IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO.

3, MARCH 2008 869

Design and Evaluation of a Reconfigurable

Antenna Array for MIMO Systems
Daniele Piazza, Student Member, IEEE, Nicholas J. Kirsch, Student Member, IEEE, Antonio Forenza, Member, IEEE,
Robert W. Heath, Jr., Senior Member, IEEE, and Kapil R. Dandekar, Senior Member, IEEE

Abstract—New reconfigurable antenna array is demonstrated novel antenna solutions in order to improve MIMO links [4],
for multiple input multiple output (MIMO) communication sys- [5]. From an antennas perspective, different array configura-
tems that improves link capacity in closely spaced antenna arrays. tions and types of element have been proposed and analyzed
The antenna system consists of an array of two printed dipoles sep-
arated by a distance of a quarter wavelength. Each of the dipoles for MIMO links [4], [5].
can be reconfigured in length using PIN diode switches. The switch It has been shown that signal correlation and mutual coupling
configuration can be modified in a manner adaptive to changes effects between MIMO array elements has a significant effect
in the environment. The configuration of switches effects the mu- on MIMO capacity [6], [7]. Recently, it has also been shown
tual coupling between the array elements, and subsequently, the that pattern diversity, which results from mutual coupling, can
radiation pattern of each antenna, leading to different degrees of
pattern diversity which can be used to improve link capacity. The
potentially lead to low inter-element received signal correlation
PIN diode-based reconfigurable antenna solution is first motivated and thus higher MIMO channel capacity [7], [8].
through a capacity analysis of the antenna in a clustered MIMO In this paper, we present a novel reconfigurable MIMO an-
channel model. A new definition of spatial correlation coefficient tenna array which will demonstrate how the ability to select
is introduced to include the effects of antenna mismatch and ra- between different pattern diversity configurations can improve
diation efficiency when quantifying the benefit of pattern diver-
MIMO system link capacity. The most closely related work [9],
sity. Next, the widespread applicability of the proposed technique
is demonstrated, relative to conventional half wavelength printed [10] has applied this type of reconfigurable antenna to achieve
dipoles, using computational electromagnetic simulation in an out- pattern and frequency diversity in &utput (SISO) links. Recon-
door and indoor environment and field measurements in an indoor figurable antennas have also been suggested for application in
laboratory environment. It is shown for the 2 2 system consid- MIMO systems [11]–[13]. The switch-based parasitic antenna
ered in this paper, that an average improvement of 10% and 8% is in [14] has also been proposed to improve MIMO system link
achieved in link capacity for a signal to noise ratio (SNR) respec-
tively of 10 dB and 20 dB in an indoor environment compared to a capacity through pattern diversity. Our reconfigurable antenna
system employing non reconfigurable antenna arrays. solution differs from [14] in that it provides a compact design,
suitable for handheld devices.
Index Terms—Antenna measurements, multiple input mul-
In this paper, the antenna system under study consists of an
tiple output (MIMO) systems, reconfigurable antennas, spatial
correlation. array of two reconfigurable microstrip dipoles; the two active
elements of the array can be reconfigured in length using PIN
diode switches. The setting of the different switches results in
I. INTRODUCTION different geometries of the antenna and, as a result, different
levels of inter-element mutual coupling and array far-field radi-
ULTIPLE-INPUT multiple-output (MIMO) wireless ation patterns. The goal of such a system is to choose the config-
M systems have demonstrated the potential to increase
communication spectral efficiency in a rich multipath environ-
uration of switches in a environment/channel adaptive fashion
to decrease MIMO spatial channel correlation and subsequently
ment [1]. Recent work in this field has focused on measuring maximize link capacity.
and characterizing the real MIMO propagation channel [2], de- The proposed reconfigurable antenna system was used to
veloping efficient space-time coding strategies [3] and creating implement a 2 2 MIMO system employing spatial mul-
tiplexing transmission. The benefits derived from using our
reconfigurable MIMO antenna system is first motivated through
Manuscript received July 27, 2006; revised April 15, 2007. This work was spatial correlation coefficient analysis in a clustered MIMO
supported by the National Science Foundation under Grants 0435041 and channel model [15]. Through this analysis, we demonstrate the
0322795. advantage of switching between different antenna configura-
D. Piazza, N. J. Kirsch, and K. R. Dandekar are with the Department of Elec-
trical and Computer Engineering, Drexel University, Philadelphia, PA 19104-
tions based on the spatial characteristics of the MIMO channel.
2875 USA (e-mail: [email protected]; [email protected]; dandekar@drexel. In conducting this analysis, we developed a new definition of
edu). spatial correlation to include the effects of antenna mismatch
A. Forenza is with Rearden, LLC, San Francisco, CA 94107 USA (e-mail:
[email protected]).
and radiation efficiency when quantifying the impact of pattern
R. W. Heath, Jr., is with the Department of Electrical and Computer En- diversity on MIMO systems. Since some limitations with
gineering, The University of Texas at Austin, TX 78712 (e-mail: heath@ece. the adopted clustered channel model have been recognized
utexas.edu).
Color versions of one or more of the figures in this paper are available online
(i.e. underestimation of actual MIMO channel capacity and
at http://ieeexplore.ieee.org. consideration of azimuthal radiation pattern instead of full 3D
Digital Object Identifier 10.1109/TAP.2008.916908 radiation) [16]–[18], in this paper, we verify the performance
0018-926X/$25.00 © 2008 IEEE
870 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 3, MARCH 2008

Fig. 1. Dipole antenna design.

