Available online at www.sciencedirect.

com

ScienceDirect
Procedia Computer Science 40 (2014) 181 189

Fourth International Conference on Selected Topics in Mobile & Wireless Networking

(MoWNet2014)

A Cooperative Beamforming Technique

for Body Area Networks
Stefan Mijovica,, Chiara Burattia , Alberto Zanellab , Roberto Verdonea
b

a
CNIT, DEI, University of Bologna, V.le del Risorgimento 2, 40136 Bologna, Italy
IEIIT, Consiglio Nazionale delle Ricerche, V.le del Risorgimento 2, 40136 Bologna, Italy

Abstract
We consider an indoor environment with multiple Body Area Networks (BANs) that have to transmit data towards specic sinks,
located in xed positions. Nodes deployed on the same body may cooperate in order to form Virtual Antenna Arrays (VAAs) and
transmit data towards one of the available sinks. Sinks are also equipped with multiple antennas, such that a virtual Multiple Input
Multiple Output (MIMO) channel is established. Beamforming is used as a cooperation technique on both transmitter and receiver
sides. A simple technique, consisting of decimating the number of cooperating nodes in the VAA with the aim of reducing energy
consumption as well as interference, is presented. Results show that the proposed technique improves the performance in terms
of energy eciency, and also in terms of block error rate when the system is interference-limited. Performance is evaluated by
applying dierent well-known scheduling strategies.
c 2014

2014
The Authors.
Published
B.V.

Published
by Elsevier
B.V. This isby
anElsevier
open access
article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of organizing committee of Fourth International Conference on Selected Topics in Mobile &
Peer-review under responsibility of organizing committee of Fourth International Conference on Selected Topics in
Wireless Networking (MoWNet2014).
Mobile & Wireless Networking (MoWNet2014)
Keywords: body area networks;cooperative communications;virtual mimo;virtual antenna array;beamforming

1. Introduction
Multiple Input Multiple Output (MIMO) technique is a well-known tool for increasing link capacity as well as
reliability. When devices cannot be equipped with multiple antennas, cooperative MIMO technique (also known as
virtual MIMO) could be exploited. Cooperative MIMO inherits most of the advantages of MIMO systems, with
the dierence that antenna elements are independent devices forming Virtual Antenna Arrays (VAAs). Cooperative
MIMO concept may be applied to Body Area Networks (BANs), where devices on the body are equipped with a
single antenna and where requirements in terms of reliability and energy eciency are very stringent. Even though it
introduces some communication overhead, cooperation among nodes in the VAA increases reliability and reduces the
total energy consumption 1.

Corresponding author. Tel.: +39-051-20-93180.

E-mail address: [email protected]

1877-0509 2014 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Peer-review under responsibility of organizing committee of Fourth International Conference on Selected Topics in
Mobile & Wireless Networking (MoWNet2014)
doi:10.1016/j.procs.2014.12.026

