Vous êtes sur la page 1sur 7

2012 5th International Conference on Computers and Devices for Communication (CODEC) Invited

Intelligent Proximity Sensors


Somnath Mukherjee (somnath3@sbcglobal.net)
RB Technology, Milpitas, CA 95035, USA

Abstract — Radio-frequency based Intelligent Proximity in proximity and the technique has not gained widespread
Sensors for detection as well as identification of objects is acceptance.
presented. A resonant structure is constructed that creates a Proximity sensors have been around for decades and are
near-field with three-dimensional extent in the same order of used to detect the presence or absence of objects. They are
magnitude as the physical size of the objects to be detected and
broadly classified into electric and magnetic varieties, with
identified. As an object is brought within the extent of the near-
field, a redistribution of electric and magnetic fields take place, each type specialized for non-metallic or metallic objects.
modifying the reflection coefficient (i.e. impedance) of the However, traditional proximity sensors are unable to
resonant structure monitored through a port. The object under identify objects along with detection. Moreover, proximity
test, with its own natural frequencies, perturbs the resonant sensors may not scale favorably in cost.
frequency of the monitoring structure to create a unique set of The present work describes an Intelligent Proximity Sensor
natural frequencies (poles and zeros). These poles and zeros, (IPS) that performs identification of objects in addition to
depending on the size, shape, material composition and
orientation, constitute the RF signature of the object and can
detection. The identification is done by comparing the
be determined from the measurement of reflection coefficient. Radio-frequency (RF) signature of the detected object with
This technique can be used to create smart shelves for a set of signatures stored in a data base. Being RF based, the
automated inventory without the need for tagging, as well for a implementation is simple and cost competitive.
variety of security applications. The technique can be used for In addition to automated inventory, the IPS is a potential
metallic and non-metallic objects, as well as for a combination candidate for various security applications, industrial control,
thereof. The basic principle is illustrated by way of robotics etc.
electromagnetic simulation, and implementation of a smart
tray using the principle is presented.
Index Terms — Intelligent Proximity Sensor, Radio Frequency II. BASIC PRINCIPLE
Identification, Near field Communication, Impedance
Spectroscopy, Discrimination Sensitivity. We will first illustrate the basic technique with the help of a
small dipole, i.e. a linear wire of length l small compared to
the operating wavelength . For such a scenario, it is valid
I. INTRODUCTION
to assume linear current distribution along the length [1].
Various industries like retail, manufacturing, healthcare etc. The distribution stays unaltered even when lumped circuit
are motivated to automate the inventory of stocks, where elements are inserted in series. As the length of the dipole is
objects are organized on shelves in retail stores, warehouses, small compared to wavelength, the radiation resistance will
manufacturing operations, healthcare facilities etc. There is be small. However, a resonance in the near-field can be
a strong need to identify the type of object, introduced by the addition of suitable structures and/or
presence/absence and quantity (count) at an instant of time. lumped elements.
Automated inventory information is very important for just-
in-time replenishment. Another application for automated Hat
inventory is for reduction of shrinkage that can be achieved
by comparing inventory on shelves with objects passing
through point-of-sales terminals. Helix
The most commonly used method for automating inventory
is to use Radio Frequency Identification (RFID) where each Ground
item needs to be tagged. Unless done during manufacturing Plane
of the item itself, individual tagging might be cost- in
prohibitive to the end user. Therefore, a method of
automated inventory that does not require tagging of objects
is highly desirable. Fig.1 Typical Sensing Structure
Image based systems for automated inventory does away the
need for tagging of items, utilizing an array of cameras to
Fig.1 shows a typical sensing structure, where a monopole
capture images of shelves. Video analytic software may then
on a ground plane, instead of a dipole, is used to conserve
be used to identify individual items. However, the cameras
space. A capacitive ‘hat’ is added to the top and an
often fail to view objects that are obstructed by other objects
inductive helix inserted around the middle of the monopole.
The region above the hat is the sensing region where the

