Nanorobot Communication Techniques:
This work presents chemical communication
techniques for nanorobots foraging in fluid environments
relevant for medical applications. Unlike larger robots, viscous
forces and rapid diffusion dominate their behaviors.
Examples
range from modified microorganisms to nanorobots using
ongoing developments in molecular computation, sensors and
motors.
I. INTRODUCTION
Nanorobots with sizes comparable to bacteria could provide
many novel capabilities through their ability to sense and
act in microscopic environments. Particularly interesting are
biomedical engineering applications [1], [2], where nanorobots
and nanoscale-structured materials inside the body provide
significant improvements in diagnosis and treatment of disease
[3], [4], [5]. The rapid progress in building nanoscale devices
should enable a wide range of capabilities [6]. For example,
ongoing development [7], [8], [9], of molecular-scale
electronics, sensors and motors provides components to enable
nanorobots [10]. Demonstrations of programmable bacteria
[11] can produce computation capability for nanodevices. The
ways to enable Nano-Build Hardware Integrated Systems has
been demonstrated and manufacturing techniques are
advancing gradually [9]. The development of nanosystems for
control of nanorobots to perform specific tasks in medicine
may also enable improvements for nanotechnology automation
[12], [13]. The next level of challenge for nanotechnology
development may address intelligent control system with
device manufacturing and automation in a broad sense [14],
[15].
We present a comprehensive work on techniques for
nanorobots communication based on extensive numerical
results and real time 3D simulation. The approach in this paper
is applied to the following biomedical problem: the nanorobots
with embedded chemical sensors perform the detection of a
single tumor cell in a small venule [16], [17], [18]. The venule
is one among many types of vessels from the human body.
Integrated nanosensors can be utilized for such a task in order
to find intensity of E-cadherin signals [19], [20]. Brownian
motion has a direct influence in a microfluid workspace,
which for active communication makes stigmergy the natural
way for near distance interaction [21]. Thus, communication
and nanorobot control reacting to changes in the environment
is quite appropriate for our study as described in this paper.
II. BEHAVIOR IN FLUID MICROENVIRONMENTS
We consider nanorobots operating in small blood vessels. The
fluid in the vessels contains numerous cells, several microns in
diameter. Viscosity dominates the nanorobots motion through
the fluid in the environment, with physical behaviors quite
different from our experience with larger organisms and robots
[22], [23]. The ratio of inertial to viscous forces for an object
of size R moving with velocity v through a fluid with viscosity
η and density ρ is given by the Reynolds number as follows: Vein internal view without the red cells. The tumor cell is the target
represented by the pink sphere located left at the wall. All the nanorobots
swim near the wall to detect cancer signals.
Typical values for density and viscosity in blood plasma are
represented by equations 2 and 3 respectively.
3 ρ = 1g / cm (2)
10 g / cm.s −2 η = (3)
Flow speeds in small blood vessels are about 1mm/s. This is
also a reasonable speed for nanorobot motion with respect to
the fluid [3],providing
3 Re 10 − ≈ for a 1-micron nanorobot,
and thus viscous forces dominate. Consequently, nanorobots
applying a locomotive force quickly reach a desired velocity
in the fluid. Hence, applied force is proportional to velocity
rather than the direct correlation applied in the acceleration of
Newton’s law F = ma . Diffusion arising from thermal
motion of molecules (Brownian motion) is also important.
Depending on the object’s size, the diffusion coefficient D
characterizes the resulting random motion. In a time t, the
root-mean-square displacement due to diffusion may be
defined as:
k = 6Dt . (4)
For a nanorobot ≈1µm operating at body temperature, this
displacement is ≈ t microns with t measured in seconds.
