1516 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 16, NO.

6, JUNE 2017

AT-MAC: Adaptive MAC-Frame Payload Tuning

for Reliable Communication in Wireless
Body Area Networks
Soumen Moulik, Student Member, IEEE, Sudip Misra, Senior Member, IEEE, and Debayan Das

Abstract—In wireless sensor networks, adaptive tuning of Medium Access Control (MAC) parameters is necessary in order to assure
the QoS requirements. In this paper, we propose an adaptive MAC-frame payload tuning mechanism for wireless body area networks
(WBANs) to maximize the probability of successful packet delivery or reliability of the associated sensor nodes based on real-time
situation. The enabling algorithm, Adaptively Tuned MAC (AT-MAC), has been proposed to tune the MAC-frame payload of a WBAN
sensor node, which is compliant with the IEEE 802.15.4 protocol. AT-MAC prioritizes sensor nodes based on the seriousness of the
health parameters that are being measured by the respective sensor nodes. Further, we consider a Markov chain-based analytical
approach that acknowledges the slotted CSMA/CA backoff mechanism with retry limits, as described in the IEEE 802.15.4 protocol.
We derive expressions for reliability, power consumption, and throughput, which are the key metrics to evaluate the network
performance of the proposed protocol, and analyze the impact of MAC parameters on them. Finally, results indicate that the low rate
and low power IEEE 802.15.4 can be used effectively in case of WBANs if the payload is tuned properly through the proposed
algorithm. The proposed AT-MAC algorithm yields around 70 percent increase in reliability of a critical node in a WBAN.

Index Terms—Wireless body area network, IEEE 802.15.4 standard, Markov chain model, adaptive payload tuning, critical sensor node,
criticality index

Ç
1 INTRODUCTION

A WBAN, as the name suggests, is a network of wireless

and wearable computing entities that sense and trans-
mit the measurements of the physiological parameters of a
of the physiological system malfunctions, is more crucial
than that of the traditional wireless sensor networks
(WSNs), as it concerns human health monitoring. Thus,
patient [1]. WBANs find applications in diverse domains maximizing reliability of a sensor node is an important con-
such as physiological and medical monitoring, and human- cern in WBANs, in order to increase the chance of successful
computer interaction [2]. The last decade had witnessed a reception of packets to the destination node.
dramatic increase in the number of such wearable comput- The motivation behind the proposed work is to increase
ing devices. Examples include wearable heart-rate monitors, the rate of successful packet delivery of the sensors, so
glucose-monitors, blood-oxygen saturation (SpO2 ) monitors, that it can be used in any WBAN application. In this work,
accelerometers, ECG sensors, and medical implants [3]. we use the concept of Criticality Index (CI) to identify the
most critical physiological parameter and the concerned
1.1 Motivation sensor node. The sensor node with maximum CI is termed
as the critical node, at a particular time. Furthermore, we
Most of the modern era e-Health applications, which are
propose an algorithm—AT-MAC—to tune the MAC-frame
being used for pervasive and ubiquitous healthcare of
payload, which is the length of data in the transmitted
patients, are based on WBANs [4], [5], [6]. Data (or packet)
frame, in order to maximize the probability of successful
loss due to collisions and network problems such as non-
packet delivery (reliability) of the critical node. Conse-
idle channel, and channel errors are the primary reliability- quently, we achieve that the information from the most
centric limitations in WBANs. Moreover, in case of WBANs, critical node reaches the destination node or the Local
data loss by any physiological sensor node, when some part Processing Unit (LPU) with the least possible delay and
maximum reliability.

S. Moulik is with the School of Medical Science and Technology, Indian
Institute of Technology Kharagpur, Kharagpur, WB 721302, India. 1.2 Contribution
E-mail: soumen.moulik@smst.iitkgp.ernet.in.