978-1-4673-2620-9/12/$31.00 ©2012 IEEE


2012 5th International Conference on Computers and Devices for Communication (CODEC) Invited

objects are introduced for detection and identification. fields need to sufficient such that introduction of objects
Complex Reflection Coefficient (in) is monitored between create a sufficiently large redistribution of fields.
the monopole and the ground plane as shown, the said - As an example, let us consider a sensor to detect and
frequency profiles being used to create RF signatures for the identify objects of dimensions within 3 to 20 cm (typical
objects. supermarket item size). With the monopole being 1.5 cms
long, hat dimensions of few cms and the inductance of the
helix in few tens of nH, the resultant series resonant
frequency of the sensor is around few hundred MHz. The
extent of the field strong enough to affect the reflection
coefficient is few tens of cms. -frequency profiles for
common objects over an octave of frequencies or so can be
created and used for detection and identification.
It is to be noted that the maximum discrimination sensitivity,
i.e. the ability to differentiate between objects, is obtained
near the resonant frequency, which can be surmised from
the following intuitive approach.
The reflection coefficient of an impedance R+jX in an
impedance environment Z0 is given by
R  jX  Z0
Γ (1)
R  jX  Z0
The phase angle of the reflection coefficient is given by
 X 
  2 arctan   (2)
 Z0  R 
If the objects are assumed lossless, the sensitivity can be
defined as
d 1

2
(3)
dX  X 
1  
 Z0  R 
which becomes maximum for X=0, i.e. under resonance
Fig. 2(a) and (b): Electric and Magnetic Field condition.
around Resonance in Sensing Region

In absence of the inductive component, the input impedance III. TYPICAL RADIO-FREQUENCY PROFILES
is capacitive and very little current flows through the We now illustrate with examples how reflection coefficient
monopole. The field above the hat (the sensing region) is changes with introduction of various objects in the sensing
predominantly electric. However, introduction of the region. Electromagnetic simulator HFSS from Ansys Corp.
inductor creates a series resonance with substantial current was used for this purpose.
flowing through the monopole. Thus, in the vicinity of
series resonance, there will be appreciable electric as well as
magnetic field in the sensing region as shown in Fig.(2) (a)
and (b). The balance between electric and magnetic fields
can be perturbed by the introduction of objects in the
sensing region, resulting in shift of resonant frequency.
Depending on the physical characteristics of the object,
there may be more than one resonant frequency. Moreover,
if a broadband measurement is carried out, it is the -
frequency profile that can be used to create the RF signature,
the said profile depending on size, shape, material Fig. 3 Electric and Magnetic Field
composition and orientation. We keep in mind that the - IPS alone 600 MHz
frequency profile is another way of formulating the pole-
The sensing structure as in Fig. 1 was created with the
zero behavior of impedance.
following parameters.
Introduction of the object under test creates a redistribution
Ground plane: 60mm X 60mm, made of copper
of the electric and magnetic fields resulting in a change of
Monopole: 15 mm long, 1.5 mm diameter, made of copper
reflection coefficient. Therefore, the physical extent of the
Hat: 25 mm diameter made of copper

978-1-4673-2620-9/12/$31.00 ©2012 IEEE


2012 5th International Conference on Computers and Devices for Communication (CODEC) Invited

A lumped inductor 100 nH was inserted in the middle Fig. 5 shows the polar plot of reflection coefficient (Smith
region of the monopole. A resistor 15K ohms was connected Chart) over 300 to 800 MHz for the above scenarios. It is
to the inductor in parallel to represent losses. observed that for IPS alone as well as for 30 and 75 mm
Fig. 3 shows the electric and magnetic field for no objects in columns of water, the reflection coefficient is close to unity
the sensing region, i.e. for the IPS alone. The sensing due to the minimal losses from the devices introduced. Most
structure is indicated with the arrows. The fields, at 600 of the losses is coming from the 100 nH inductor which has
MHz, are displayed on a plane (150 mm X 256 mm) passing a Q-factor of around 20. The above scenarios mainly store
through the axis of symmetry containing the monopole. We rather than dissipate energy, but the distribution of electric
notice significant electric and magnetic field at the display and magnetic fields are different in each case. As a result,
frequency of 600 MHz, which is somewhat higher than the the distinction between the profiles arises mainly in phase.
series resonance of the IPS viz. ~ 550 MHz. Though not clear from Fig. 5, a phase-frequency plot (not
shown) reveals a significant change of phase (~18°) near
resonance between 30 and 75 mm columns of water.

Water

0.5

Fig. 4(a) Electric and Magnetic Field IPS


30
50 X 30 mm water cylinder 600 MHz 0
200
-1 -0.5 0 0.5 1
75

-0.5

-1

Fig.5 Polar plot (Smith Chart) of from


simulation (1)

For the 200 mm column of water, radiation is responsible


Fig. 4(b) Electric and Magnetic Field for a reduction in magnitude of reflection coefficient.
50 X 75 mm water cylinder 600 MHz Operation in this region, though possible in principle, needs
to be handled with caution due to the restrictions imposed
by regulatory agencies.