Brownian motion also randomly changes the nanorobot
orientation. Chemicals have much larger diffusion coefficients
than nanorobots. Because displacement grows as Ο( t)
instead of linearly in t, diffusion is fast at short distances and
relevant for coordinating activity among nearby nanorobots,
but slow over long distances. Chemicals can signal medically
relevant events [24], and also be used for nanorobots
communication. Communication ideally involves chemicals
not otherwise found in the body (to produce low signal noise
level) which are biologically inert over the relevant time scale
of the nanorobot task, and can later be cleared from the body
by existing biological processes [3].
III. NANOROBOT DESIGN
Virtual Reality was considered a suitable approach for
nanorobot design and for the use of macro- and micro-robotics
concepts given certain theoretical and practical aspects that
focus on its domain of application. The nanodevice design
must be robust enough to operate in an aqueous environment
with movement having six-degrees of freedom. The nanorobot
design is derived from biological models and is comprised of
components such as molecular sorting rotors and a robot arm
(telescoping manipulator) [25]. The nanorobot exteriors
considered is comprised of diamondoid new material [26],
[27] to which may be attached an artificial glycocalyx surface
that minimizes fibrinogen (and other blood protein) absorption
and bioactivity, thus ensuring sufficient biocompatibility for
the nanorobot to avoid immune system attack [3]. Different
molecule types are distinguished by a series of chemotactic
sensors whose binding sites have a different affinity for each
kind of molecule [3].
The control system must ensure a suitable performance. It
can be demonstrated with a determined number of nanorobots
responding as fast as possible for a specific task based
scenario. In our work, we consider nanorobots flowing in a
blood vessel with a small target area on the wall emitting a
specific chemical. Manufacturing better sensors and actuators
with nanoscale sizes is advancing [28], [9]. The nanorobots,
designed with sensors for this chemical, must find the source
in a vessel wall. In the 3D workspace the target has surface
chemicals allowing the nanorobots to detect and recognize it.
IV. PHYSICAL PARAMETERS
The microenvironments of the circulatory system vary
considerably in size, flow rates, and other physical properties.
Moreover, chemicals in the blood have a range of diffusion
coefficients, and there is a range of plausible designs for the
nanorobots.
We use typical values for these properties, but our control
techniques can be modified for other values such as adjusting
detection thresholds. Small vessels have diameters of up to
several tens of microns, and lengths of about a millimeter. The
workspace used in the simulator comprised an environment
consisting of a segment of the vessel of length L with a small
target region on the wall emitting a chemical into the fluid
(Fig. 1). Cells and nanorobots continually enter one end of the
workspace along with the fluid flow. We treat any nanorobots
not responding while within the workspace as if they did not
detect any signal, so they flow with the fluid as it leaves the
workspace.
Thus, we choose the workspace length sufficient
PARAMETERS
Chemical signal
production rate Q 10 molecule/s 4 =
⋅
diffusion coefficient D 100 m /s 2 = µ
background concentration 3 3 6 10 molecule/(µm) − ×
Parameter Nominal value
average fluid velocity v =1000µm/s
vessel diameter d = 20µm
workspace length L = 50µm
density of cells 3 3 2.5 10 cell /(µm) − ×
density of nanorobots 4 3 2 10 robot/(µm) − ×
to include the region where the chemical from the target is
significantly above the background level. The cells occupy
about 1/5-th of the workspace volume, a typical hematocrit
value for small blood vessels.
Table I lists the details included in the simulator for
rendering in real time the 3D environment, including
nanorobots and chemical signal parameters. We treat the
nanorobots as cylinders, 1µm in length and 0.5µm in
diameter. Most of the cells are red blood cells, with diameter
6µm . The number densities of platelets and white blood cells
are about 1/20-th and 1/1000-th that of the red cells,
respectively. The nanorobot density equals 12 10 nanorobots in
the entire 5-liter blood volume of a typical adult. Thus a
similar number of nanorobots may be used in medical
applications [3]. The total mass of all the nanorobots is about
0.2g. Due to fluid drag and the characteristics of locomotion in
viscous fluids, nanorobots moving through the fluid at
≈1mm/s dissipate a picowatt [29]. Thus, if all the
nanorobots moved simultaneously they would use about one
watt, compared to a typical person’s 100-watt resting power
consumption.
