A Proposal for Network Coding with the IEEE 802.15.6 Standard

Xiaomeng Shi and Muriel M´edard

Department of Electrical Engineering and Computer Science,

Massachusetts Institute of Technology, 77 Massachusetts Ave., Cambridge, MA, U.S.A.

Keywords:

Network Coding, IEEE 802.15.6, WBAN, Energy Efﬁciency.

Abstract:

We examine the Medium Access Control sublayer of the IEEE 802.15.6 Wireless Body Area Network

(WBAN) standard, and propose minor modiﬁcations to the standard so that linear random network coding can

be included to help improve energy efﬁciency and throughput of WBANs compatible with the standard. Both

generation-based and sliding window approaches are possible, and a group-block acknowledgment scheme can

be implemented by modifying block acknowledgment control type frames. Discussions on potential energy

and throughput advantages of network coding are provided.

1 INTRODUCTION

The IEEE 802.15.6 WBAN standard (IEEE, 2012)

was published by the IEEE 802.15 Task Group 6

(TG6) in 2012 to overcome limitations of other Per-

sonal Area Network standards such as IEEE 802.15.4

(Zigbee) and IEEE 802.15.1 (Bluetooth). Since its

initial draft proposal, many studies have been con-

ducted to analyze the performance of medium access

mechanisms of this standard. For example, (Ullah and

Kwak, 2011) presents theoretical throughput and de-

lay limits and bandwidth efﬁciency results for ideal

channels without transmission errors, while (Tach-

tatzis et al., 2010) analyzes the energy lifetime of pe-

riodic scheduled access modes.

In this article, we propose to incorporate the use

of network coding (NC) into a WBAN arranged in a

star topology. By examining the IEEE 802.15.6 stan-

dard, we propose modiﬁcations to frame acknowledg-

ment schemes so that linear NC can be included in the

communication process, thus potentially improve the

energy efﬁciency and/or throughput of the system.

Although still a relatively young ﬁeld, NC has

shown great potential in addressing issues such as net-

work throughput, robustness, and security (Fragouli

and Soljanin, 2007). Unlike traditional routing, NC

encourages nodes within a network to mix data be-

fore forwarding. Data ﬂows are viewed as informa-

tion which can be combined algebraically. As long

as enough degrees of freedom, in the form of linearly

independent coded packets or symbols, are received,

the original uncoded data can be recovered.

The concept of NC is introduced in the seminal

paper (Ahlswede et al., 2000). The authors in (Koetter

and M´edard, 2003) provide an algebraic framework

for linear network codes, while (Ho et al., 2006) intro-

duces random linear network codes (RLNC), showing

that RLNC achieves multicast capacity probabilisti-

cally in a distributed manner, exponentially in the ﬁ-

nite ﬁeld size for coding operations. Several practi-

cal approaches of NC have since been proposed to

bridge theoretical studies to real applications. To take

into account randomly generated coding coefﬁcients,

a generation-based scheme with concurrently trans-

mitted coefﬁcients is considered in (Chou et al., 2003)

to allow NC to co-exist with current network pro-

tocols. Simple XOR codes can be implemented in

two-way relay networks and four-way cross topolo-

gies by inserting an additional NC layer between the

IP and MAC layers of the 802.11 protocol (Katti et al.,

2008). Apart from such generation-based schemes, a

sliding-window approach similar to that used in the

Transmission Control Protocol (TCP) for ﬂow and

congestion control can be introduced for NC, such

that end-to-end packet losses in wireless networks us-

ing TCP can be masked (Sundararajan et al., 2009).

Moreover, the important potential of NC in pro-

viding improved efﬁciency in lossy networks in a dis-

tributed manner makes it an attractive option for en-

ergy constrained WBANs. The rateless nature of net-

work codes can lower energy consumption by reduc-

ing the amount of medium access control needed, and

by reducing energy used by nodes to wake up peri-

odically to receive timing allocations from the control

145

Shi X. and Médard M..

A Proposal for Network Coding with the IEEE 802.15.6 Standard.

DOI: 10.5220/0005245701450150

In Proceedings of the 4th International Conference on Sensor Networks (SENSORNETS-2015), pages 145-150

ISBN: 978-989-758-086-4

 2015 SCITEPRESS (Science and Technology Publications, Lda.)

base station (Shih et al., 2001). In a star network,

by adapting redundancy based on transmission and

reception energies for data and control packets, sig-

niﬁcant energy reduction may be achieved when NC

is incorporated and as the receiving energy per con-

trol packet increases from equal to much larger than

the transmitting energy per data packet (Shi et al.,

2011). Achievable gains also increase as more nodes

are added to the star network, and when the channels

seen by different nodes become more asymmetric.