S. Misra is with the School of Information Technology, Indian Institute of The specific contributions of this work are as follows:
Technology Kharagpur, Kharagpur, WB 721302, India.
E-mail: smisra@sit.iitkgp.ernet.in.
In order to quantify the severity of a WBAN sensor

D. Das is with the Department of ETCE, Jadavpur University, Kolkata, node, we use the concept of CI.
WB 700032, India. E-mail: ddebayan50@gmail.com.
The Markov chain-based analytical model proposed
Manuscript received 7 Aug. 2015; revised 26 May 2016; accepted 20 July in this work achieves much better reliable communi-
2016. Date of publication 3 Aug. 2016; date of current version 3 May 2017. cation with less delay and less power consumption.
For information on obtaining reprints of this article, please send e-mail to:
reprints@ieee.org, and reference the Digital Object Identifier below.
We optimize the MAC-frame payload of the critical
Digital Object Identifier no. 10.1109/TMC.2016.2598166 node in order to maximize the reliability of that node.
1536-1233 ß 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
MOULIK ET AL.: AT-MAC: ADAPTIVE MAC-FRAME PAYLOAD TUNING FOR RELIABLE COMMUNICATION IN WIRELESS BODY AREA... 1517

2 RELATED WORK 2.3 Application-Based Works

Among the application-based works, the ones by Rodrigues
In this section, we categorically discuss the pros and cons of
et al. [20] and Pereira et al. [21] are prominent. Rodrigues
the existing literature that are focused on the IEEE 802.15.4
et al. [20] addressed the issue of processing of raw biosensor
protocol [7]. We categorize the works based on their nature,
data to achieve appropriate and medically relevant visuali-
and analyze their pros and cons.
zation of monitored physiological parameters. The authors
setup a platform for health-monitoring, and it runs on
2.1 Simulation-Based Works phones that have Java and Bluetooth enabled. Extending
In spite of the deficiency of MAC payload tuning-based works the previous approach, Pereira et al. [21] presented a
for IEEE 802.15.4, there have been several simulation-based mobile-based biofeedback monitoring system that operates
works [8], [9] to study the delay, throughput, and power con- on major smart-phone platforms such as Symbian, Android,
sumption of this protocol. Koubaa et al. [8] simulated the per- Windows Mobile, and iPhone. Another interesting work by
formance of slotted CSMA/CA for diverse settings of Anjum et al. [22] proposed a priority-based load aware
network to analyze the influence of protocol components MAC protocol for body sensor networks. In this work, the
such as beacon order, superframe order, and backoff expo- data packets are served based on some priority that
nent on network performance. Zheng and Lee [9] further con- depends on the data-type and the generation rate. However,
sider several other issues such as association through tree these works lack tuning of MAC parameters, specifically the
formation, coordinator relocation, and guaranteed time-slot tuning of frame payload, which is necessary to optimize
allocation in their simulation-based study. On the other hand, QoS attributes such as reliability of the critical-most sensor
Vishnevsky et al. [10] studied the problems of beacon colli- node in a WBAN. Moulik et al. [23] and Misra et al. [24] con-
sions in case of simultaneous joining of more than one devices sidered the tuning of payload and data-rate of the sensor
to a piconet. Pang et al. [11] and Ko et al. [12] proposed TCP- nodes associated with a WBAN in their works, using deci-
like window adjustment mechanisms for IEEE 802.15.4 to sion process and bargaining solution, respectively. How-
maximize network throughput. The algorithms correspond- ever, the tuning of payload and data-rate in these works are
ing to these works analyze attributes such as the channel sens- not sufficiently adaptive, and the detailed analysis of reli-
ing state, successful packet transmission, and packet collision ability, delay, throughput, and other network parameters