As a specific example, we consider a typical protein
produced in response to injury, with concentration near the
injured tissue of ≈ 30ng / ml and background concentration in
the bloodstream about 300 times smaller. A typical molecular
weight of 4 10 Dalton leads to the parameter values for the
chemical signal in Table I. This choice provides an interesting
nanorobot task, though we could equally well study tasks
involving chemicals with different concentrations relevant for
other biomedical engineering applications [21]. In our study,
the chemical signal was taken to be produced uniformly over
the target region at the rate Q. The background concentration,
listed in Table I, is a significant sensory parameter, because
the signal rapidly dilutes as it diffuses from the source.
V. NANOROBOT BEHAVIORS CONTROL
In our research, with aims of addressing analyses and
validation for feasible nanorobot control design automation,
the Nanorobot Control Design (NCD) simulator was
developed, which is software for nanorobots in environments
with fluids dominated by Brownian motion and viscous rather
than inertial forces. We examine several practical control
techniques for nanorobot motions. First, as a point of
comparison, we use the nanorobots’ small Brownian motions
to find the target by random search. In a second method, the
nanorobots monitor for chemical concentration significantly
above the background level. After detecting the signal, a
nanorobot estimates the concentration gradient and moves
toward higher concentrations until it reaches the target. In the
third approach, nanorobots at the target release another
chemical which others use as an additional guiding signal to
the target. With our signal concentrations, only nanorobots
passing within a few microns of the target are likely to detect
the signal. Thus, we improve the response by having the
nanorobots maintain positions near the vessel wall instead of
floating throughout the volume flow in the vessel (Fig. 1). In
the render modeling was used a vein wall with grid texture to
enable better depth and distance perception in the 3D
workspace. A key choice in chemical signaling is the
measurement time and detection threshold at which the signal
is considered to be received. Due to background concentration,
some detection occurs even without the target signal. As a
guide for the choice of threshold, we use the diffusive capture
rate α for a sphere of radius R in a region with concentration
as:
α = 4πDRC (5)
where the concentration for other shapes such as cylinders are
about the same [29]. With independent random motions for the
molecules, detection over a time interval ∆t is a Poisson
process with mean value α∆t .
Using Table I,
α ≈ 0.5molecule / s at the background concentration and
≈ 150 near the source.
With the target on the vessel wall, fluid
velocity near the target is lower than the average velocity v in
Table I.
When objects occupy only a small fraction of the
volume the velocity at distance r from the center of the vessel
is:
2 (1 ( /( / 2)) )
2 w = v − r d , (6)
and with the cells, the velocity shows somewhat a parabolic
flow [3], but similar enough for this parabolic profile to give a
useful design guideline.
The control design has to avoid nanorobots to miss the target
as well as to spend power in unnecessary active locomotion.
Obviously, after detecting the signal, for the nanorobot to
move far against the bloodstream, and go around numerous
blood cells, to reach the target may waste precious time and
energy. Thus, a reasonable design choice is for nanorobots to
respond within at most 10µm downstream of the target.
View of simulator workspace showing the vessel wall, cells and
nanorobots. The nanorobot is considerably smaller than the 6µm cell
diameter.
Figure 2: Nominal behavior of a nanorobot passing above the target
(small gray circle) with the fluid moving to the right. Thick dashed line
shows initial passive motion, lasting about 10ms, as the nanorobot
determines signal concentration is significantly above background.
Distances are in microns.
nanorobot 2µm from the wall encounters fluid velocity
≈ 400µm/s .
Therefore, it takes about 30ms to move 10µm,
during which time it will detect, on average, ≈ 3 signal
molecules while the background concentration has ≈ 1%
probability to give even a single detection in this time. Thus,
to save power with sensor processing, the activation threshold
to detect signals is setup for intervals of 30ms. In the
measurement to estimate the concentration gradient, the
sensors are positioned on the surface of nanorobot’s
extremities. After detecting the signal, the nanorobot estimates
the direction to the target from the concentration gradient. The
process consists of alternate short movements with random
changes in direction, at a rate depending on how the
concentration changes during the move. If no signal was
detected, the nanorobot just keeps flowing with the
bloodstream saving power consumption.