of the proposed reconfigurable antenna using both i) com-

putational electromagnetic simulation (simulation technique
described in [19]) in outdoor and indoor environment, and
ii) experimental field-testing in an indoor environment. In
conducting this analysis the two antennas have been assumed to
transmit/receive simultaneously including therefore the effects
caused by the superposition of the antenna’s radiation patterns Fig. 2. Antenna structure: top and bottom view.
at the receiver and transmitter.
This paper is organized as follows. In Section II, we intro-
duce the MIMO channel model used to evaluate the reconfig- and known at the receiver, the signal vector at the transmitter is
urable antenna array. In Section III, the physical parameters of composed of independent signals with equal power. In this
the reconfigurable antenna system design and radiation pattern case, the capacity is given by [20]
simulations are presented. In Section IV, a new formulation of
spatial correlation coefficient for quantifying MIMO pattern di- (2)
versity is presented. This spatial correlation analysis in a MIMO
clustered channel model motivates the use of the proposed re- where is an identity matrix, denotes a com-
configurable antenna in MIMO systems. In Section V, simu- plex conjugate transpose operation, and SNR is the average
lation results are presented for the proposed antenna obtained signal-to-noise-ratio over all receiver array elements.
using computational electromagnetic simulations in an outdoor In a correlated Rayleigh-fading MIMO channel the channel
environment. In Section VI we analyze capacity results from matrix is defined, according to the Kronecker model [21],
field-measurements taken on a 2 2 MIMO testbed using the [22], as
reconfigurable MIMO array, and we compared them with re-
sults obtained with computational electromagnetic simulations (3)
performed in the same indoor environment. Finally, conclusions
where and denote respectively the transmit and receive
are drawn in Section VII.
spatial correlation matrices and is a matrix of
II. MIMO SYSTEM AND CHANNEL MODEL complex Gaussian fading coefficients. In this paper, the special
case of is considered.
In a narrowband MIMO wireless system with transmit
The “Kronecker” model [21] in (3) is adopted in this paper to
antennas and receive antennas, the signal collected at the
illustrate the potential of the proposed reconfigurable antenna.
receiver is related to the signal outgoing from the transmitter
The same model has been used previously in the literature [8]
through the relation
to demonstrate the benefits of pattern diversity in MIMO com-
(1) munication systems.
The spatial correlation coefficients (entries of the matrix
where is the signal vector at the receiver array, and ) and the channel capacity can be generated through
is the signal vector at the transmit antenna array, a clustered MIMO channel as described in [23]. In clustered
is the complex additive white Gaussian noise MIMO channel models, the scattering objects around the
(AWGN) vector and is the channel transfer ma- transmit/receive arrays are modeled as “clusters”. Each cluster
trix. Assuming that the channel is unknown at the transmitter is characterized by a mean angle of arrival (AOA) , where
PIAZZA et al.: DESIGN AND EVALUATION OF A RECONFIGURABLE ANTENNA ARRAY FOR MIMO SYSTEMS 871

Fig. 3. The measured S for the “long” and “short” antenna configuration in a 2 element array with inter-element separation of =4 when (a) the other antenna
is in the “short” configuration, and (b) the other antenna is in the “long” configuration.

represents the solid angle consisting of azimuth TABLE I

and elevation components. For simplicity, it is as- LIST OF STRUCTURAL PARAMETERS OF THE PRINTED DIPOLE ANTENNA
sumed that there is a single cluster (as in [8]) and that the
cluster location is distributed only over azimuth directions with
mean AOA, (as in [8] and [23]). Depending on the system
bandwidth, the excess delay across different propagation paths
may not be resolvable. In this case, multiple AOAs are defined
with an offset with respect to the mean AOA of the cluster
. This angle of arrival is generated according to a certain
probability density function (PDF) that models the power
angular spectrum (PAS). The variance of the PAS is a measure
of the angle spread (AS), , of the cluster.

III. ANTENNA DESIGN

The proposed antenna solution has been realized using an
array of two microstrip dipoles [24], the geometry of which is
shown in Fig. 1 and Fig. 2. A quarter-wavelength microstrip
balun acts as an unbalanced-to-balanced transformer from the flowing back to the RF input, an inductor of 56 nH is used to
feed coaxial line to the two printed dipole strips [25]. The block the RF from flowing in the dc supply trace, and a resistor
ground plane of the microstrip line and the dipole strips are of 100 is mounted to limit the voltage across the diode as
in the same plane. As indicated in this figure, the presence of shown in Fig. 2. The structural parameters of the printed dipole
a via-hole permits feed point 2 of a printed dipole strip to be antenna are listed in the Table I.
shifted in phase by 180 with respect to the feed point 1 of The antenna structure has been analyzed and simulated using
the other printed dipole strip. This occurs because of the 180 the method of moments (MoM) (FEKO software) [26]. The be-
phase difference between the top strip and the ground plane of havior of the switch has been modeled using a microstrip line
the microstrip line. with the same scattering parameters of the PIN diode in both
The lengths of the dipole-arm strip, and therefore the ge- ON and OFF configurations.
ometry of the antenna, can be changed using two PIN diode The close proximity of the two printed dipoles has been se-
switches (MA4P789), which guarantee an isolation of 36 dB lected such that there is a strong mutual coupling between the
and low insertion loss (0.7 dB). In this way, it is possible to de- two dipoles. This coupling is effectively used to have different
fine two configurations for the antenna, one when both of the radiation patterns for each array’s geometry. In particular be-
switches are turned on (“long” configuration) and another when cause of the vicinity of the two reconfigurable dipoles, varying
they are turned off (“short” configuration). In order to drive the length of one of the dipoles, effects the input impedance and
the switches, a direct bias voltage of 1.7 V is supplied to the the current distribution of the other dipole, changing therefore
diode with two thin traces, the presence of which, as analyzed its radiation properties [25]. Fig. 3 shows the coupling level be-
through electromagnetic simulations, does not corrupt the radia- tween the different possible configurations of antenna geometry
tion characteristics of the antenna. A surface mounted capacitor in an array (both “short,” both “long,” first antenna “short” and
of 47 pF is placed on the balun to block the dc current from second “long,” first antenna “long” and second “short”). It can
872 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 3, MARCH 2008

Fig. 4. The measured antenna input impedance for the “long” and “short” antenna configuration in a 2 element array with inter-element separation of =4 for
frequencies from 1.5 GHz to 3.5 GHz when (a) the other antenna is in the “short” configuration, and (b) the other antenna is in the “long” configuration. A marker
selects the frequency of 2.45 GHz.