In what follows, we provide a summary of the

medium access control (MAC) sublayer speciﬁca-

tions of the 802.15.6 WBAN standard, and describe

how NC can be included with minimal changes to the

MAC layer, with discussions on the potential through-

put and energy beneﬁts of using a coded system.

2 THE MAC SUBLAYER

SPECIFICATION OF THE IEEE

802.15.6 STANDARD

In a WBAN compatible with the IEEE 802.15.6 stan-

dard, there are multiple nodes but one and only one

hub. The hub coordinates medium access among dif-

ferent nodes. Nodes can be arranged into a star-

shaped topology, or into a two-hop extended star

topology, where information are exchanged between

the hub and a node via another relay-capable node.

In a WBAN, a hub may or may not provide time

referenced allocations to all nodes. When it does, the

MAC sublayer uses a slotted time reference model,

with the time axis divided into beacon periods (su-

perframes) of equal lengths, and each beacon period

divided into allocation intervals of equal length. A

frame transmission may span more than one alloca-

tion interval, with start and end times not necessarily

aligned to interval boundaries. A beacon frame may

be broadcasted by the hub at the beginning of or at a

shifted location within each beacon period.

There are three possible access modes within a

WBAN: beacon mode with beacon periods (super-

frames), non-beacon mode with superframes, and

non-beacon mode without superframes. In beacon

mode with beacon periods, a beacon frame is broad-

casted at the beginning of each beacon period to spec-

ify the length of allocation intervals within the super-

frame, the start and stop times of the access phases

present, and the number of inactive superframes to

follow the current active superframe. Four types of

access phases can be present in a beacon frame, cov-

ering contention-based random access, scheduled ac-

cess, and improvised polling/posting.

In beacon and non-beacon modes with super-

frames, scheduled access may be used to obtain up-

link and downlink allocations, and scheduled-polling

may be used to obtain bilink allocations. To obtain

one or more new scheduled allocations, a node sends a

Connection Request frame to the hub, which in return

sends a Connection Assignment frame to the node to

grant access when possible. Existing scheduled allo-

cations can be modiﬁed or terminated by the trans-

mission of a new Connection Request frame.

2.1 MAC Frame Format, Frame Types

and Subtypes

A MAC frame consists a header, a variable-length

frame body, and a cyclic redundancy check sequence.

The MACframe body has a maximum possible length

of 255 octets. The MAC header contains the sender

and receiver IDs, as well as a frame control sequence.

The contents of a MAC frame differ among dif-

ferent frame types and subtypes. In the MAC frame

header, a Frame Type ﬁeld can take on four dif-

ferent values, representing management, data, con-

trol and reserved frames. A management type frame

can be further divided into beacon, security-related,

connection-related, and reserved frame subtypes. In

beacon mode with beacon periods, a beacon frame is

multicasted by the hub at the beginning of every bea-

con period. Its frame payload contains the length of

the current beacon period, and timing allocation in-

formation such as the time duration of each alloca-

tion slot and start and stop times of each access in-

terval. Data type frames are divided into subtypes

depending on user priority levels and data subtypes.

A control type frame is used by a node or a hub to

acknowledge the recipient of a management type or

a data type frame. A node or hub receiving a control

type frame does not generate any further acknowledg-

ments. In each transmitted management type or data

type frame, an Ack Policy ﬁeld in the MAC header in-

dicates the type of acknowledgment requested by the

current frame.

2.2 Modes of Frame Acknowledgments

The Ack Policy ﬁeld of the MAC header indicates the

type of acknowledgment requested. Different values

represent four different policies.

No acknowledgment (N-Ack): an N-Ack indicates

that the current frame does not require an acknowl-

edgment from the recipient. A frame with an N-Ack

request can be transmitted by either the hub or a node.

Control type frames have their Ack Policy ﬁelds set to

N-Ack.

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

146

Group acknowledgment (G-Ack): G-Acks are ap-

plicable to data frames sent to a hub. A G-Ack is

requested if the frame is of data type, its Ack Policy

ﬁeld is set to N-Ack, and its frame subtype ﬁeld is

set to a pre-deﬁned constant. The hub acknowledges

frames with G-Ack requests from multiple nodes to-

gether through the use of a G-Ack multicast, in the

form of a frame subtype called B2 frames. A B2

frame contains a set of node IDs (NIDs) to indicate

nodes from which the hub has received a frame with

G-Ack requests since the last transmitted B2 frame.