to modify the contention window size based on the analysis. are not addressed in these works.
A fair backoff algorithm is also proposed by Fang et al. [13]. Synthesis: Most of the existing studies were either con-
However, the authors did not consider any health specific ducted in contexts other than IEEE 802.15.4-based WBANs,
parameters in their works. Further, physiological severity or overlooked the importance of reliable packet transmis-
measured by the individual sensors is not considered in these sion as a significant aspect of QoS. We address this lacuna
works. The primary focus of some of these works [11], [12], in this paper, and show how we can exploit the low data
[13] are maximizing network throughput. Though, in case of rate and the low power of IEEE 802.15.4 for reliable data
health monitoring applications, maximum throughput cannot transmission in WBANs. In this paper, we present AT-
guarantee the reliability of packet transmission. MAC, an adaptively tuned MAC protocol for IEEE 802.15.4,
to maximize the probability of successful data packet deliv-
2.2 Analytical Works ery (reliability) of a ‘critical’ node at a certain time instant,
by tuning the MAC-frame payload. From hardware imple-
Inspired by Bianchi’s work [14], different authors contrib-
mentation perspective payload tuning can be achieved by
uted by developing analytical models for IEEE 802.15.4.
varying the sensing time of physiological data, which is
Misic et al. [15] proposed a three-dimensional Markov chain
much easier to implement in comparison with the MAC
considering a non-saturated network. The authors consid-
parameters considered by Park et al. in [18], [19]. The pro-
ered M/G/1/K queues to acknowledge the uplink trans-
posed adaptive MAC improves the reliability of a critical
mission. In a similar approach, Sahoo and Sheu [16] derived
WBAN sensor node, while guaranteeing less collision prob-
an analytical model for IEEE 802.15.4 CSMA/CA based on a
ability, power efficiency and delay constraints of the entire
three-dimensional Markov chain considering the packet
system.
retry limits. Pollin et al. [17] proposed a two-dimensional
Markov chain and evaluate both uplink and acknowledged
traffic. However, none of these studies realized the necessity 3 FRAMEWORK AND PROBLEM SCENARIO
of tuning protocol parameters in order to achieve better net- In this section, we discuss the overall framework and the
work performance. Park et al. [18] proposed a three-dimen- problem scenario of the proposed model in the context of
sional Markov chain model for IEEE 802.15.4 protocol, ubiquitous health monitoring [25]. An end-user is equipped
while taking into account reliability, energy consumption, with a WBAN setup that contains different physiological
and delay. In another work, Park et al. [19] analyzed reli- sensors (such as heart rate sensor, accelerometer, pulse
ability, delay and power consumption for the IEEE 802.15.4 oximeter etc.), and one LPU associated with them. These
protocol, and optimized these performance metrics. How- sensors sense physiological parameters, and convey the
ever, the works do not consider the severity of health measurements to the LPU, as illustrated in Fig. 1. In such
parameters which are monitored by specific physiological systems, it should be ensured that at every turn, the sensor
sensor nodes. Moreover, these works are mostly analytical node that shows the maximum abnormality in sensed phys-
in nature, and from the hardware implementation perspec- iological data, i.e., the critical body sensor node must be
tive, the direct tuning of the considered MAC parameters in detected by the LPU in order to improve it’s own transmis-
these works is difficult to achieve. sion reliability through MAC-frame payload optimization.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
1518 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 16, NO. 6, JUNE 2017

