controller (7.18) resulted in a critically low velocity where the microswimmers could not overcome environmental flows leading to a failure in reaching a destination. By adding in the nested I controller (7.19), the error in velocity due to external disturbances could be compensated over time; in other words, the microswimmer rotated faster in order to swim faster to overcome environmental conditions. A representative experiment using controller (7.19) is showed in Fig. 7.6 which illustrates an achiral microswimmer reaching a target position (red triangle) from an initial position (green triangle). Using real-time image processing, a microswimmer's velocity can be maintained at any reference velocity.
The parameters k, h, γ, and were chosen to be 0.5, 0.65, 0.2, and 0.9, respectively. The parameters k and h serve to give a curvature to the trajectory, while γ controls the swimming velocity with respect to the distance from the target. From the representative experiment in Fig. 7.6, an achiral microswimmer was able to move from the initial position of
,
, and
to the target position of
,
, and
in 24.23 s.
In summary, the experiments successfully demonstrated feedback control of achiral microswimmers using a nonlinear controller and an integral controller. Qualitative data were obtained from repeated experiments to establish a relationship between the swimming speed of microswimmers and the applied rotation frequencies. Based on this relationship, a quantitative kinematic model was estimated. A nonlinear feedback controller was designed and implemented to control the rotation frequency and turning rate
of the microswimmers. Through simulations, it was shown that the feedback controller was able to guide the microswimmer from an initial position to a target position. To increase the performance of the controller, as well as to address the uncertainties from environmental factors, an integral (I) controller was added as a nested loop to compensate for the velocity error. Simulations showed a significant improvement in performance. The simulation results were validated through experiments under a real environment.
This work was funded by National Science Foundation (DMR 1306794), Korea Institute of Science Technology (K-GRL program), Army Research Office (W911NF-11-1-0490), and Ministry of Trade, Industry, and Energy (MOTIE) (No. 10052980) awards to MinJun Kim.
3.142.197.95