A node may retry the frame with the G-Ack request if

it fails to receive the expected B2 frame, or if its NID

does not appear in a received B2 frame. Figure 61 of

(IEEE, 2012) gives an example of G-Acks.

Immediate acknowledgment (I-Ack): a frame with

an I-Ack request can be transmitted by either the hub

or a node. The frame is acknowledged by an I-Ack

immediately upon the frame’s reception. The recip-

ient sends back the I-Ack frame after the end of the

frame reception, after a given interframe spacing, de-

noted by pSIFS (e.g. 75µs). Management type frames

are always transmitted with an I-Ack request. Data

type frames can be transmitted with a request for any

one of the four acknowledgment modes.

Block acknowledgment later (L-Ack) and Block

acknowledgment (B-Ack): a frame with an L-Ack or a

B-Ack request can be transmitted by either the hub or

a node. The transmitted frame must contain a whole

MAC Service Data Unit (MSDU). Frames with L-

Ack requests are not followed by any I-Acks. Instead,

blocks of frames with L-Acks are acknowledged by

a B-Ack frame pSISF after the reception of the last

frame with a B-Ack request. Figure 63 of (IEEE,

2012) illustrates the block transmission scheme estab-

lished through the use of L-Acks and B-Acks. During

block transmission, the source shall send frames in

the order of non-decreasing sequence numbers, with

repetitions if retransmissions take place. Figure 63 of

(IEEE, 2012) provides two instances of retries. All

frames in a block transmission should be of the same

type and subtype. A block transmission may span

more than one allocation interval.

Three ﬁelds of a B-Ack frame function together

to provide acknowledgments to a block of data: Old-

est Frame Expected, Next Block Size, and Frame Sta-

tus Bitmap. The Oldest Frame Expected ﬁeld is not

present if there is only one frame in the current block.

When present, it is set to the sequence number of

the oldest frame that is expected but has not been re-

ceived. If all frames in the block has been received,

the Oldest Frame Field is set to one plus the sequence

number of the frame preceding the B-Ack frame.

Thus, the Oldest Frame Expected ﬁeld indicates the

next expected frame, assuming all frames are trans-

mitted in order. This is similar to acknowledgments

sent in the Transmission Control Protocol (TCP). The

Next Block Size ﬁeld is not present if only one new

frame is expected and allowed. When present, it is

set to the maximum number of frames that can be in-

cluded in the next block transmission. The one byte-

long Frame Status Bitmap ﬁeld indicates the reception

status of each successive frame in the block transmis-

sion, starting from the one immediately next to the

oldest expected frame in sequence number. Succes-

sive frames are frames with successivesequence num-

bers. Hence, acknowledgments are packet speciﬁc;

any un-acknowledged packet is retransmitted in the

next block. Moreover, since frames in a block trans-

mission are required to contain whole MSDUs, the re-

cipient may implement a timeout mechanism so that

MSDUs can be released to the upper layers without

the reception of all frames in the block.

Other than I-Ack, B-Ack, and B2 frames, I-

Ack+Poll and B-Ack+Poll control type frames can

also be transmitted by the hub to grant immediate

polled allocations or to announce a future poll or

post request, in addition to frame acknowledgments.

Moreover, a node or a hub shall treat an expected

frame, such as an I-Ack or B-Ack frame, as not ar-

riving after waiting for the physical layer preamble of

the frame for a given amount of time.

3 INCORPORATION OF

NETWORK CODES

A comparison between the acknowledgment modes

offered by the IEEE 802.15.6 standard, and the net-

work coding (NC) scheme in (Shi et al., 2011) shows

that NC can be easily incorporated into the standard.

In this subsection, we explain how this can be carried

out. The quantitative advantages of network coded

transmissions require detailed analysis of actual im-

plementations of the standard. We do not attempt to

characterize the performance of such systems here.