In analogy with quorum sensing in bacteria, from monitoring
the concentration of a signal from others, a nanorobot can
estimate the number of nanorobots at the target. So, the
nanorobot uses this information to determine when enough
nanorobots are at the target, thereby terminating any additional
“attractant” signal a nanorobot may be releasing. In our
investigation, the nanorobots stop attracting others once
enough nanorobots have responded. The amount is considered
enough when the target region is densely covered by
nanorobots. Thus, for investigation purposes, values of N={10,
20} were set up in the simulator as a reasonable amount of
nanorobots to the plaque target lesion. A feasible continuation
of this procedure would be to have the nanorobots emit a
different signal that others, not already at the target, interpret
as an indication they no longer need to respond, thereby
leaving them free to continue monitoring for other target areas.
Detecting multiple signaling chemicals requires sensors for
more than one chemical.
The following control methods were considered:
• Random: nanorobots moving passively with the fluid
reaching the target only if they bump into it due to
Brownian motion.
• Follow gradient: nanorobots monitor concentration
intensity for E-cadherin signals, when detected, measure
and follow the gradient until reaching the target. If the
gradient estimate subsequent to signal detection finds no
additional signal in 50ms, the nanorobot considers the
signal to be a false positive and continues flowing with
the fluid.
• Follow gradient with attractant: as above, but nanorobots
arriving at the target, they release in addition a different
chemical signal used by others to improve their ability to
find the target.
The third technique involving communication among the
nanorobots is quite suitable to improve the nanorobots’
behavior performance. By comparing these techniques, we can
evaluate the benefit of chemical communication among
nanorobots to work on typical biomedical applications.
VI. SIMULATOR RESULTS
Q , which is the chemical signal as molecules per second.
The diffusion coefficient is represented by D, and
the diffusion
equation is:
D∇ C = v∂C / ∂x 2 , (7)
TABLE II Detailed simulation depicting 90 experiments with the
amount of 1/3 for each control method. Respective colors represent
the cases for nanorobot behavior based on (a) dark for gradient with
attractant, (b) blue for follow gradient, or (c) green for random
motion.
origin and no net flux across the boundary plane at y = 0,
determines the steady-state concentration C,
which is
molecules per 3 µm or chemical concentration at point (x, y, z):
( )/(2 )
2 ( , , ) v r x D e
Dr
Q C x y z − −
⋅
= π
(8)
where
2 2 2 r = x + y + z (9)
is the distance to the chemical signal source [29].
Fig. 2 is an illustration of nanorobot behavior. The fluid flow
pushes the concentration of the diffusing signal downstream.
Consequently a nanorobot passing more than a few microns
from the source won’t detect the signal while it is still
relatively near the source. As an example, considering the
parameters from Table I, when nanorobots passing close
enough, they detect on average the higher signal concentration
within about 10ms.
Thus, keeping their motion near the vessel
wall, the signal detection happens after these have moved at
most 10µm past the source. Therefore, it provides about
5nanorobot/s arriving at the tumor cell in the small venule.
Eq. (8) also illustrates a design trade-off for chemical signals
the nanorobots could release. Instead of the diffusion
coefficient associated with the chemical from the target, such
additional signals would use other molecules which could, by
design, have a different diffusion coefficient.
From Eq. (8), the
effect of the fluid motion becomes significant at distances
beyond O(D/ v). Thus, notwithstanding the fluid flow, larger
diffusion constants allow further spread upstream.
On the
other hand, the O(1/ D) overall factor in Eq. (8) means lower
concentrations. Furthermore, the concentration of the new
signal is time dependent since the source strength increases as
more nanorobots reach the target and the signal from each
nanorobot requires time ( / ) 2 O r D to reach a distance r.