Fig. 5. Radiation pattern (in dB) in the E plane (a) and in the H plane (b) of the printed dipole in the “short” configuration for an operation frequency of 2.45 GHz.

be noted that the mutual coupling is different for each configu- figurations of antenna geometry in an array (both “short”, both
ration of the array. In Fig. 4 is depicted the input impedance of “long”, first antenna “short” and second “long”, first antenna
one of the two dipoles for each array configurations. It can be “long” and second “short”) the simulated radiation pattern of
noted that the reconfigurable dipole is characterized by an input the two elements is reported in Fig. 6 and Fig. 7 in the and
impedance that is different for each array’s configuration. These planes for the operation frequency of 2.45 GHz. The radia-
differences in input impedance are due to the different lengths tion pattern of each dipole configuration is measured leaving the
of the dipole and the different level of mutual coupling of each other dipole of the array open circuited according to [27]. The
configuration. A difference in input impedance is in fact ob- radiation pattern, which is linearly polarized, changes from one
served not only when the “active” dipole length is changed from configuration to another. Fig. 6 shows the main differences in
“short” to “long” but also when the “parasitic” dipole length radiation pattern between the different antenna configurations.
is changed from “short” to “long” leaving the “active” dipole In particular the plot shows that the “long” configuration has a
length unchanged. more prominent secondary lobe than the “short” configuration.
The strong mutual coupling effects between the two antennas This difference also makes the “short” configuration more direc-
influence also the radiation propertied of the array’s elements. tive than the “long” configuration. Another difference between
Fig. 5 shows the simulated radiation pattern of a single element configurations is in the direction of the main beam. In Fig. 6(b)
half wavelength printed dipole. The radiation pattern is greatly and Fig. 6(c), the radiation patterns are tilted with re-
modified when the antenna is used in a two element array with spect to the radiation patterns in Fig. 6(a) and Fig. 6(d). Fig. 7
inter element separation of . For the different possible con- does not show as many differences in the patterns between con-
PIAZZA et al.: DESIGN AND EVALUATION OF A RECONFIGURABLE ANTENNA ARRAY FOR MIMO SYSTEMS 873

Fig. 6. Radiation pattern (in dB) in the H plane ( = 90 ) of the two printed dipole separated by =4 in all the configurations for an operation frequency of 2.45
GHz: (a) antenna 1, “short,” antenna 2, “short;” (b) antenna 1, “long,” antenna 2, “short;” (c) antenna 1, “short,” antenna 2,“ long;” (d) antenna 1, “long,” antenna
2, “long”.

Fig. 7. Radiation pattern (in decibels) in the E plane ( = 90 ) of the two dipole separated by =4 in all the configurations for an operation frequency of
2.45 GHz: (a) antenna 1, “short,” antenna 2, “short;” (b) antenna 1, “long,” antenna 2, “short;” (c) antenna 1, “short,” antenna 2,“ long;” (d) antenna 1, “long,”
antenna 2, “long.”

figurations as Fig. 6, however the difference in directivity can be The return loss for the antenna is determined when the antenna
observed. Fig. 6 and Fig. 7 also show the different levels of ra- is placed in a two element array with inter-element spacing of
diated power (i.e., efficiencies) between the four possible array so that the effects of mutual coupling can be incorporated
configurations. into the antenna input impedance. The system has been designed
The simulated as well as the measured values of the return to work in the frequency band between 2.4 GHz and 2.48 GHz
loss of the antenna in the two configurations are shown in Fig. 8. typical of an 802.11-like MIMO-aware WLAN and Fig. 8 shows
874 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 3, MARCH 2008

Fig. 8. The measured reflection coefficient for the “long” and “short” antenna configuration in a 2 element array with inter-element separation of =4 when (a)
the other antenna is in the “short” configuration, and (b) the other antenna is in the “long” configuration. The respective simulated curves are also shown dashed.

Fig. 9. The measured VSWR for the “long” and “short” antenna configuration in a 2 element array with inter-element separation of =4 when (a) the other antenna
is in the “short” configuration, and (b) the other antenna is in the “long” configuration.

that the antenna configurations are matched at the frequency IV. ANTENNA ANALYSIS IN CLUSTERED
band of interest (for a target of 10 dB). In particular the MIMO CHANNEL MODEL
“short” configuration (half wavelength microstrip dipole) per- In this section, we propose a new definition of spatial correla-
formed better than the “long” configuration (three quarter wave- tion to quantify the only pattern diversity while including the ef-
length microstrip dipole) both in radiation efficiency (“long” fects of antenna input impedance and radiation efficiency. Using
configuration , “short” configuration ) and this new definition, the proposed reconfigurable antenna array
in matching condition. In Fig. 9 is then shown the measured for MIMO systems is motivated through spatial correlation and
for all the array’s configuration, from which it is evi- ergodic capacity performance analysis in clustered channels.
dent that each antenna’s configuration is matched over the band-
width of interest. A. Spatial Correlation Coefficients
The proper choice of the receiver and transmitter array switch
The spatial correlation between the th and th elements of
configurations (i.e. the length of each antenna) is the one which
MIMO arrays, previously used to quantify pattern diversity is
decreases MIMO spatial channel correlation and maximizes
given by [23]
channel capacity for a particular multipath environment. For
example, the optimal choice could be determined by switching
(4)
through all the possible length configurations at transmitter and
receiver whenever the capacity experienced by the link falls
below a predefined threshold. The scope of this paper is to where is the solid angle, is the PAS of the
demonstrate proof-of-concept capacity improvements in static scattered fields, is the far-field radiation pattern of the -th
networks in indoor and outdoor environments. antenna of the reconfigurable antenna array and is the
PIAZZA et al.: DESIGN AND EVALUATION OF A RECONFIGURABLE ANTENNA ARRAY FOR MIMO SYSTEMS 875

far-field radiation pattern of ideal isotropic radiators. Moreover,

it is assumed that

(5)