Fig. 4(c) Electric and Magnetic Field Fig. 6 Electric and Magnetic Field
50 X 200 mm water cylinder 600 MHz 50 X 30 mm Al cylinder 600 MHz
Fig. 6 and 7 demonstrate the effect of introduction of
Next, Fig.4(a), 4(b) and 4(c) display the fields in the same metallic objects, the former corresponding to an Aluminum
plane when a cylindrical column of water of diameter 50 cylinder of 50 mm diameter and 30 mm height, whereas the
mm and lengths 30, 75 and 200 mm respectively are later relates to an Aluminum box of dimensions 50mm X
introduced. The reduced electric field in the water column is 50mm X 30mm. As expected, negligible field exists inside
clearly visible; the magnetic field undergoes no significant the metallic enclosures, and the profile will be solely
changes except for 200 mm.

978-1-4673-2620-9/12/$31.00 ©2012 IEEE


2012 5th International Conference on Computers and Devices for Communication (CODEC) Invited

determined by surface geometry and material properties. In was measured at the SMA connector using a Vector
other words, it is not possible to decipher the contents inside Network Analyzer (VNA), with the microstrip trace being
a metallic container – a task possible for dielectric properly de-embedded through calibration.
containers. Fig. 9(a) shows polar plot of reflection coefficients of
objects determined experimentally over 300 to 800 MHz.
The green trace refers to no object, i.e. the ISP alone. The
red trace corresponds to an aluminum soft drink can of
diameter 64mm and height 122mm, while the blue trace
refers to a plastic bottle 60mm diameter X 130mm height
containing water based beverage.

1.0

Fig. 7 Electric and Magnetic Field


50 X 50 X 30 mm Al box 600 MHz 0.5
ISP
Can
Fig. 8 shows the polar plot of reflection coefficient of Bottle
scenarios of Fig. 6 and 7. For comparison purpose, they are
overlaid with a water cylinder of 50 mm diameter and 30 0.0
-1.0 -0.5 0.0 0.5 1.0
mm height. A phase-frequency plot (not shown) reveals a
change of phase of ~5° near resonance between water and
Aluminum cylinders of identical dimensions (50 mm X 30 -0.5

mm). The corresponding number is ~20° near resonance for


Aluminum box and cylinder.
-1.0
1

Fig. 9(a) Polar plot (Smith Chart) of


0.5
from experiment (1)

water cyl It is observed that the general behavior of the simulated and
0 Al cyl experimental data is similar, but an additional resonance is
-1 -0.5 0 0.5 1 Al box
present in the experiment. This resonance is possibly due to
parasitics in the experimental structure, especially in the
-0.5 helix, and its representation by a lossy lumped inductor is
inadequate. Moreover, since the parameters of the simulated
structure and the experimental structure are not identical, we
-1
refrain from quantitative comparison here.
Fig. 8 Polar plot (Smith Chart) of
from simulation (2)

IV. EXPERIMENTAL RESULTS


A sensing structure was constructed with the following
parameters.
Ground plane: 40mm X 40mm, made of copper
Monopole containing helix: 18 mm long, made of copper
Helix diameter: ~3 mm
Helix length: ~ 12 mm
Number of turns in helix: 6
Hat: 25 mm diameter, made of copper
The monopole/helix was constructed using 28AWG bus
wire.
A microstrip trace was present from the monopole end to an
SMA connector for measurement. Reflection coefficient

978-1-4673-2620-9/12/$31.00 ©2012 IEEE


2012 5th International Conference on Computers and Devices for Communication (CODEC) Invited

If the application requires close range operation (e.g.