Therefore, faster diffusion results in lower concentrations,
requiring more time for other nanorobots to determine
gradients. Hence, chemical diffusion could be more efficient
for nanorobot communication, if the signals are increasing in
a steady, constant and progressively manner.
Nanorobots passing within ≈ 0.1µm of the target usually
bump into it. Those passing within a few microns often detect
the signal, which spreads a bit further upstream and away from
the single tumor due to the slow fluid motion near the venule’s
wall and the cells motion. Nanorobots close to the wall also
benefit from the slower fluid motion by having more time to
detect the signal, as discussed previously. Thus, the present 3D
simulation provides guidelines for nanorobot communication
and activation control, as well as for sensor manufacturing
design. We use an “attractant” signal with the same value of D
as the original signal. Each nanorobot can release at one-tenth
the rate of the target over the times considered here.
Distinct performances were observed throughout a set of
analyses obtained from the NCD software, where the
nanorobots use also chemical sensors as the communication
technique to interact dynamically with the 3D environment,
and to achieve a more successful collective coordination. Fig.
3 shows the virtual environment in our study, comprised a
small venule vessel which contains nanorobots, the red blood
cells (RBCs) and a single tumor cell, which is the target area
on the vessel wall. Here, the target area is overleaped by the
RBCs.
In the simulation, the nanorobots search for possible
small cancer tumor into the workspace crowded by RBCs.
In Fig. 4 it could be observed in a detailed fashion the
information about the nanorobots behaviors.
Table II provides
a summary and comparison of the control techniques
evaluated using the NCD simulator. It shows the time required
for 10 and 20 robots to identify and reach the target. Each
value is the mean of 30 repetitions of the simulation, with
standard deviation in parentheses. The error estimate for these
mean values is 30 times smaller than the standard deviations
listed here. For comparison, if every nanorobot passing
through the vessel found the target, 20 nanorobots would
arrive at the target in about 0.2s. As one would expect,
enabling nanorobots to detect and follow gradient
concentration increases the probability for nanorobots to find
the target, where in comparison with random motion the
nanorobots show a better performance of 23%. Further, for
gradient with “attractant”, we see that using the signals allows
the nanorobots to find and reach the target in the 3D
workspace 46% faster than that with random motions. This is
a remarkable improvement in performance for response time.
IEEE ICARCV 2006 International Conference on Control, Automation, Robotics and Vision
VII.
CONCLUSION AND REMARKS
The development of nanorobots may provide remarkable
advances for diagnosis and treatment of cancer. Using
chemical sensors they can be programmed to detect different
levels of E-cadherin and beta-catenin in primary and
metastatic phases. Our work has shown a comprehensive
methodology on tracking single tumor cell in a small venule,
where nanorobots using communication techniques to increase
their collective efficiency. The simulation has clearly
demonstrated how better time responses can be achieved for
tumor detection, if chemical signals are incorporated as part of
nanorobot control strategy. As observed in the study, the
follow gradient with attractant signal is a practical method for
orientation and coordination of nanorobots. It has enabled a
better performance for nanorobots to detect and reach
cancerous targets. This approach can be useful in the treatment
of many patients for a detailed examination and intervention.
A single tumor cell can be characterized as a typical
endothelial cell mutation with profound consequences for
patients suffering from cancer. Endothelial cells have a large
number of functions and may play an important role in human
health. They also serve as part of the structure forming the
inside blood vessels, which are spread throughout every single
organ or system comprising our body. An abnormal cell
mutation and reproduction can represent a wide variety of
malignant cases in the oncology field. Thus, a better
understanding and the development of new tools based on
nanotechnology through chemical sensors may represent
important advances to identify, and combat the initial stage of
tumor development. Nanorobots can help with significant
improvement on cell therapy techniques, and unprecedented
positive results to save lives.
No comments:
Post a Comment