3.1 Generation-based Approach

NC can be carried out at individual nodes using a

generation-based approach (Chou et al., 2003). First,

the block acknowledgment mode with L-Acks/B-

Acks can be used with coded frames transmitted by

a node or the hub. At the transmitter side, the MAC

frame payload of a given number of packets, to be

transmitted in a block, can be coded together. Each

block is thus a generation. With RLNC, coding coefﬁ-

cients can be attached as part of the payload. With de-

AProposalforNetworkCodingwiththeIEEE802.15.6Standard

147

Hub

Transmits

Beacon

pMIFS

Node 1 uplink

Data

(GB-

Ack)

Nodes

Transmit

Data

(GB-

Ack)

Data

(GB-

Ack)

Data

(GB-

Ack)

Node 1 uplink

Data

(GB-

Ack)

Data

(GB-

Ack)

Node 2 uplink

Data

(GB-

Ack)

Data

(GB-

Ack)

Data

(GB-

Ack)

Node 1 uplink

allocation interval

Burst Mode

Node 1 uplink

allocation interval

Node 2 uplink

allocation interval

Figure 1: Example of coded data blocks with Group-Block Acknowledgments (GB-Ack).

terministic codes, coding coefﬁcients are pre-deﬁned.

In the MAC frame control header of the transmitted

frame, a reserved bit can indicate whether the current

packet is coded. With coding, redundant packets can

be added to compensate for potential losses. As dis-

cussed in (Shi et al., 2011), a reasonable level of re-

dundancy can be computed heuristically. Packet loss

rates can be estimated and redundancy levels can be

adjusted accordingly. A systematic network code can

be used to minimize coding energy. All coded frames

except the last one carry an L-Ack request to the re-

ceiver, while the last coded frame carries a B-Ack re-

quest to the receiver.

At the receiver side, assuming well-chosen coef-

ﬁcients, each arriving packet is a unique degree of

freedom (DOF), innovative to the receiver with high

probability. Whenever a B-Ack request is received,

the receiver can use the Frame Status Bitmap ﬁeld to

acknowledge to the transmitter the number of DOFs

received, or the number of DOFs still expected. The

Oldest Frame Expected ﬁeld can be modiﬁed to indi-

cate the generation number. When uncoded, only 9

frames can be acknowledged in each block (indicated

by the Oldest Frame Expected ﬁeld, plus 8 bits in the

Frame Status Bitmap). With coded data, more than 9

frames can be included in a single block.

Moreover, B-Ack and G-Ack modes can be com-

bined into a Group-Block acknowledgment(GB-Ack)

mode. Figure 1 shows a possible use of GB-Acks

in scheduled access mode, for data blocks uploaded

from two nodes to a hub. At each node, data frames in

a block are coded with random or deterministic coefﬁ-

cients. In the MAC frame control header of the trans-

mitted frame, a reserved bit can be used to indicate if

the current packet is coded. At the hub, a B2 frame is

broadcasted at a pre-determined allocation slot when

G-Acks are requested by multiple nodes. A B2 frame

contains a set of NIDs, each representing a node to be

group acknowledged. Since there can be a maximum

of 64 nodes within a WBAN, each with a single-octet

NID, but the B2 frame payload can be of 255 octets

long, it is possible to append to the current B2 frame

payload the numbers of DOFs successfully received

from each node for its respective current generation.

For example, a one-octet ﬁeld can be added to the B2

frame payload for each node from which the hub has

received coded frames in the current block. The ﬁrst

4 bits can represent the generation number, while the

last 4 bits can represent the number of DOFs received.

There are several possible advantages of using

coded transmissions with B-Acks or GB-Acks. First,

independently coded packets are individual DOFs,

thus acknowledgments do not need to specify which

individual packet has been received successfully. In

addition, with coded payloads, more than 9 frames

can be included in a block transmission.

From a throughput perspective, in cases where

an acknowledgment is lost and retransmission occurs

without the loss of the actual data frame, a retrans-

mitted data frame is wasteful if uncoded. By com-

parison, if the retransmitted frame is coded with a set

of coefﬁcients linearly independent from the previous

frame, the retransmission is innovative, carrying one

more DOF to the hub. In addition, in burst mode block

transmissions, interframe spacing between successive

frames is smaller than the I-Ack case. Thus with less

frequent transmissions, the overall throughput of in-

dividual nodes can be higher. Figure 2 shows an sam-

ple timing diagram of I-Ack, B-Ack (coded or un-

coded) and GB-Ack schemes. Here Guard Intervals

(GT) are centrally added in between allocation inter-

vals instead of distributively included in each alloca-

tion interval. Although we do not provide detailed

computations here, it is not difﬁcult to show that the

number of data frames transmitted in an allocation in-

terval with GB-Acks can be higher than the number of

data frames transmitted in the same allocation interval

with I-Acks or B-Acks, although if the allocation in-

terval is very short, with high probability, the same

number of frames ﬁts into the interval regardless if B-

Acks or GB-Acks is used. It is also not hard to see

that if the number of nodes in the WBAN is large,

using GB-Ack is beneﬁcial in terms of reducing the

total number of acknowledgment frames transmitted

in one superframe.