182

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189

To the authors knowledge, the rst paper to suggest the application of the MIMO concept to BAN is 2 , where it
is shown that the use of MIMO signicantly improves the channel capacity with respect to conventional systems. In
this work cooperating nodes deployed on the body, are assumed to be connected through wired links. Wireless communication among cooperative nodes is considered in 3 , where results show that, apart from the increase of capacity,
signicant advantage can be obtained in terms of interference rejection. The application of cooperative MIMO in
BANs is further investigated in 4 where the authors develop a simple but eective cooperative diversity scheme for
Ultra Wide Band based BAN that has the objective of improving bit error rate performance with respect to Single
Input Single Output (SISO) case.
As far as the study of Medium Access Control (MAC) protocols for cooperative MIMO systems is concerned, 5
presents NetEigen, a protocol mitigating interference and maximising the desired received signal power. The latter eect is achieved by properly setting both transmitter and receiver weight vectors. Most recently, a cooperative
scheduling framework that closely relies on NetEigen MAC has been proposed in 6 . A network-coding-based cooperative ARQ MAC protocol is presented in 7 . The scheme proposed in that work achieves better energy eciency with
respect to state of the art protocols without compromising the oered quality of service. A good overview of MAC
protocols for cooperative communication systems is given in 8 .
It is worth noting that cooperative beamforming implies additional transmit power constraints with respect to
standard beamforming. Since each element is an actual device with its own power amplier, a constraint on transmit
power of each element needs to be imposed. It is well known that this limits the performance gain achieved by the
standard beamforming, but still provides great advantage over SISO case 9 .
In contrast with the above works, we present B-MIMO, a cooperative MIMO scheme envisaged for BANs. BMIMO applies cooperative beamforming to BANs, by optimising the number of cooperating nodes according to their
channel conditions. Due to the heterogeneity of the BAN channel, nodes experience very dierent behaviors: nodes
shadowed by the body need to cooperate to reach the requested signal to noise ratio; while nodes in good channel
condition may not benet from the cooperation with nodes that suer from bad channel quality. B-MIMO reduces the
number of cooperating nodes, depending on channel status, with the objective of reducing the energy consumption as
well as the level of interference.
B-MIMO is applied to an indoor scenario where dierent bodies have to transmit data to a selected sink. Realistic
settings, including nodes distribution and channel model, are considered. The proposed solution is compared to a
cooperative MIMO solution, where all nodes in the BAN cooperate and no decimation is applied. Results show that
the proposed technique improves the performance in terms of energy eciency, and also in terms of Block Error
Rate (BLER) when the system is interference-limited. Performance is evaluated by considering dierent well-known
scheduling strategies, that are maximum throughput, proportional fairness and round robin. The design of a novel
scheduling strategy is out of the scope of this work.
The rest of the paper is organised as follows: Sec. II introduces the reference scenario and the channel model; Sec.
III presents the communication protocol and the cooperative MIMO scheme based on beamforming. Sec. IV describes
the proposed B-MIMO scheme and Sec. V briey presents the well-known scheduling algorithms considered in the
paper. Numerical results and conclusions are reported in Sec.s VI and VII, respectively.
2. Reference Scenario and Channel Model
We consider an indoor environment (i.e., a hospital room) with several people (patients) (see Fig. 1). Each patient
is equipped with a BAN consisting of three nodes placed on left hip, heart and right ear. The nodes located on the
body have to transmit multimedia data to the sinks. Cooperative beamforming is used to transmit data towards a sink.
Sinks, with multiple antennas, are placed on the walls of the room. Bodies, represented by elliptical cylinders in 3D,
are located in random positions and with random orientations in the area.
Human body is a very specic environment where body shadowing plays a major role. In order to properly account
for the propagation environment, the channel model based on an extensive measurement campaign presented in 10 is
used in this work. In 10 measurements are performed in an indoor laboratory furnished with tables, chairs and some
general equipment such that the model captures the eects of real environment. The identied model is the most
suitable for our scenario, since it is based on measurements performed by locating nodes exactly in the same positions
considered in this work.

183

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189

10

y [m]

Sinks
Nodes

BANs
0

10

x [m]

Fig. 2. Scenario X-Y plane, denition of angle .

Fig. 1. Scenario 3D.

According to 10 the channel attenuation can be modeled as:

P(d, )|dB = G0 (d)|dB + S ()|dB + F|dB

(1)