Fig. 1. Overall framework of the proposed AT-MAC algorithm.

However, apart from the payload optimization there is no However, a case always may exist that re-transmission of
difference among the critical and the non-critical sensor a packet fails continuously. In such cases, the proposed AT-
nodes. They all follow the same protocol for periodically MAC algorithm considers the approach stated in the IEEE
transmitting physiological data. If one of the nodes which 802.15.4 protocol itself [7], i.e., re-transmitting the packet
was previously a ‘non-critical’ node, exhibits the maximum only a fixed number of times, which is defined by the
health severity, the LPU considers it as the ‘critical’ node in parameter re-transmission counter ðrÞ in this paper. If packet
the current turn, and optimize it’s payload. Thus, critical transmission still fails then AT-MAC drops the current
and non-critical notions associated with the sensor nodes packet and chooses a new packet from the buffer, even if
are not static, rather changes with time, depending on the the current packet holds information that is sensed from a
severity of the sensed physiological data. critical node. It won’t be a problem as the next packet,
At a particular instant of time, it may be the case that the which contains fresh updated information, is picked up
sensed data have different range of medical severity. There- immediately. This whole process of transmission, re-trans-
fore, the default MAC-frame payload of these sensors may mission, dropping of the current packet, and selection of the
not always guarantee optimal satisfaction of QoS attributes, next packet for fresh transmission is very fast, approxi-
such as reliability of data transmission. For the sake of mately in a scale of fraction of seconds. Thus, it is highly
receiving critical health information reliably it is necessary unlikely to misjudge a critical node as non-critical, even if
to grant a privilege to the physiological sensor node that the patient faces irregular heart rate, or any such similar
senses the most critical data with respect to the other sen- health abnormality, as frequent ups and downs in physio-
sors at that particular time instant. Fig. 1 briefly elaborates logical severity is assumed not to occur within fraction of
all the major steps involved in the whole operation. In this seconds.
work, the proposed AT-MAC algorithm identifies the most
critical node at a particular time instant, and then based on 4 ANALYTICAL MODEL
the severity value of the measured physiological parameter, In the proposed work, we consider a WBAN with N physio-
i.e., the CI of that node, the system selects a priority mode logical sensor nodes that contend to send physiological data
for it. Three priority modes are envisioned in this work, viz. to the LPU. The Markov model analysis proposed in this
High Priority, Medium Priority, and Normal Mode. Each prior- work, is purely based on the principles of slotted CSMA/
ity mode has its own payload range, which is optimized in CA algorithm discussed in IEEE 802.15.4 protocol [7].
order to maximize its data transmission reliability. This pay- Thus the ideas of Clear Channel Assessments (CCAs),
load range also depends on the minimum reliability acknowledged transmission, macMaxCSMABackoffs (p), and
requirement of that mode, i.e., RhighPmin or RmedP normP
min or Rmin . macMaxFrameRetries (r) are used in Markov chain, as it is
Furthermore, the proposed AT-MAC algorithm handles discussed in the protocol.
packet re-transmissions, until the retry limit reaches, Let a1 and a2 be the probability of finding the channel
through payload tuning within the resultant payload range. busy during CCA1 due to data and ACK transmissions,
All associated calculations such as selecting the maximum respectively. Hence, the total probability of a channel to be
CI among a pool of CIs, assigning priority mode based on busy during CCA1 is given by: a ¼ a1 þ a2 . Let b denotes
the value of maximum CI, and optimization of MAC-frame the probability of finding the channel busy during CCA2 .
payload of the critical node, are done by the LPU. We consider the default values of macMaxCSMABackoffs (p)

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
MOULIK ET AL.: AT-MAC: ADAPTIVE MAC-FRAME PAYLOAD TUNING FOR RELIABLE COMMUNICATION IN WIRELESS BODY AREA... 1519

and macMaxFrameRetries (r) as 4 and 3, respectively, as pre-

scribed in the IEEE 802.15.4 protocol [7]. This model holds
good for other values of r and p too, and it is verified in
Section 6. The slot lengths Ls and Lc represent the slot
length for successful packet delivery and the slot length for
packet collision, respectively. They are defined as, Ls ¼
Lp þ LSIFS þ LACK , and Lc ¼ Lp þ LSIFS þ LmACK , where Lp
is the slot length of the transmitted packet, LSIFS is the slot
length of inter-frame spacing after a packet is transmitted,
LACK is the slot length of the acknowledgement frame
received by the sensor node, and LmACK is the
macAckWaitDuration in terms of slot length. The different
terminologies considered in the proposed analytical model,
are defined as follows.
Definition 1 (Collision Probability). The collision probabil- Fig. 2. Variation of a and b with MAC frame payload (Lp ).
ity (Pcoll ) of a sensor node is the probability that the data packet
transmitted from that node encounters a collision with another
data packet or ACK packet transmitted by some other node of The failure probability is primarily affected by channel
the network at the same time slot. error (Pec ) and collision, and is given as follows:

Proposition 1. If t is the probability that a node attempts the Pf ¼ 1  ð1  Pec Þð1  Pcoll Þ: (4)
first carrier sensing CCA1 in a random time slot, then the colli-
sion probability of the sensor node is given as In the simulation of the proposed model, we consider the
probability of the channel error (Pec ) as a function of
Pcoll ¼ 1  ð1  tÞN1  ða þ ð1  aÞbÞN1 : received signal strength indicator (RSSI), modulation, and
channel coding. RSSI depends on the path-loss model, shad-
Proof. Consider one sensor node (N1 ) among N sensor owing standard deviation, and Gaussian fading model. We
nodes, for which we derive the expression for collision also incorporate the packet error rate for the corresponding
probability. Now, a denotes the busy channel probabil- modulation scheme in our simulation. Interference due to
ity during CCA1 , and b is the busy channel probability multiple body sensor nodes is considered in the shadowing
during CCA2 . Hence, ð1  aÞb is the probability of and fading model.
finding the channel busy during CCA2 , given that the
channel is idle during CCA1 . Hence, the probability Definition 3 (Reliability). The reliability (R) of a sensor node
that the channel is busy for all other sensor nodes, is defined as the probability of successful packet delivery by that
except N1 , is given by node [18].
The expression of reliability of a sensor node is given by
Pidle ¼ ða þ ð1  aÞbÞN1 : (1)
R ¼ 1  ðPdcf [ Pdrl Þ ¼ 1  Pdcf  Pdrl ; (5)
We have, ð1  tÞ, the probability that a node does not
perform carrier sensing. The probability that all ðN  1Þ where, Pdcf is the probability that the packet is discarded
nodes, other than N1 , do not attempt carrier sensing is due to channel access failure, and Pdrl is the packet drop
ð1  tÞN1 . Hence, for the sensor node N1 , the probability probability due to retry limits.
that there is no chance of collision, i.e., other ðN  1Þ
nodes are not sensing or they find the channel busy dur- Definition 4 (Payload). The payload (L) of the transmitted
ing CCA is given as data frame by a sensor node is defined as the length of data or
message in the transmitted frame.
Pcoll ¼ ð1  tÞN1 þ ða þ ð1  aÞbÞN1 : (2) From the simulation results obtained, we can express a
and b in terms of the MAC frame payload as shown below.
Therefore, for the sensor node N1 , the collision probabil-
ity is given as Lemma 1. The expressions of the busy channel probabilities a
and b can be approximated as linear functions of the MAC
Pcoll ¼ 1  ð1  tÞN1  ða þ ð1  aÞbÞN1 : (3) frame payload (Lp ).
u
t Proof. We study the variation of a and b for different values
of MAC-frame payload. From Fig. 2, it is evident that the
Park et al. [18] considered t as the carrier sensing proba-
variation of a and b with the payload (L) is almost linear.
bility as well as the transmission probability. However, in
For the sake of convenience, we interpolate the set of
practice, this may not be the case.
obtained values using polynomial fitting. We find that, in
Definition 2 (Failure Probability). The failure probability is the expressions of a and b, the coefficients of the higher
the conditional probability that a packet is not received by a sen- order terms are negligible in value. The coefficient of the
sor node, given that it was transmitted successfully from squared term in the expressions of a and b is of the order
another node. of 105 . Hence, only the linear term along with a constant

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
1520 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 16, NO. 6, JUNE 2017

is considered for the sake of our calculations in this work. Definition 6 (Critical Node). A Critical Node (sensor) in a
Therefore, the expressions of a and b are given as WBAN is defined as that node, which has the maximum CI at a
certain time slot.
a ¼ c1 L þ c2 (6)
In this paper, we ensure that the critical node is able to
b ¼ c3 L þ c4 ; (7) send data with maximum reliability, at any instant of time.
where c1 lies in the range (0.0029, 0.0033), and c2 lies in
the range (0.3051, 0.3428) so that the overall frame pay- 4.1 AT-MAC: The Proposed Algorithm
load is less than 127 bytes, as specified in the IEEE The proposed algorithm, AT-MAC, tunes the MAC-frame
802.15.4 standard [7]. Again, maintaining the above con- payload adaptively, as the name suggests. Algorithm 1 gov-
straint, the range of c3 is (0.0015, 0.0019) and that of c4 is erns the whole procedure of AT-MAC by sequentially call-
(0.1120, 0.1542). u
t ing the functions—FindCritical, SetPriority, and ACKSensing.

Corollary 1. Let us consider that the payload assigned to a critical Algorithm 1. AT-MAC Algorithm
node, corresponding to maximum reliability, at an instant be L0 ,
and let Lp be the payload assigned to all other sensor nodes. There- Input: Sensed physiological data from each sensor node.
Output: Assignment of payload that leads to maximum
fore, using Lemma 1, the normalized equations for a and b are
reliability for the critical node.