where the first term of the equality is the condition for to

be a PDF.
Since the spatial correlation is normalized with respect to the
antenna gain of ideal isotropic radiators, the envelope of (4) is
not guaranteed to be lower than one, as provided by the conven-
tional definition of correlation [28], [29]. This guarantee cannot
be satisfied because in (4) the power radiated by the isotropic
radiator and the th and th elements of the MIMO arrays are
assumed to be equal. For the case of the proposed reconfigurable Fig. 10. Single-sided correlated MIMO channel ergodic capacity for the four
antenna, however, the above assumption does not hold because possible array configurations as a function of the cluster mean AOA, with
each antenna configuration has different levels of radiated power SNR = 20 dB. The erogdic capacity for a double-sided uncorrelated channel
is also shown as a reference.
given equal input power. This difference in radiated power oc-
curs because of the different input impedance and efficiency of
each configuration, as shown in Section III.
In order to model the effects of antenna radiation efficiency The spatial correlation coefficients were numerically evalu-
and input impedance, the definition in (4) has been revised ated by employing realistic radiation patterns in the azimuthal
plane (see Fig. 6) for all the switch configurations of a two ele-
ment reconfigurable antenna array. The MIMO channel capacity
was computed by using the spatial correlation coefficients in (6).
A tight upper bound on the ergodic capacity for spatial multi-
(6) plexing systems (with equal power allocation across the transmit
antennas) [32], was adopted in this paper. The upper bound for
double sided correlated MIMO channel capacity is expressed as
where and are the voltage reflection coefficients at
the th and th antenna input ports and and describe the
percentage of power lost due to ohmic and dielectric losses in
the respective antennas. Note that in this definition of spatial
correlation coefficient the radiation pattern of an array element
is defined for a unit driving current and all other elements in the
array terminated in an open circuit. (7)
We observe that the definition in (6) accounts for the
cross-polarization components of the radiated/incident fields.
In fact, from equation (4) in [29], we define the PAS as where and ,
, where and are the angular are the minor determinants of and (i.e., a determinant
power densities of the and components of the incident of the matrix lying in the rows and in the
field, respectively. For the sake of simplicity it is assumed columns of and .
, which is equivalent to say that the We first considered the performance of the reconfigurable
channel is characterized by a unitary cross-polarization-dis- array in single-sided correlated MIMO channels. In single-sided
crimination as defined in [29], [30]. Thus correlated channels, either or , with the other
. It is also assumed, matrix determined by the clustered MIMO channel model.
consistent with the measurements results in [31], that most of Thus, the channel capacity was computed for a MIMO system
the scattered power propagates over the azimuth directions. employing the reconfigurable antenna at only one end of
Therefore , the link. Fig. 10 shows the channel capacity achieved with
where denotes the convolution operator, is the mean AOA the four possible array configurations as a function of the
of the cluster and is generated according to the truncated mean AOA of the cluster, for an angular spread of
Laplacian distribution as in [15]. and . Fig. 10 also shows the ca-
pacity of a double-sided uncorrelated 2 2 MIMO channel
B. Performance in Clustered MIMO Channel Model as a reference. It can be observed that for each
antenna configuration, the capacity varies over the mean AOA
In this section, we evaluate the capacity performance of the of the cluster. The highest value of achievable capacity can be
proposed reconfigurable antenna in clustered MIMO channel guaranteed by switching between the four array configurations,
models described in Section II. The capacity is computed from since each configuration outperformed the others in certain
the spatial correlation coefficients defined in (6). mean AOA ranges.
876 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 3, MARCH 2008

Fig. 11. Contour plot of the double-sided correlated MIMO channel ergodic capacity for the (a) reconfigurable antenna array, and of the (b) non-reconfigurable
=2 printed dipole as a function of the cluster mean AOA and per-cluster AS, with SNR = 20 dB.

In order to complete a channel capacity analysis of the re-

configurable antenna in a clustered channel model, the channel
capacity defined in (7) was computed for a double sided cor-
related channel (i.e. both and determined by the clus-
tered MIMO channel model). In this case, the proposed recon-
figurable antenna was used at both ends of the MIMO link. The
capacity of the channel was found according to (7) for each pos-
sible configuration of the transmitting and receiving antenna
array, for a total of 16 different configurations. The capacity
achievable with the reconfigurable antenna array was defined as
the highest one among the 16 antenna configurations. Fig. 11(a)
depicts the double-sided correlated MIMO channel ergodic ca-
pacity of the reconfigurable antenna as a function of the cluster
mean AOA and cluster angle spread (assuming these parame-
ters to be the same at both the transmitter and the receiver side),
with . In Fig. 11(b) the ergodic capacity achiev-
able with the same printed dipole in the “short” configuration
(as specified previously, the most efficient in terms of radiation
and matching) is reported as a reference non-reconfigurable an-
tenna system. The results show the variation of the capacity with
the mean AOA of the cluster and the AS. The reconfigurable
antenna outperforms the non reconfigurable printed dipole for
every mean AOA and every AS. With respect to the single-sided
correlated channel of Fig. 10, the situation in Fig. 11(a) shows
the capacity achieved is lower, since both transmitter and re-
ceiver sides are correlated. However, the double-sided corre-
lated channel better describes the real channel when closely
spaced antenna array elements are used. Moreover, the relative
capacity improvement achievable with the reconfigurable an-
tenna in a double-sided correlated channel is higher than the
capacity improvement achievable in a single-sided correlated Fig. 12. Outdoor and indoor environments models: (a) 3D model of downtown
channel. Austin (TX location is shown with some of the RX locations), (b) orientation of
reconfigurable antenna array in the indoor environment illustrating the multiple
The obtained results motivate then the use of this reconfig- receiver locations and the transmitter location.
urable antenna in a MIMO link and demonstrate how a decision
rule for switching configuration can be generated using the in-
formation of the channel spatial correlation, that is a function of Since some limitations with the adopted clustered channel
mean AS and AOAs. model have been recognized [16]–[18], in the next two sec-
PIAZZA et al.: DESIGN AND EVALUATION OF A RECONFIGURABLE ANTENNA ARRAY FOR MIMO SYSTEMS 877

Fig. 13. CDF of capacity for the reconfigurable and fixed antenna array in a (a) ray tracing simulated outdoor environment, and in a (b) measured indoor
environment.

tions, the reconfigurable antenna performance is analyzed using

computational electromagnetic simulation in an outdoor and in-
door environment, and experimental field-testing in an indoor
environment. In particular through electromagnetic simulations
and experimental field-testing the effect of considering the full
3D antenna radiation pattern is investigated in realistic environ-
ments that consist of more than one cluster and in which the
effect of cross-polarization is considered.