From an energy perspective, given a ﬁxed num-

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

148

Hub

pSIFS

Hub

pMIFS

pSIFS

Hub

Transmits

Data

(I-

Ack)

Ack

Nodes

Transmit

Data

(I-

Ack)

Ack

Hub

Transmits

Nodes

Transmit

Data

(L-

Ack

)

Data

(L-

Ack

)

Data

(B-

Ack

)

-Ack

•••

Node 1 uplink

allocation interval

pSIFS

Ack

)

Ack

)

Ack

)

Node 1 uplink

allocation interval

Node 2 uplink

allocation interval

•••

Node 2 uplink

allocation interval

•••

pSIFS

pMIFS

(a) (b)

Hub

Transmits

pSIFS

pMIFS

Node 1 uplink

allocation interval

Burst Mode

Data

(GB-

Ack)

Nodes

Transmit

Data

(GB-

Ack)

Data

(GB-

Ack)

Data

(GB-

Ack)

Node 2 uplink

allocation interval

Data

(GB-

Ack)

Data

(GB-

Ack)

Burst Mode

(c)

Figure 2: Timing examples (a) uncoded transmission with I-Acks; (b) coded data transmission with B-Acks and L-Acks; and

ber of data frames to transmit by a node, block trans-

mission requires the node to be active for a shorter

period of time than when I-Acks are used. Such en-

ergy savings become more signiﬁcant when switching

between transmission and reception modes take up a

non-negligibleamount of power. Moreover, with cod-

ing, it is always possible to transmit more redundan-

cies when listening to acknowledgment packets costs

more energy. Nonetheless, coding requires additional

energy, and a tradeoff exists between encoding energy

and transmission/reception energy.

From a reliability perspective, coded transmis-

sions allow redundant coded frames to be transmit-

ted a priori to compensate for possible losses. NC

thus provides another reliability mechanism, allowing

less stringent requirements on PHY design. In cases

where channel quality is low and scheduled alloca-

tions for a node is aborted because none of the trans-

mitted frames is receivedsuccessfully at the hub, such

provision for losses may prolong the duration of the

connection between a node to the hub.

3.2 Sliding Window Approach

An alternative to the generation-based NC scheme is

to use a sliding window approach, as proposed by

(Sundararajan et al., 2009) for interfacing NC with

TCP/IP. At the transmitter, data frame payloads are

put into a coding window of a ﬁxed size as they ar-

rive from the upper layer, as long as the coding win-

dow is not full. A frame is removed from the coding

window when it is acknowledged by the receiver. At

the receiver, coding coefﬁcients from received pack-

ets are put into a matrix and passed through Gaus-

sian elimination into a reduced row echelon form.

Frames on the pivot columns are taken as seen, and

are acknowledged in the next available acknowledg-

ment frame. In short, a receiver is said to have seen

a packet p

if it can compute a linear combination of

the form p

∑

l>k

, where α

are coding coefﬁ-

cients, and k represents frame sequence number. It

can be shown that if all frames involved in the coding

operations have been seen, all coded packets can be

decoded. Seen packets are thus dropped from the cod-

ing window at the transmitter side. The notion of seen

packets deﬁnes an ordering for the DOF conveyed by

coded packets: for every additional DOF received at

the hub, the next data frame in the coding window is

acknowledged and dropped from subsequent coding

operations.

The sliding window approach can be used in a

WBAN similar to the generation-based approach. At

the transmitter, data frame payloads are coded and

sent with redundancies. At the receiver, either B-Acks

or GB-Acks are used. The Oldest Expected Frame

ﬁeld of a B-Ack frame can be modiﬁed to represent

the oldest expected seen frame. Frames which have

been seen are subsequently removed from the coding

AProposalforNetworkCodingwiththeIEEE802.15.6Standard

149

window at the transmitter. For GB-Ack frames, an

Oldest Expected Seen Frame ﬁeld can be added to in-

dicate frames to be removed from the coding window

of each transmitting node. A systematic code can also

be used, and the transmission scheme can operate in a

rateless fashion if needed.