0 
L N 1 1: In a WBAN, N number of nodes communicate with the
a ¼ c1 þ Lp þ c2 (8) LPU at fixed MAC parameters.
N N
2: critical node = FindCriticalðNÞ

0  3: Mode = SetPriorityðCI½CriticalNodeÞ
L N 1
b ¼ c3 þ Lp þ c4 : (9) 4: ACKSensingðModeÞ
N N

Definition 5 (Criticality Index). The Criticality Index of the Algorithm 2 identifies the critical body sensor node in the
sth sensor is the measure of seriousness of the health parameter, WBAN. The function FindCritical returns the CI of that criti-
which is being measured by that physiological sensor at the cal node whose payload scheme is to be optimized, and the
time instant t [23]. function SetPriority described in Algorithm 3, assigns a par-
In crisp set theory, we interpret the severity of a health ticular priority mode for data transmission to that node,
parameter as ‘low’, ‘moderate’, and ‘high’, but cannot quan- depending on its CI. We envision three priority modes: high
tify the deviation, i.e., how much it is low, moderate, or high priority, medium priority, and normal mode. After selecting
from the expected measure. Thus, from one recent existing the priority mode, the most critical sensor node starts data
work [23], we incorporate the concept of fuzzy inference in transmission using the slotted CSMA/CA algorithm. The
order to achieve a justified formulation of CI. Mamdani model payload range varies between modes, and is set depending
[26], the most used fuzzy inference technique is used to derive on the reliability requirement of each mode—such as RhighPmin ,
the criticality of the physiological parameters. RmedP normP
min , and Rmin . These additional constraints represent
The CI considers the criticality of data collected through the minimum reliability requirement for the high priority,
different fuzzy sets and membership functions. In the fuzzy medium priority, and normal mode, respectively. This
rules, along with the internal factors (human physiological threshold can be varied in application level, depending on
parameters), external influences such as age that has signifi- the requirement, and accordingly the MAC-frame payload
cant effect on health parameters, are considered. Linguistic is tuned. The function ACKSensing finally assigns the pay-
fuzzy sets such as ‘LOW’, ‘MODERATE’, and ‘HIGH’ are load, as explained in Algorithm 4. This algorithm is
considered to categorize the severity of each physiological repeated till successful transmission occurs, or the retry
parameter. On the other hand, the external variable, human limit r becomes zero; whichever is earlier. For the case,
age, is represented through the fuzzy sets—‘YOUNG’, r ¼ 0, the packet is dropped. However, it is highly undesir-
‘ADULT’, and ‘AGED’. Next, some rules are considered able and should not occur for a high or medium priority
such as “if temperature = LOW and age = HIGH, then criti- critical node. Hence, the payload is tuned in order to ensure
cality = STAGE 3”. The Mamdani model fuzzifies the input highest reliability. The values of Lmax and Lmin are derived
variable combinations (such as the combination of tempera- from Markov chain analysis of the data transmission on the
ture and age), evaluates the rules, aggregates the rule out- critical node.
puts, and defuzzifies to get a crisp value, which is
considered as the CI of that particular physiological sensor. 4.2 Markov Chain Model for Critical Node
The value of CI strictly ranges between 0 and 1. The mathe- In this paper, Markov chain has been used for analytical
matical expression of the CI is given as follows: modeling. Exploiting this model we derive the expression
R1 of reliability, which, in turn, helps to derive the range of
mR;t ðcÞ:c dc payload for maximum reliability of the critical node.
CI ¼ R0 1 ; (10)
Using the Markov model we efficiently consider different
0 mR;t ðcÞ dc
steps of the slotted CSMA/CA algorithm, and deduce
where mR;t ðcÞ denotes the aggregated output of all the fuzzy corresponding expressions for power consumption, and net-
rules that are stored in a rule-base. The details of this mech- work parameters such as—throughput and delay, for fur-
anism are available in the original published article [23]. ther analysis.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
MOULIK ET AL.: AT-MAC: ADAPTIVE MAC-FRAME PAYLOAD TUNING FOR RELIABLE COMMUNICATION IN WIRELESS BODY AREA... 1521