V. RAY TRACING SIMULATION RESULTS

In this section, the reconfigurable antenna array is studied in
terms of channel capacity, via numerical computation using an
electromagnetic ray tracer, FASANT [33]. FASANT is a deter-
ministic ray tracing program based on geometric optics and the
uniform theory of diffraction. A 3D model of Austin downtown Fig. 14. Percentage capacity improvement achieved with the reconfigurable
was simulated as the geometry input of FASANT. There was a antenna versus receiver location, and mean percentage capacity improvement
transmission point located in the center of the model, at a height (dashed line) in outdoor environment.
of 1.5 m, and a receiver, at a height of 1.5 m, moving along the
streets, as shown in Fig. 12(a), occupying 2400 different loca-
to the “short” configuration channel matrix. The normalization
tions. Note that the orientation of the array has been selected
factor is defined as:
such that the maximum degree of pattern diversity between the
different antenna’s configuration is in the azimuthal plane.
The radiation patterns (Fig. 6 and Fig. 7), obtained from
" (8)
FEKO antenna design software, for the different configurations
of the antenna spaced by were used in the ray tracing
simulation both at the receiver as well as at the transmitter in a The capacity of the link was then determined using (2), as-
2 2 MIMO system. Ray tracing considers the effects caused suming a SNR of 20 dB.
by the superposition of the antenna’s radiation patterns at the The capacity of the channel was found in this way for each
receiver and transmitter. possible configuration of the transmitting and receiving antenna
The channel matrix was computed as described in [34]. array, for a total of 16 different configurations per position, and
To define the capacity of the MIMO communication link, a the optimal solution of the reconfigurable antenna was the one
Frobenius normalization of the channel matrix was computed which guaranteed the highest channel capacity. The channel
in order to remove the differences in path loss among a number capacity achieved using the reconfigurable antenna was com-
of channel matrices [35]. Specifically, to preserve the relative pared with the capacity of a MIMO link where the same printed
antenna gain effects of each configuration, all the channel ma- dipole in the “short” configuration (as specified previously to
trices for each receiver location were normalized with respect be the most efficient in terms of radiation and matching) was
878 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 3, MARCH 2008

Fig. 15. MIMO OFDM testbed block diagram.

used at the transmitter and receiver array. Based on the data

collected from simulations, a cumulative distribution function
(CDF) for the capacity was numerically obtained and shown
in Fig. 13(a), where one curve represents the CDF relative to
a system where the reconfigurable antenna solution was used,
and the other was relative to one where the “short” configuration
(fixed antenna) was used. The percentage improvement in ca-
pacity that the system provided on a location-by-location basis,
with respect to the fixed antenna solution, is shown in Fig. 14. As
shown by the dashed line the average percentage improvement
in capacity results is 7% for a SNR of 20 dB, reaching peaks of
40% for certain receiver positions. It can also be observed, as
shown in Section IV, that the operation of the system was de-
pendent on the particular environment between the transmitter
and the receiver.

VI. MEASUREMENTS RESULTS Fig. 16. Percentage capacity improvement versus SNR for 5 different receiver
locations in indoor environment (improvement defined as the normalized dif-
The proposed reconfigurable antenna was then tested taking ference in capacity between the best configuration between the sixteen possible
measurements on a 2 2 MIMO orthogonal frequency division
2
solutions of the reconfigurable antenna system and a 2 2 MIMO system with
fixed length antennas).
multiplexing (OFDM) testbed communication system in an in-
door environment. Each node of the experimental platform con-
sisted of frequency agile transceivers operating in the ISM and (BPSK) was transmitted independently over the two transmit-
UNII radio bands and a baseband process computer. The base- ters. This training pattern was then received and used to estimate
band chassis provided by National Instruments had two major the channel matrix [36]. The orientation of the array, depicted
functional roles. First, the unit contained the analog to digital in Fig. 12(b), has been selected such the maximum degree of
(A/D) and (D/A) converters required for the two transceivers. pattern diversity between the different antenna’s configuration
The converters operated at 100 MS/s with 14-bit quantization. is in the azimuthal plane.
Second, the baseband unit was a software defined radio (SDR) To determine the capacity of the MIMO wideband communi-
which allowed the physical layer to be flexible in implementing cation link, a normalization of the channel matrix for each sub-
different experiments. An overview of the testbed can be seen carrier was computed using the Frobenius norm as described
in Fig. 15. in Section IV, in order to remove the differences in path loss
Two nodes have been used for measurement. As shown in among a number of channel matrices [35]. Specifically, all the
Fig. 12(b), the measurements were taken in the hallway of the channel matrices for each receiver location and for each sub-car-
3rd floor of the Bossone Research building on Drexel Univer- rier was normalized with respect to the “short” configuration
sity campus. The transmitter was stationary, while the receiver channel matrix at the respective location and sub-carrier.
was moved between several different locations. The channel ma- The capacity of the link was then determined using the
trix was measured for 5 different locations of the receiver and equation
for all 16 possible configurations of the antenna system. The
measurements were performed at 2.484 GHz. We used BPSK to
generate the analog baseband signal. The analog signal obtained (9)
was modulated using OFDM with the data being sent on each of
the 52 sub-carriers. The spacing between each sub carrier was where was the total number of sub-carriers. Since the channel
312.5 kHz. A training pattern using binary phase shift keying was characterized over a broad frequency band, the capacity of
PIAZZA et al.: DESIGN AND EVALUATION OF A RECONFIGURABLE ANTENNA ARRAY FOR MIMO SYSTEMS 879

Fig. 17. Percentage capacity improvement versus SNR for 5 different receiver locations in indoor environment for a narrow-band channel at 2.48 GHz: (a) results
from measurements, (b) results from simulations.