An advantage of the sliding window approach is to

allow new data to be added to the transmission queue

incrementally as they arrive at the MAC sublayer and

removed as soon as they are seen at the receiver. A

pre-deﬁned generation size is not required, although

the size of the coding window can vary, depending on

the amount of computation and memory available at

sensor nodes. When data rate is not very high, fast re-

moval of data packets from the coding window is ben-

eﬁcial, since fewer coding operations are performed

for subsequent transmission rounds. Another advan-

tage of the sliding window approach is that its retrans-

mission requests are very short, since it only needs

to acknowledge the last seen DOF. A disadvantage

of the sliding window approach is that even though

memory and energy are assumed to be abundant at

the hub, decoding delay is larger. Seen packets may

have to stay in the receiver queue for longer periods

than the generation-based case, before they can be de-

coded. This problem can be mitigated if the transmit-

ter ﬂushes its coding window periodically.

4 CONCLUSIONS

In this article, we provide a brief description of the

medium access control sublayer of the IEEE 802.15.6

WBAN standard, and consider the incorporation of

network coding (NC) in WBANs compatible with the

standard. Both generation-based and sliding window

approaches are possible, and a group-block acknowl-

edgment can be implemented by modifying block ac-

knowledgmentcontrol type frames. With NC, retrans-

mission requests are simpler, and redundant packets

can be transmitted a priori, in expectation of packet

losses. Such redundancies can be seen as a reliabil-

ity mechanism in addition to forward error correc-

tion schemes in the physical layer. Allowing more

coded packets to be sent in each round also allows

the overall transmission process to terminate in fewer

rounds, leading to potential savings in energy use.

Since the standard only provides general rules and

guidelines, actual implementations can differ in terms

of hardware and signal processing algorithm design.

Whether coding helps in terms of energy or through-

put thus depends on how the standard is realized.

ACKNOWLEDGEMENT

This work was supported by the Interconnect Focus

Center (IFC), one of the six research centers funded

under the Focus Center Research Program (FCRP),

a Semiconductor Research Corporation (SRC) pro-

gram, under subcontract # RA306-S 1.

REFERENCES

Ahlswede, R., Cai, N., Li, S., and Yeung, R. (2000). Net-

work information ﬂow. IEEE Trans. on Info. Theory,

46(4):1204–1216.

Chou, P., Wu, Y., and Jain, K. (2003). Practical network

coding. In Proceedings of the Annual Allerton Con-

ference on Communication, Control and Computing,

volume 41, pages 40–49. The University; 1998.

Fragouli, C. and Soljanin, E. (2007). Network coding fun-

damentals. Foundations and Trends

 in Networking,

2(1):1–133.

Ho, T., M´edard, M., Koetter, R., Karger, D., Effros, M., Shi,

J., and Leong, B. (2006). A random linear network

coding approach to multicast. IEEE Trans. on Info.

Theory, 52(10):4413–4430.

IEEE (2012). IEEE Standard for Local and metropolitan

area networks - Part 15.6: Wireless Body Area Net-

works.

Katti, S., Rahul, H., Hu, W., Katabi, D., M´edard, M., and

Crowcroft, J. (2008). XORs in the air: practical wire-

less network coding. IEEE/ACM Trans. on Network-

ing (TON), 16(3):497–510.

Koetter, R. and M´edard, M. (2003). An algebraic approach

to network coding. IEEE/ACM Trans. on Networking

(TON), 11(5):782–795.

Shi, X., M´edard, M., and Lucani, D. (2011). When both

transmitting and receiving energies matter: an appli-

cation of network coding in WBANs. In NETWORK-

ING 2011 Workshops, pages 119–128. Springer.

Shih, E., Cho, S., Ickes, N., Min, R., Sinha, A., Wang, A.,

and Chandrakasan, A. (2001). Physical layer driven

protocol and algorithm design for energy-efﬁcient

wireless sensor networks. In Proceedings of the 7th

annual international conference on Mobile comput-

ing and networking, pages 272–287. ACM New York,

NY, USA.

Sundararajan, J., Shah, D., M´edard, M., Mitzenmacher, M.,

and Barros, J. (2009). Network coding meets TCP. In

Proceedings of IEEE INFOCOM, pages 280–288.

Tachtatzis, C., Di Franco, F., Tracey, D., Timmons, N.,

and Morrison, J. (2010). An energy analysis of IEEE

802.15.6 scheduled access modes. In GLOBECOM

Workshops, 2010 IEEE, pages 1270–1275. IEEE.

Ullah, S. and Kwak, K. (2011). Throughput and delay limits

of IEEE 802.15.6. In WCNC, 2011 IEEE, pages 174–

178. IEEE.

SENSORNETS2015-4thInternationalConferenceonSensorNetworks

150