where d is the distance between the transmit node on body and the sink, is the relative angle between the body and
the sink (see Fig. 2), G0 (d) is the mean channel gain, S() is the body shadowing component of the channel transfer
function, F is its fading component. For more information on the channel model and the relevant parameters, refer
to 10 . We also apply a xed 20 dB loss when the signal passes through a body. As for the on-body communication, we
assume that all nodes may communicate among them without losses by employing techniques presented in 11 and 12 .
3. The Communication Protocol
We assume that: i) a wired connection among sinks is present, such that they can maintain synchronisation; ii) time
is divided into frames, starting with a query sent by the sink(s). Upon reception of queries each BAN will associate to
the strongest sink and the association is announced by short packets transmitted by each node in a BAN. Considering
that a CSMA-based protocol is used and the small size of packets, we can assume that there are no losses in this phase.
The importance of this step is twofold: i) a sink will know how many BANs are associated to it; ii) a sink can estimate
the uplink channel for each BAN associated to it.
We assume that each frame is divided into time slots, assigned by each sink to the BANs associated to it. Once a sink
denes the schedule, at the beginning of each time slot, it will poll the BAN which is scheduled to transmit in that time
slot. Poll packets include previously computed channel estimation and the target source (one out of the three nodes,
randomly selected by the sink). Once the source node receives the poll, it computes the beamforming weight vector
(based on the full channel estimated by the sink) and it transmits a broadcast packet, containing the beamforming
weights and the data to be transmitted, to the cooperating nodes. This packet is also used for synchronisation, and it
is followed by the cooperative data transmission towards the sink.
In this paper we consider a cooperative MIMO scheme, where we account for the fact that each antenna element
has its own transmit power limitations. By denoting as NT(0) the number of antennas at the useful transmitter and as NR
the number of antennas at the receiver, the received signal, when aected by Nint interferers each one equipped with
NT(i) antennas, with i = 1, 2, ..., Nint, can be expressed as:
y = H(0) (0) x(0) +

Nint

i=1

H(i) (i) x(i) + n

184

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189

where y CNR 1 is a vector, x(i) C is the input symbol of transmitter i, H(i) CNR NT is the channel gain matrix
between transmitter i and the receiver, (i) CNT 1 is the weight vector of transmitter i, and n CNR 1 is the thermal
noise vector. We assume that E{n nH } = 2n I, where 2n is the thermal noise power per antenna element, I CNR NR
is the identity matrix, E{} denotes expectation, and {.}H denotes the conjugate transpose. The output symbol after the
linear combiner at the sink can be expressed as
z = w H H(0) (0) x(0) +

Nint


w H H(i) (i) x(i) + w H n

i=1

where w CNR 1 is the weight vector of the receiver.

The signal-to-noise ratio (SNR) and the signal-to-interference ratio (SIR) can be written as
S NR =

H (0) (0) 2
P(0)
t |w H |
2n

P(0) |w H H(0) (0) |2

S IR = N t (i)
.
int
H (i) (i) 2
i=1 Pt |w H |

(2)

(3)

In the case of collocated MIMO beamforming, the method for computing the transmit weight vector which optimises
SNR is well known and presented in 13 . On the other hand, in order to properly account for the fact that antennas are
distributed and not collocated, we impose the following constraint on the transmit power of each element:
2
|(i)
j | =

1
NT(i)

, i = 1,2,. . . ,Nint , j = 1,2,. . . ,NT(i)

(4)

which means that all cooperating devices, equipped with their own power ampliers, present the same transmitted
power. Transmit weight coecients can thus only aect the signal phase, ensuring constructive summing at the
receiver.
Unfortunately, the presence of the constraint (4) makes the problem of the optimization of the vectors w and (i)
very cumbersome. As a consequence, some sub-optimal techniques are generally applied. In particular, we use the
algorithm for the evaluation of the transmit and receive weight vectors presented in 9 .
4. The B-MIMO Scheme
Due to the specic body environment, and the heterogeneity of the channel, nodes of the same BAN experience
very dierent channel conditions, causing dierent performance. The latter justies the need of nodes experiencing
bad channel to cooperate with other nodes, to reach the requested quality of service. On the other hand, when the
source has good channel conditions, cooperation with nodes having a bad channel could be useless. In fact, nodes
experiencing a bad channel do not contribute signicantly to the power received by the destination sink, while they
may cause signicant interference towards other sinks. The other drawback is that some energy is wasted for a
transmission which makes a negligible impact on the overall performance.
To demonstrate the channel heterogeneity we consider a scenario with 10 bodies where nodes in three dierent
positions on body transmit packets towards sinks. Fig. 3 presents the cumulative distribution function (CDF) of SNR
achieved by the three nodes. As can be seen, the performance strongly depends on the position of a node. Moreover,
if we would consider a channel model typical for wireless sensor networks, the nodes would achieve not only similar
performance but also the SNR, at the considered distances, would be much higher, both eects originating from the
lack of the dominant shadowing component.
To ameliorate the above described eects, we introduce B-MIMO, a cooperation scheme where each node will participate in the cooperative transmission only if its contribution to the overall power received by the intended destination
is above a given threshold, denoted as . Each node can compute its contribution once the poll packet, containing the
channel estimation, is received. According to 9 , to compute its transmit weight, each node needs to compute also the