the wideband channel was defined as an average of the capaci- Fig. 13(a)] confirming the improvement that the reconfigurable
ties over all the sub-carriers of the MIMO-OFDM link [37] MIMO antenna solution achieved in terms of capacity.
and the optimal solution for the reconfigurable antenna was the In Fig. 17 is then shown the percentage improvement in
one which guaranteed the highest average capacity. capacity achievable with the reconfigurable antenna in a nar-
In Fig. 16 the percentage improvement in channel capacity, rowband channel for an indoor environment. The percentage
achieved with the novel reconfigurable antenna array with re- improvement is defined in the same way as in Fig. 16, but
spect to a system where a fixed “short” configuration was used it is calculated for a single subcarrier centered at 2.48 GHz.
for every antenna array element at both transmitter and receiver, In Fig. 17 the percentage improvement obtained through
was plotted versus different SNR values for all 5 receiver lo- measurements for the five different locations of the receiver
cations. At each location the capacity improvement was de- [see Fig. 17(a)] is compared with the one achieved through
fined as the difference in capacity between the best configuration ray-tracing simulations conducted using FASANT, as in
between the sixteen possible configurations (i.e. reconfiguring Section V, for the same five locations and for the same fre-
both transmit and receive arrays) of the reconfigurable antenna quency of 2.48 GHz [Fig. 17(b)]. The results show a substantial
system and a 2 2 MIMO system with fixed, “short” config- agreement between the measured and the simulated results,
uration antennas. In order to define a “relative” improvement, confirming therefore the benefit of using the proposed recon-
the channel capacity improvement was normalized for each lo- figurable antenna in indoor environments.
cation with respect to the capacity measured with the “short”
configuration antennas. Fig. 16 shows that the reconfigurable VII. CONCLUSION
antenna solution increased MIMO link capacity with respect to A MIMO system using a reconfigurable antenna was intro-
a conventional, fixed antenna system. Only for receiver posi- duced and analyzed in terms of capacity. The variation in the
tion 1 there was no improvement since, in that location, the best antenna geometry and in the mutual coupling effect between the
configuration of the reconfigurable antenna corresponded to the radiating elements was exploited in order to generate different
reference “short” configuration. On average, the novel antenna radiation patterns. This pattern diversity was then exploited on
solution achieved a 10% improvement in capacity (with a peak a channel by channel basis to identify the antenna configuration
of 28%) for a SNR of 10 dB and 8% (with a peak of 16%) for a which provided the greatest level of channel capacity.
SNR of 20 dB with respect to a system with a fixed antenna con- The benefits deriving from using a reconfigurable MIMO
figuration. Fig. 16 also shows how the improvement in capacity antenna system was first motivated through a spatial cor-
achieved with this novel antenna solution varied with receiver relation coefficient analysis in a clustered MIMO channel
location and so with the particular multipath environment, as model. Through this analysis we demonstrated the advantage
previously verified with simulations in Sections IV and V. of switching between different antenna configurations based on
A narrowband analysis of the channel, like the MIMO clus- the spatial characteristics of the MIMO channel.
tered channel model analysis (Section IV), could be performed Channel capacity was simulated using computational electro-
by considering the channel on each OFDM sub-carrier to be an magnetics in an outdoor environment and was also computed
independent narrowband channel realization. This technique al- taking indoor field measurements with a 2 2 MIMO system.
lows for a total of 260 narrowband channel realizations (5 loca- Results have shown how the reconfigurable antenna solution de-
tions times 52 sub-carriers for each location). For this narrow- scribed in this paper could provide an average improvement in
band analysis, results were shown in terms of the capacity CDF. capacity, for a SNR of 10 dB and 20 dB, of 10% and 8% respec-
In Fig. 13(b) it was shown how the results from measurements tively in a 2 2 MIMO link with respect to a system which did
were comparable to the results obtained from the simulation [see not have this reconfiguration property.
880 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 56, NO. 3, MARCH 2008