Algorithm 2. Algorithm for Determining the Critical We propose a Markov chain model considering the MAC
Node PIB attributes macMaxCSMABackoffs ¼ p and
macMaxFrameRetries ¼ r for each packet transmitted by
Input: Sensed physiological data from N sensor nodes. the critical node, which follows slotted CSMA/CA algo-
Output: The sensor node that senses critical physiological data rithm and acknowledged transmission. We design the
at that time instant.
Markov model with three variables sðtÞ, lðtÞ and rðtÞ repre-
1: function FindCriticalðNÞ
senting the backoff stage, payload of the transmitted data
2: temp = 0
frame, and the value of re-transmission counter at time t,
3: for all i = 1 to N do
respectively, as illustrated in Fig. 3. The transition to any of
4: Compute CI as described in [23]
5: if CI[i] > temp then the three priority mode considered in the Markov model, is
6: temp = CI[i] a stochastic process. Hence, we assign P1 , P2 , and P3 as the
7: CriticalNode = i transition probabilities respectively for high priority,
8: end if medium priority, and normal mode data transmission, such
9: i iþ1 that P1 þ P2 þ P3 ¼ 1. We consider t as the stationary inde-
10: end for pendent probability that a node performs carrier sensing in
11: return(CI[CriticalNode]) a randomly chosen time slot. The tuple (sðtÞ, lðtÞ, rðtÞ) repre-
12: end function sents each state in the proposed three-dimensional Markov
chain. We do not include the Markov model states for the
normal mode data transmission as it is the trivial case and
Algorithm 3. Algorithm for Assigning the Priority Mode payload tuning is not necessary for this case. We consider
Input: CI of the critical node. ð0; 0; 0Þ as the idle state, when the node does not have any
Output: The suitable priority mode for the critical node. packet to transmit or receive, and waits for new packet
1: function SetPriorityðCI½CriticalNodeÞ arrivals. In every scheme, the node, initially, performs Clear
2: if CI[CriticalNode] > 0:75 then Channel Assessment (CCA1 and CCA2 ), and if the channel
3: Assign Mode High Priority is idle, data transmission starts. The states ði; 0; 0Þ and
4: Set minimum reliability requirement as RhighP
min ði; Lmax ; 0Þ denote CCA1 and CCA2 , respectively. The suc-
5: Assign L0 Lmax , r 3 cessful and failed packet delivery are represented by the
6: else states ðps ; 0; 0Þ and ð0; 0; rÞ, respectively. The stationary dis-
7: if CI[critical-node] > 0:5 then tribution of the proposed Markov chain model is given by
8: Assign Mode Medium Priority
9: Set minimum reliability requirement as RmedP
min Si;j;k ¼ lim P ðsðtÞ ¼ i; lðtÞ ¼ j; rðtÞ ¼ kÞ;
10: Assign L0 Lmax , r 3 t!1
(11)
11: else 8i 2 ½0; p; j 2 ½Lmin ; Lmax ; k 2 ½0; r:
12: Assign Mode Normal Mode
13: Set minimum reliability requirement as RnormP The state transition probabilities in Fig. 3 are
min
14: Assign L0 ¼ Lp , r 3
15: end if P ðp; 0; 0 j 0; 0; 0Þ ¼ Px t; 8x 2 ½1; 3 (12)
16: end if
17: return(Mode) P ði; Lmax ; 0 j i; 0; 0Þ ¼ 1  a; 8i 2 ½1; p (13)
18: end function
P ði; Lmax ; r j i; Lmax ; 0Þ ¼ 1  b; 8i 2 ½1; p (14)

Algorithm 4. Algorithm for Sensing the Acknowledge- X

k1
ment Corresponding to a Transmitted Packet P ðps ; 0; r j i; j; kÞ ¼ ð1  Pcoll Þ x
Pcoll ¼ 1  Pcoll
k
;
x¼0 (15)
Input: The operating mode of the critical node.
Output: No output. It handles retry counter depending on the 8i 2 ½1; p; j 2 ½Lmin ; Lmax ; k 2 ½1; r
status of packet transmission.
1: function ACKSensing(Mode) P ði; j2 ; k  1 j i; j1 ; kÞ ¼ Pcoll ;
2: if success then (16)
8i 2 ½1; p; k 2 ½2; r; j1 ; j2 2 ½Lmin ; Lmax 
3: Critical sensor node goes to idle state.
4: else
5: while r 6¼ 0 do P ð0; 0; r j i; j; kÞ ¼ Pcoll
k
;
(17)
6: if r = 1 then 8i 2 ½1; p; j 2 ½Lmin ; Lmax ; k 2 ½1; r
7: L0 Lmin
8: else
9: r r1 P ð0; 0; 0 j i; j; kÞ ¼ 1;
(18)
10: L0 RandðL0  1; Lmin þ 1Þ 8i 2 ½1; p; j 2 ½Lmin ; Lmax ; k 2 ½1; r
11: end if
12: end while P ði  1; 0; 0 j i; Lmax ; 0Þ ¼ b; 8i 2 ½2; p (19)
13: end if
14: end function P ði  1; 0; 0 j i; 0; 0Þ ¼ a þ ð1  aÞb; 8i 2 ½1; p (20)