185

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189

1
0.9
0.8

CDF of SNRdB

0.7
0.6
0.5
0.4
0.3
0.2

Left hip
Heart
Right ear

0.1

SNR

-10

10

20

30

40

50

Fig. 3. CDF of SNR for nodes placed in dierent positions on body.

receive weight vector of the sink. Given the channel estimation and the receiver weight vector, an equivalent MISO
(0)
channel, h , can be estimated
(0)
(5)
h = w H H(0) .
In order to quantify the individual contribution to the overall received power of each node, we dene the ratio i as
i 

(0)
h (0)
i i
(0)
h (0)

, i = 1,2,. . . ,NT(0)

(6)

Therefore, if i is larger than , node i will cooperate with the source, otherwise it will not participate in the transmission towards the sink.
The method for selecting the value of parameter will be described in section 6.
5. Scheduling Algorithms
The proposal of a new scheduling algorithm is out of the scope of this paper and three well-known algorithms are
considered to schedule transmissions from BANs connected to the same sink: maximum throughput, proportional fair
and round robin. The goal of the rst one is to maximise the overall network throughput, disregarding the fairness.
Proportional fair, on the other hand, tends to equalize the throughput among all the BANs associated to the same sink,
thus maximizing the Jain index, which represents a measure of fairness 14 . Round robin is the most simple scheduling
algorithm in which all BANs get the same number of time slots, disregarding any performance metric. All three
algorithms can be modeled using the same mathematical formula. Suppose that in each time slot weight coecient
can be assigned to each BAN. Each sink will assign the next time slot to the BAN having the highest weight. Weight
of user i in slot k is given by:
xi,k
Wi,k =
(7)
k


( xi, j )
j=1

where xi,k is the estimation of performance achieved by user i if slot k is assigned to it, and , are coecients
dening the behavior of the scheduling algorithm. As a performance metric to be used by the scheduling algorithm
we choose the information rate, given by R = log2 (1 + S NR) [bits/s/Hz], regardless of the presence of interference.
By setting = 1 and = 0, the weight coecient depends only on the nominator. This means that the slot will be
assigned to the user that might achieve the best performance, which leads to maximum throughput algorithm. On the
other hand, if we set = 0 and = 1, the slot will be assigned to the user achieving the least throughput in the current
frame, thus leading to proportional fair algorithm. By setting both = 1 and = 1, we give equal importance to the
numerator and denominator. In our scenario, where channel is xed during a frame, the latter brings to round robin
scheduling algorithm. Fine tuning can be done by making dierent combinations of values of and .

186

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189











 








































































































         










Fig. 4. Scheduling scheme - example of round robin.