The reconfigurable antenna for MIMO systems is an attrac- [16] H. Ozcelik, M. Herdin, W. Weichselberger, J. Wallace, and E. Bonek,
tive solution for handheld devices (where space is an important “Deficiencies of ‘Kronecker’ MIMO radio channel model,” Electron.
Lett., vol. 39, no. 16, pp. 1209–1210, 2003.
constraint) to maintain good communication link capacity. [17] S. Wyne, A. F. Molisch, P. Almers, G. Eriksson, J. Karedal, and F.
The proposed antenna is an initial demonstration of reconfig- Tufvesson, “Statistical evaluation of outdoor-to-indoor office MIMO
urable MIMO system with only two switches per antenna array measurements at 5.2 GHz,” in IEEE Vehicular Technology Conf., 2005,
element. Future research in this field can then be envisioned vol. 61, no. 1, pp. 146–150.
[18] C. Oestges, E. Ozcelik, and H. Bonek, “On the prectical use of ana-
in: i) developing novel antenna solutions with higher level lytical MIMO channel models,” in IEEE Antennas and Propagation
of reconfigurability to achieve higher capacity improvement; Society Int. Symp., 2005, vol. 3B, pp. 406–409.
ii) studying techniques for integrating a switch configuration [19] K. R. Dandekar, G. Xu, and H. Ling, “Computational electromagnetic
simulation of smart antenna systems in urban microcellular environ-
selection process into practical MIMO-aware ad hoc network ments,” IEEE Trans. Veh. Technol., vol. 52, no. 4, pp. 733–742, 2003.
nodes; and iii) determining the interval of time over which [20] M. Jankiraman, Space Time Codes and MIMO Systems. Boston, MA:
a switch configuration is optimal before a new configuration Artech House, 2004.
needs to be selected. [21] J. P. Kermoal, L. Schumacher, K. I. Pedersen, P. E. Mogensen, and
F. Frederiksen, “A stochastic MIMO radio channel model with exper-
imental validation,” IEEE J. Sel. Areas Commun., vol. 20, no. 6, pp.
ACKNOWLEDGMENT 1211–1226, 2002.
[22] D.-S. Shiu, G. J. Foschini, M. J. Gans, and J. M. Kahn, “Fading corre-
National Instruments has provided equipment donations sup- lation and its effect on the capacity of multielement antenna systems,”
porting this work. The authors would like to thank R. Spring, IEEE Trans. Commun., vol. 48, no. 3, pp. 502–513, 2000.
L. Chao, and A. S. Khemka for their helpful comments and sug- [23] A. Forenza, G. Wan, and R. W. Heath, Jr., “Optimization of 2-element
gestions concerning this work. arrays of circular patch antennas in spatially correlated MIMO chan-
nels,” presented at the IEEE Int. Waveform Diversity and Design Conf.,
Jan. 23–27, 2006.
REFERENCES [24] H. R. Chuang and L. C. Kuo, “3-D FDTD design analysis of a 2.4-GHz
[1] G. J. Foschini and M. J. Gans, “On limits of wireless communications polarization-diversity printed dipole antenna with integrated balun and
in a fading environment when using multiple antennas,” Wireless Per- polarization-switching circuit for WLAN and wireless communication
sonal Commun., vol. 6, no. 3, pp. 311–335, 1998. applications,” IEEE Trans. Microw. Theory Tech., vol. 51, no. 2 I, pp.
[2] A. F. Molisch, M. Steinbauer, M. Toeltsch, E. Bonek, and R. S. Thoma, 374–381, 2003.
“Capacity of MIMO systems based on measured wireless channels,” [25] C. A. Balanis, Antenna Theory: Analysis and Design. New York:
IEEE J. Sel. Areas Commun., vol. 20, no. 3, pp. 561–569, 2002. Wiley, 1997.
[3] V. Tarokh, N. Seshadri, and A. R. Calderbank, “Space-time codes for [26] D. B. Davidson, I. P. Theron, U. Jakobus, F. M. Landstorfer, F. J. C.
high data rate wireless communication: Performance criterion and code Meyer, J. Mostert, and J. J. van Tonder, “Recent progress on the antenna
construction,” IEEE Trans. Inf. Theory, vol. 44, no. 2, pp. 744–765,
simulation program FEKO,” in Proc. South African Symp. on Commu-
1998.
nications and Signal Processing, COMSIG, 1998, pp. 427–430.
[4] T. Svantesson, “An antenna solution for MIMO channels: The multi-
mode antenna,” in Conf. Record of the Asilomar Conf. on Signals, Sys- [27] M. L. Morris and M. A. Jensen, “Network model for MIMO systems
tems and Computers, 2000, vol. 2, pp. 1617–1621. with coupled antennas and noisy amplifiers,” IEEE Trans. Antennas
[5] C. B. Dietrich, Jr., K. Dietze, J. R. Nealy, and W. L. Stutzman, “Spa- Propag., vol. 53, no. 1, pt. II, pp. 545–552, 2005.
tial, polarization, and pattern diversity for wireless handheld termi- [28] R. G. Vaughan and J. B. Andersen, “Antenna diversity in mobile com-
nals,” IEEE Trans. Antennas Propag., vol. 49, no. 9, pp. 1271–1281, munications,” IEEE Trans. Veh. Technol., vol. T-36, no. 4, pp. 149–172,
2001. 1987.
[6] J. P. Kermoal, L. Schumacher, P. E. Mogensen, and K. I. Pedersen, [29] M. A. Jensen and Y. Rahmat-Samii, “Performance analysis of an-
“Experimental investigation of correlation properties of MIMO radio tennas for hand-held transceivers using FDTD,” IEEE Trans. Antennas
channels for indoor picocell scenarios,” in Proc. IEEE Vehicular Tech- Propag., vol. 42, no. 8, pp. 1106–1112, 1994.
nology Conf., 2000, vol. 1, no. 52D, pp. 14–21. [30] P. Soma, D. S. Baum, V. Erceg, R. Krishnamoorthy, and A. J. Paulraj,
[7] T. Svantesson and A. Ranheim, “Mutual coupling effects on the ca- “Analysis and modeling of multiple-input multiple-output (MIMO)
pacity of multielement antenna systems,” in Proc. IEEE Int. Conf. on radio channel based on outdoor measurements conducted at 2.5 GHz
Acoustics, Speech and Signal Processing ICASSP, 2001, vol. 4, pp. for fixed BWA applications,” in Proc. IEEE Int. Conf. on Communi-
2485–2488.
cations, 2002, vol. 1, pp. 272–276.
[8] A. Forenza and R. W. Heath, Jr., “Benefit of pattern diversity via two-
element array of circular patch antennas in indoor clustered MIMO [31] L. M. Correia, Wireless Flexible Personalized Communications. New
channels,” IEEE Trans. Commun., vol. 54, no. 5, pp. 943–954, 2006. York: Wiley, 2001.
[9] W. H. Weedon, W. J. Payne, and G. M. Rebeiz, “MEMS-switched re- [32] H. Shin and J. H. Lee, “Capacity of multiple-antenna fading channels:
configurable antennas,” in IEEE Antennas and Propagation Society, Spatial fading correlation, double scattering, and keyhole,” IEEE Trans.
AP-S Int. Symp. Digest, 2001, vol. 3, pp. 654–657. Inf. Theory, vol. 49, no. 10, pp. 2636–2647, 2003.
[10] G. H. Huff, J. Feng, S. Zhang, and J. T. Bernhard, “A novel radia- [33] M. F. Catedra, J. Perez, A. Gonzalez, O. Gutierrez, and F. Saez de
tion pattern and frequency reconfigurable single turn square spiral mi- Adana, “Fast computer tool for the analysis of propagation in urban
crostrip antenna,” IEEE Microw. Wireless Compon. Lett., vol. 13, no. cells,” in Proc. Annual Wireless Communications Conf., 1997, pp.
2, pp. 57–59, 2003. 240–245.
[11] B. A. Cetiner, H. Jafarkhani, J. Y. Qian, H. J. Yoo, A. Grau, and F. De [34] J. Perez, J. Ibanez, L. Vielva, and I. Santamaria, “Capacity estimation
Flaviis, “Multifunctional reconfigurable MEMS integrated antennas of polarization-diversity MIMO systems in urban microcellular envi-
for adaptive MIMO systems,” IEEE Commun. Mag., vol. 42, no. 12, ronments,” in IEEE Int. Symp. on Personal, Indoor and Mobile Radio
pp. 62–70, 2004. Communications, PIMRC, 2004, vol. 4, pp. 2730–2734.
[12] B. A. Cetiner, J. Y. Qian, G. P. Li, and F. De Flaviis, “A reconfig- [35] M. A. Jensen and J. W. Wallace, “A review of antennas and propagation
urable spiral antenna for adaptive MIMO systems,” Eurasip J. Wireless
for MIMO wireless communications,” IEEE Trans. Antennas Propag.,
Commun. Networking, vol. 2005, no. 3, pp. 382–389, 2005.
[13] B. A. Cetiner, E. Akay, E. Sengul, and E. Ayanoglu, “A MIMO system vol. 52, no. 11, pp. 2810–2824, 2004.
with multifunctional reconfigurable antennas,” IEEE Antennas Wire- [36] J. W. Wallace, M. A. Jensen, A. L. Swindlehurst, and B. D. Jeffs, “Ex-
less Propag. Lett., vol. 5, no. 12, pp. 463–466, 2006. perimental characterization of the MIMO wireless channel: Data ac-
[14] M. Wennstrom and T. Svantesson, “An antenna solution for MIMO quisition and analysis,” IEEE Trans. Wireless Commun., vol. 2, no. 2,
channels: The switched parasitic antenna,” in Proc. IEEE Int. Symp. on pp. 335–343, 2003.
Personal, Indoor and Mobile Radio Communications, PIMRC, 2001, [37] H. Bolcskei, D. Gesbert, and A. J. Paulraj, “On the capacity of OFDM-
vol. 1, pp. 159–163. based spatial multiplexing systems,” IEEE Trans. Commun., vol. 50,
[15] V. Erceg et al., TGn Channel Models IEEE 802.11-03/940r4, 2004. no. 2, pp. 225–234, 2002.
PIAZZA et al.: DESIGN AND EVALUATION OF A RECONFIGURABLE ANTENNA ARRAY FOR MIMO SYSTEMS 881