1.0 inventory on a shelf), near field approximations
(distance<<1) may be used. For excitation current
staying same, lowering the frequency will increase the
0.5 electric field (Er and E) and therefore sensitivity.
Can1
Can2
For security applications, there is a requirement for
somewhat extended range, and intermediate field
0.0 approximation (distance>1) may be used. Here,
-1.0 -0.5 0.0 0.5 1.0
increasing the frequency increases the magnetic field (H)
and electric field (E).
-0.5
Hat I
Hat – top view
-1.0

Side view
Fig. 9(b) Polar plot (Smith Chart) of
from experiment (2)
Distance

From Fig. 9(a) it is clear that the signatures of the bottle Fig. 10 Current Distribution in Monopole
(dielectric) and the can (metal) are significantly different
Introduction of large enough objects (metal or dielectric)
and therefore can be distinguished easily. Fig. 9(b) shows
encourages radiation, as these objects in conjunction with
that it is also possible to distinguish between objects with
the ISP, constitute a radiating structure. Therefore, while
somewhat similar physical characteristics. Here, the red
estimating unintentional radiation, representative objects of
trace (can1) refers to an aluminum soft drink can of
largest possible size need to be considered. Since coherent
diameter 64mm and height 122mm, whereas the blue trace
detection requiring very narrow bandwidth is used to
(can2) refers to steel can of 70 mm diameter and 100 mm
measure reflection coefficient, fortunately RF power can be
height. The losses (i.e. composition) in the two different
restricted to low values.
objects appear to have a negligible impact on the magnitude
Measurement of phase is very important as a large category
of the reflection coefficient. However, the small change in
of objects present minimal losses and therefore has
dimensions creates a significant shift in balance of electric
insufficient impact on the magnitude of reflection
and magnetic energies stored, manifesting as substantial
coefficient. This category of objects however identified by
change (~50° near resonance) in phase-frequency
the appreciable phase change produced due to re-
characteristics.
distribution of stored electric and magnetic energies.
Fortunately, accurate phase measurement to a degree or so
V. DESIGN CONSIDERATIONS can be performed economically.
It is obvious that the contents inside a dielectric container
If the hat is circular and symmetric about the axis of the contribute to the signature, and therefore could be used for
monopole, then the current at the tip of the monopole can be non-invasive measurement of contents. However, for metal
assumed zero because of radial currents flowing in all containers the inside contents do not play a role in
directions (Fig. 10). Further, if the length of the monopole, generating a profile.
is small compared to the wavelength, the current distribution
can be assumed linear [1] as shown in Fig. 10. Introduction
of a lumped component in series is not expected to change VI. PROTOTYPE SMART TRAY
the current distribution appreciably as by definition the
A photograph of the prototype ‘Smart Tray’ is shown in Fig.
lumped component is small compared to wavelength.
11.The prototype tray consisted of ten IPS’s arranged in a
Consequently, the fields generated by the sensing structure
grid fashion - with each IPS selectable through a RF
can be determined from the corresponding expressions for a
switching matrix. The reflection coefficient can be
monopole [1].
determined by a scheme as in Fig. 12, where C1 and C2 are
Increasing the length of the monopole increases the fields,
directional couplers that separate the transmitted and
but space constraints often limits the length. Also, length
reflected waves and are compared in a Magnitude-Phase
cannot increase beyond a point such that the upper limit of
Comparator (e.g. AD8302 from Analog Devices Inc.) to
unintentional radiation set by regulatory agencies is violated.
obtain complex reflection coefficient. The device could

978-1-4673-2620-9/12/$31.00 ©2012 IEEE


2012 5th International Conference on Computers and Devices for Communication (CODEC) Invited

provide an effective phase resolution of less than 2° at the radiating fields since the effect of permissible radiation on
operating frequencies. discrimination sensitivity need to be determined.
A more accurate circuit model for the sensing structure
needs to be established for explaining the parasitic
resonances discussed in section IV.

Object to
be
Identified
Radio
Frequency IPS
Synthesizer
RF Switch Matrix
Fig. 11 Prototype Smart Tray
Each IPS can be trained to identify N number of objects
based on same number of -frequency profiles stored in a C1 C2
computer. Each profile contains M number of frequency
points, and therefore the i-th profile can be defined as