We assume each sink schedules transmissions disregarding the schedule of other sinks. Consequently, there might
be interference among BANs associated to dierent sinks. The communication among sinks is limited to synchronisation maintenance. An example of round robin is shown in Fig. 4.
6. Numerical Results
In this section, we rst introduce the metrics we use for performance estimation, then we provide some guidelines
for setting parameter and nally we compare the system performance when dierent scheduling algorithms and
cooperation schemes are employed.
Results have been obtained through a proprietary simulator, written in C++. A simulation round represents one
frame which consists of one hundred time slots. In each round BANs position and orientation are randomly and
uniformly distributed in the observed area. Results are obtained averaging over 10000 rounds. In simulations we set:
2
2
P(i)
t = 0 dBm, n = 110 dBm, S NR = 10 dB, S IR = 3 dB, NR = 3 and a room of 10 10 m , where the bodies have
a height of 180 cm and sinks are at 120 cm from the ground. We also set the packet size equal to 1 kB, according to
the multimedia application, and the slot time equal to 8 ms (i.e., a bit rate of 1 Mbit/s is considered).
6.1. Performance Metrics
Performance is evaluated in terms of Block Error Rate (BLER), that is the percentage of packets (i.e., blocks of
bits) generated by the dierent BANs which are not received correctly by the destination sink. BLER is determined
based on SNR and SIR: if during transmission both, SNR and SIR, are above two given thresholds S NR and S IR ,
respectively, the packet is considered to be successfully received, otherwise the packet is lost. We also evaluate the
energy eciency, [bit/s/mJ], that is the average number of bits per second received by the sinks, per mJ of energy
spent. Since the overhead generated in the network in order to establish the VAAs and to perform the beamforming
transmission is the same for all the cooperative strategies and scheduling algorithms considered, in the results we just
consider the energy spent for the cooperative data transmission.
6.2. The Impact of the Threshold
In section 4 we introduced parameter which represents the threshold for the individual contribution of each
node to overall received power at the sink. An increase of lowers the number of nodes participating in cooperative
transmission, meaning lower transmit power per BAN. The eect of the reduction of transmitting nodes is twofold: the
number of packets lost due to low SNR conditions increases but the number of packets lost due to low SIR decreases,
since the interference power is lower as well. Since the considered scenario is mostly noise-limited, the overall eect
is that the BLER value is slowly growing with (see Table 1). On the other hand, the energy consumption per BAN
decreases since the transmit power is lower. As a metric, energy eciency takes into account both BLER and energy

187

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189

2400

2300

[kbit/s/mJ]

2200

2100

2000

1900

1800

[%]

10

Fig. 5. Energy eciency as a function of the threshold .

consumption. The latter is the dominant factor as it can be seen in Fig. 5 which shows the energy eciency as a
function of .
Table 1. BLER for dierent values of parameter .

2%

4%

6%

8%

10%

BLER

0.2817

0.2950

0.3025

0.3108

0.3148

0.3193

In the following we set = 5%, since further increasing keeps energy eciency almost constant, while causing
an increasing of the BLER.
6.3. B-MIMO versus Cooperative MIMO
Fig. 6 shows BLER as a function of number of BANs in the scenario, denoted as NBAN . Both cooperative MIMO
and B-MIMO schemes are considered when the three previously described scheduling algorithms are employed. The
maximum throughput algorithm oers the best performance in terms of BLER, which is expected since sinks will
assign all the slots to the BAN having the best link quality, thus overall BLER of the system will be equal to the one
of the best performing BAN. As the number of BANs increases, the variety of link qualities increases, meaning that
the probability that at least one BAN has a good connection with its sink is higher. The latter justies the increasing
of BLER with NBAN . In the case of proportional fair if a BAN is performing badly, it will be assigned more slots to
reduce the performance gap with the other BANs. The latter implies that if there is at least one BAN with a bad link to
its sink, it will be assigned most of the slots in the frame, thus causing an increase in BLER. Round robin represents
an intermediate solution, where BLER slightly increases with NBAN due to the increasing of the level of interference.
In such scenario, cooperative MIMO outperforms B-MIMO in terms of BLER, since less power is transmitted in the
B-MIMO and the system is mostly noise-limited, therefore the advantage of B-MIMO over conventional cooperative
MIMO vanishes.
On the other hand, B-MIMO outperforms cooperative MIMO in terms of energy eciency because the energy
saved due to lower transmit power is more signicant then performance loss in terms of BLER. The latter is presented
in Fig. 7, where energy eciency is shown as a function of NBAN .
Since the energy consumption is independent on scheduling algorithm, the energy eciency of dierent scheduling
algorithms depends only on the BLER. Having taken into account the previous statement, the best performance is
achieved by maximum throughput algorithm since it achieves the lowest BLER. On the other hand BLER of round
robin and proportional fair is increasing with the number of BANs in scenario, which brings to the inverse behaviour
of the energy eciency curves.