Daniele Piazza (S’03) received the B.S. and Laurea Robert W. Heath, Jr. (S’96–M’01–SM’06) re-
(with high honors) degrees in telecommunication ceived the B.S. and M.S. degrees from the University
engineering from the Politecnico di Milano, Italy, of Virginia, Charlottesville, in 1996 and 1997,
in 2003 and 2006, respectively, and the M.S. degree respectively, and the Ph.D. from Stanford University,
in electrical and computer engineering from Drexel Stanford, CA, in 2002, all in electrical engineering.
University, Philadelphia, PA, in 2006, where he is From 1998 to 2001, he was a Senior Member
currently working toward the Ph.D. degree. of the Technical Staff then a Senior Consultant at
While working toward the M.S. degree at Drexel Iospan Wireless Inc, San Jose, CA, where he worked
University, he conducted research focused on recon- on the design and implementation of the physical and
figurable antennas for MIMO communications and is link layers of the first commercial MIMO-OFDM
currently developing reconfigurable antenna systems communication system. In 2003, he founded MIMO
for adaptive MIMO communications. He is also working as a researcher at Po- Wireless Inc., a consulting company dedicated to the advancement of MIMO
litecnico di Milano in the field of radio frequency identification (RFId). His re- technology. Since January 2002, he has been with the Department of Electrical
search interests concentrate mainly on reconfigurable antennas, smart antenna and Computer Engineering, the University of Texas at Austin, where he is
arrays, MIMO communications and RFId systems. currently an Associate Professor and member of the Wireless Networking and
Communications Group. His research interests cover a broad range of MIMO
communication including limited feedback techniques, multihop networking,
multiuser MIMO, antenna design, and scheduling algorithms as well as 60
Nicholas J. Kirsch (S’00) received the B.S. degree GHz communication techniques.
in electrical engineering from the University of Dr. Heath is the recipient of the David and Doris Lybarger Endowed Faculty
Wisconsin-Madison, in 2003 and the M.S. degree Fellowship in Engineering. He was an Editor for the IEEE TRANSACTIONS
in telecommunications engineering from Drexel ON COMMUNICATION and an Associate Editor for the IEEE TRANSACTIONS
University, Philadelphia, PA, in 2006, where he is ON VEHICULAR TECHNOLOGY. He is a member of the Signal Processing for
currently working toward the Ph.D. degree. Communications Technical Committee in the IEEE Signal Processing Society.
From 2001 to 2002, he worked at W. L. Gore & As- He was a technical Co-Chair for the 2007 Fall Vehicular Technology Confer-
sociates on fiber optic link modules and long-wave- ence, and is the organizer of the 2008 Communication Theory Workshop, and
length lasers. His research interests include MIMO a Co-Organizer of the 2009 Signal Processing for Wireless Communications
communication systems, ad hoc networking, adap- Workshop.
tive radio systems, and reconfigurable antennas.

Kapil R. Dandekar (S’95–M’01–SM’07) received

Antonio Forenza (S’04–M’06) received the M.S. de- the B.S. degree in electrical engineering from the
gree in telecommunications engineering from the Po- University of Virginia, Charlottesville, in 1997
litecnico di Torino, Italy, and the Eurecom Institute, and the M.S. and Ph.D. degrees in electrical and
Sophia Antipolis, France, in 2001, and the Ph.D. de- computer engineering from the University of Texas
gree in electrical and computer engineering from the at Austin, in 1998 and 2001, respectively.
University of Texas at Austin, in 2006. In 1992, he worked at the U.S. Naval Observatory
In 2001, he interned as a Systems Engineer and from 1993 to 1997, he worked at the U.S.
at Iospan Wireless, Inc., San Jose, CA, a startup Naval Research Laboratory. In 2001, he joined the
company that developed the first commercial Electrical and Computer Engineering Department,
MIMO-OFDM communication system. His main Drexel University, Philadelphia, PA, as an Assis-
research focus was on link-adaptation and physical tant Professor. He is currently the Director of the Drexel Wireless Systems
layer algorithm design. In fall 2001, he joined ArrayComm, Inc., San Jose, CA, Laboratory (DWSL), which has been supported by the U.S. National Science
as a Systems Engineer. In ArrayComm, he was actively involved in the design Foundation, Army CERDEC, National Security Agency, Office of Naval
and implementation of smart antenna systems for the 3G WCDMA wireless Research, and private industry. His current research interests involve MIMO
platform. During summers 2004 and 2005, he interned as a Research Engineer ad hoc networks, reconfigurable antennas, free space optical communications,
at Samsung Advanced Institute of Technology (SAIT), Suwon, Korea, and ultrasonic communications, and sensor networks. He has published articles
Freescale Semiconductor, Inc., Austin, TX, respectively, developing adaptive in several journals including the IEEE TRANSACTIONS ON ANTENNAS AND
MIMO transmission and MU-MIMO precoding techniques for 3GPP, IEEE PROPAGATION, IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, IEEE
802.11n and IEEE 802.16e standards systems. Since June 2006, he has been TRANSACTIONS ON COMMUNICATIONS, IEEE TRANSACTIONS ON VEHICULAR
working for Rearden, LLC, San Francisco, CA, as a Senior Systems Engineer. TECHNOLOGY, and IEE Electronics Letters.
His research interests include MIMO antenna design, adaptive MIMO trans- Dr. Dandekar currently serves on the editorial board of IEEE Expert Now
mission techniques, precoding methods for MU-MIMO, smart antenna signal and serves on the Pre-University Education Committee of the IEEE Educational
processing. Activities Board. He is a continuing active member of the Technical Program
Committee for the IEEE Radio and Wireless Symposium. He served as an Asso-
ciate Editor for the IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY and is
currently the Chairman of the Philadelphia chapter of the IEEE Vehicular Tech-
nology Society.