i = {i,j}j=1,M i=1,N (4) Magnitude and Phase Comparison

The profile of the test object is similarly described as

Magnitude Phase
test = {testj}j=1,M (5)
Fig. 12 Measurement of Reflection Coefficient
A figure of merit Ei can be defined as
The sensing structure discussed herewith, though of
2 moderate height (~18mm), is three-dimensional in nature.
M
Γtest j Certain applications, e.g. volumetric sensing for security
Ei   2
(6) and automated inventory, require a two-dimensional sensing
j1
Γtest j  Γ i, j structure for space conservation and convenience. A two-
dimensional structure can be fabricated using very low cost
If Ei > T, T being a pre-defined positive number (threshold), printing or etching techniques. A logical start for the two-
it can be inferred that a match has occurred between the dimensional structure could be the dual of the resonant
object under test with the i-th profile. It is to be noted that monopole, e.g. a resonant loop.
the parameter Ei is analogous to Signal to Noise Ratio (SNR) To correlate the signature of a measured profile with the
and can often be treated as such. ones in the database, a straightforward mean-square-error
As an object is moved away from its reference location (i.e. approach has been used. More sophisticated algorithms to
the location at which the Smart Tray has been trained), Ei distinguish between objects that have similar
decreases, but it is still possible to identify objects up to a electromagnetic behavior (and therefore signatures), could
certain wander distance if T is not set too high. This makes be investigated. Such algorithms, such as matched filter
the system robust in performance under practical situations. detection, might provide increased discrimination sensitivity
and robustness against ‘noise’ introduced due to change in
position and orientation. Artificial intelligence may be used
VII. FUTURE WORK to perpetually train the sensor for improved performance.
As we demonstrated in the section III, a container made of
A rigorous theoretical framework needs to be established for
dielectric material will generate unique signatures for
validating the hypothesis that a particular object with unique
geometry, material composition and dimensions create a various levels of liquid (or granular solid) it contains.
Therefore, it is possible to use this technique for non-
unique -frequency profile at a measurement port. For
invasive measurement of contents such as liquids or
simplicity, the initial treatment can be based on quasi-static
granular solids [3]. A further advantage of this method is
(non-radiating) assumptions, but eventually should include
absence of any guiding structure as in [2].

978-1-4673-2620-9/12/$31.00 ©2012 IEEE


2012 5th International Conference on Computers and Devices for Communication (CODEC) Invited

The smart tray (section VI) has a grid structure (Fig. 13) ACKNOWLEDGEMENT
with each ‘square’ earmarked for items of comparable
footprint. Items too large compared to the grid size cannot The author thanks Sanschip Inc. of San Jose, CA, USA for
be accommodated as shown in top right of ‘Fixed Grid’ in supporting the work.
Fig. 13. On the other hand, if the item is too small compared
to grid dimensions (bottom right in ‘Fixed Grid’ in Fig. 13), REFERENCES
the figure of merit Ei (equation (6)) introduced due to
[1] Balanis C.A, Antenna Theory – Analysis and Design, Harper and
position wander could be too low. To overcome this, the Row, 1982.
smart tray can include grids of various sizes, but that [2] Mukherjee S., Non-invasive Measurement of Liquid Content inside a
restricts its operational flexibility and versatility. To Small Vial, Proc. IEEE Radio and Wireless Symposium (RWS),
construct a truly versatile smart tray capable of identifying New Orleans, January 2010
[3] Mukherjee, S.,Non-invasive Level Measurement for Liquid or
objects with random footprints, the grid can be made be
Granular Solids, US Provisional Patent Application May 2010
small compared to the item footprints. By processing [4] Mukherjee S., Method to Detect and Identify Objects using
adjacent non-empty grids, an item with random footprint Intelligent Proximity Sensors, US Provisional Patent Application
can be identified. Furthermore, the position of the item is January 2011
determined automatically. However, the complexity of the [5]
RF switch matrix is increased several times.

Fixed Grid Versatile Grid

Fig. 13 Grid Structure – Fixed and Universal

CONCLUSION

An intelligent proximity sensor for detection and


identification of objects is presented. Being electromagnetic
in nature (i.e. utilizing both electric and magnetic fields), it
has the ability to detect and identify objects that are metallic,
non-metallic or a combination thereof. This is in contrast to
capacitive (electric field based) proximity sensors suitable
for dielectric materials, and magnetic (eddy current based)
proximity sensors for conducting materials. The intelligent
sensor derives its identification property from operation
around the resonance region, in contrast to traditional
proximity sensors operating near DC. The intelligent sensor
operates on the fact that a certain object uniquely perturbs
the balance of stored electric and magnetic energies,
creating unique frequency-reflection coefficient profiles at a
measurement port. The frequency-reflection coefficient
profiles constitute signatures that can be used for
identification. As the sensor does not necessarily depend on
the power absorption property of the objects, it can detect
and identify lossless objects based on change of phase of
reflection coefficient occurring due to redistribution in
stored electric and magnetic energies. The technique is
amenable to low cost implementation and has the potential
for multiple applications.

978-1-4673-2620-9/12/$31.00 ©2012 IEEE

Vous aimerez peut-être aussi