188

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189

0.4

3200

Maximum throughput-BMIMO
Round robin smart-BMIMO
Proportional fair smart-BMIMO
Maximum throughput-cooperative MIMO
Round robin-cooperative MIMO
Proportional fair-cooperative MIMO

3000
2800
2600

BLER

[kbit/s/mJ]

0.3

0.2

2400
2200
2000
1800

0.1

1600

Maximum throughput-BMIMO
Round robin smart-BMIMO
Proportional fair smart-BMIMO
Maximum throughput-cooperative MIMO
Round robin-cooperative MIMO
Proportional fair-cooperative MIMO

10

10

NBAN

BAN

Fig. 7. Energy eciency of dierent scheduling algorithms and

cooperation schemes.

Fig. 6. BLER of dierent scheduling algorithms and cooperation

schemes.

0.12

0.1

Maximum throughput-BMIMO
Round robin-BMIMO
Proportional fair-BMIMO
Maximum throughput-cooperative MIMO
Round robin-cooperative MIMO
Proportional fair-cooperative MIMO

BLER

0.08

0.06

0.04

0.02

10

NBAN

Fig. 8. BLER of dierent scheduling algorithms and cooperation schemes in interference-limited scenario.

It is worth noting that even though maximum throughput oers the best performance in terms of both BLER and
energy eciency, it is the least fair scheduling algorithm, i.e., its Jain index is the minimum possible. On the other
hand, proportional fair maximises this metric at the cost of higher BLER and lower energy eciency.
In order to consider an interference-limited system, we set the transmit power to 10 dBm and the noise power to
120 dBm, such that no packets are lost for power conditions and losses are only due to low levels of SIR. Note that
in the presence of more sinks (i.e., more BANs transmitting at the same time), the scenario can become interferencelimited, even though the receiver is characterised by a larger noise power.
Fig. 8 presents BLER as a function of the number of BANs in an interference limited scenario. First we can note
that all the three scheduling algorithms have similar behaviour. This is explained by the fact that the scheduling is
based on SNR which is not the dominant metric for BLER performance. Maximum throughput is still performing the
best, since BAN with the highest useful signal power has higher probability of having good SIR, while proportional
fair and round robin have almost the same performance.
An interesting fact is that by employing B-MIMO we achieve not only better energy eciency, but also a noticeable performance gain in terms of BLER over cooperative beamforming, due to lower interference introduced. The
conclusion is that apart from energy savings, B-MIMO systems can achieve performance improvement in interferencelimited scenarios.

Stefan Mijovic et al. / Procedia Computer Science 40 (2014) 181 189

7. Conclusions
In this paper we presented B-MIMO, a cooperative beamforming scheme thought for body area networks. We
considered an indoor scenario where multiple BANs are present and nodes belonging to the same BAN establish
VAAs to transmit data towards a properly selected sink. A novel cooperation scheme, B-MIMO, where the number
of cooperating nodes varies depending on channel conditions, is proposed. Comparison between B-MIMO and a
cooperative solution where all nodes in the BAN belong to the VAA demonstrates the advantage of the novel solution
in terms of energy eciency and also in terms of BLER, in interference-limited scenarios. The decreasing of the
number of cooperating nodes, in fact, decreases the energy consumption and the level of interference generated by
BANs.
Acknowledgment
This work was supported by the European Commission in the framework of the FP7 Network of Excellence in
Wireless COMmunications NEWCOM (Grant agreement no. 318306).
References
1. X. Li, G. Kang, X. Zhang, D. Huang, An energy-ecient cooperative MIMO strategy for Wireless Sensor Networks with intrabody channel, in: International Symposium on Communications and Information Technologies (ISCIT), 2012, 2012, pp. 679684.
doi:10.1109/ISCIT.2012.6380987.
2. Y. Ouyang, D. Love, W. Chappell, Body-Worn Distributed MIMO System, IEEE Transactions on Vehicular Technology 58 (4) (2009) 1752
1765. doi:10.1109/TVT.2008.2004491.
3. I. Khan, P. Hall, Y. Nechayev, L. Akhoondzadeh-Asl, Multiple antenna systems for increasing on-body channel capacity and reducing BANto-BAN interference, in: International Workshop on Antenna Technology (iWAT), 2010, 2010, pp. 14. doi:10.1109/IWAT.2010.5464736.
4. J. Ding, E. Dutkiewicz, X. Huang, Performance evaluation of virtual MIMO for UWB based body area networks, in: 7th International Symposium on Medical Information and Communication Technology (ISMICT), 2013, 2013, pp. 2832. doi:10.1109/ISMICT.2013.6521693.
5. P. Zhao, B. Daneshrad, Net-eigen MAC: A new MIMO MAC solution for interference-oriented concurrent link communications, in: MILITARY COMMUNICATIONS CONFERENCE, 2011 - MILCOM 2011, 2011, pp. 248253. doi:10.1109/MILCOM.2011.6127669.
6. P. Zhao, Y. Lu, B. Daneshrad, M. Gerla, Cooperative spatial scheduling in distributed MIMO MAC with interference/concurrency awareness,
in: MILITARY COMMUNICATIONS CONFERENCE, 2012 - MILCOM 2012, 2012, pp. 16. doi:10.1109/MILCOM.2012.6415728.
7. A. Antonopoulos, C. Verikoukis, Network-Coding-Based Cooperative ARQ Medium Access Control Protocol for Wireless Sensor Networks,
International Journal of Distributed Sensor Networks 2012. doi:10.1155/2012/601321.
8. R. Khan, H. Karl, MAC Protocols for Cooperative Diversity in Wireless LANs and Wireless Sensor Networks, IEEE Communications
Surveys Tutorials, 16 (1) (2014) 4663. doi:10.1109/SURV.2013.042313.00067.
9. X. Zheng, Y. Xie, J. Li, P. Stoica, MIMO Transmit Beamforming Under Uniform Elemental Power Constraint, IEEE Transactions on Signal
Processing 55 (11) (2007) 53955406. doi:10.1109/TSP.2007.896058.
10. R. Rosini, R. DErrico, O-Body channel modelling at 2.45 GHz for two dierent antennas, in: 6th European Conference on Antennas and
Propagation (EUCAP), 2012, 2012, pp. 33783382. doi:10.1109/EuCAP.2012.6206591.
11. K. S. Prabh, J.-H. Hauer, Opportunistic Packet Scheduling in Body Area Networks, in: Proceedings of the 8th European Conference on
Wireless Sensor Networks, EWSN11, Springer-Verlag, Berlin, Heidelberg, 2011, pp. 114129.
12. K. Prabh, F. Royo, S. Tennina, T. Olivares, BANMAC: An Opportunistic MAC Protocol for Reliable Communications in Body Area
Networks, in: IEEE 8th International Conference on Distributed Computing in Sensor Systems (DCOSS), 2012, 2012, pp. 166175.
doi:10.1109/DCOSS.2012.37.
13. P. Dighe, R. Mallik, S. Jamuar, Analysis of transmit-receive diversity in Rayleigh fading, IEEE Transactions on Communications 51 (4)
(2003) 694703. doi:10.1109/TCOMM.2003.810871.
14. R. K. Jain, The Art of Computer Systems Performance Analysis: Techniques for Experimental Design, Measurement, Simulation, and
Modeling, 1st Edition, Wiley, 1991.

189