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
1522 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 16, NO. 6, JUNE 2017

Fig. 3. Markov chain model for the IEEE 802.15.4-based AT-MAC algorithm.

P ð0; 0; r j 1; 0; 0Þ ¼ a (21)
Proposition 2. The idle state S0;0;0 is expressed as


P ð0; 0; r j 1; Lmax ; 0Þ ¼ b: (22) S0;0;0 ¼ ða þ ð1  aÞbÞp þ ð1  ða þ ð1  aÞbÞp Þ



1
1 1 1
Eq. (12) derives the probability that the sensor node ð1  Pcoll
r
Þ 1þ þ Pcoll
r
þ 1þ :
1  Pcoll 1b 1a
selects either of the three modes, as decided by the LPU.
(24)
Eq. (13) shows the probability of transition to the CCA2 state
from CCA1 . Eq. (14) denotes the probability that the sensor
node finds the channel idle in CCA2 and goes to the next Proof. The proof is given in the Appendix. u
t
state ði; Lmax ; rÞ, where maximum payload is assigned to
the transmission data frame of the critical sensor node, 4.3 Maximization of Reliability
depending on the priority mode. When the CI of the critical In a WBAN, we must ensure the successful data transmis-
node is less than 0.5, we choose normal mode of data trans- sion from the critical body sensor node and reception at the
mission, and set the maximum payload to Lp , a fixed opti- LPU, such that the reliability is maximum and collision
mal value. Eq. (16) denotes the decrement of retransmission probability, delay, and failure probability of the critical
counter due to collision of transmitted data packet. The size node is minimum. Hence, we develop a mathematical
of payload is set randomly between the previous payload model for tuning the payload of the critical node, in order to
value and the minimum payload for that scheme, as repre- find the range of payload for which the reliability of that
sented below node is maximum. The critical node in the WBAN is deter-
mined using the CI, as mentioned earlier in this section.
L0 ¼ RandðL0  1; Lmin þ 1Þ: (23) From Definition 3, reliability is given as R ¼ 1  Pdcf 
Pdrl . In this expression, Pdcf is the probability of a packet
Eq. (17) derives the probability of packet failure after being discarded due to channel access failure. Channel
sensing the channel as idle in both CCA1 and CCA2 due access failure can occur if either of CCA1 or CCA2 is unsuc-
to r successive packet collisions. Eq. (18) derives the cessful. Hence, the overall probability of packet drop owing
probability of reaching the idle state from the transmit- to channel access failure is given by
ting state, which is unity. Eq. (19) denotes the probability
of finding the channel busy in CCA2 , and of selecting the X
r1
next state in the backoff stage of CCA1 . Eq. (20) repre- Pdcf ¼ ða þ ð1  aÞbÞp ½Pcoll ð1  ða þ ð1  aÞbÞp Þx
sents the transition from one backoff state to another x¼0 (25)
backoff state. Eqs. (21) and (22) denote the total probabil- ’p ð1  r Þ
¼ :
ity of packet drop due to channel access failure in CCA1 ð1  Þ
and in CCA2 , respectively.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
MOULIK ET AL.: AT-MAC: ADAPTIVE MAC-FRAME PAYLOAD TUNING FOR RELIABLE COMMUNICATION IN WIRELESS BODY AREA... 1523

On the other hand, Pdrl is the probability of packet drop due 4.4 Energy Consumption
to packet collision after finding the channel idle, that is due Total energy dissipation by a sensor node is derived
to retry limits. Hence through the consideration of different states of a body sen-
sor node. When the node performs CCA, the total energy
Pdrl ¼ Pcoll
r
ð1  ða þ ð1  aÞbÞp Þr ¼ r : (26) consumption due to sensing is given as
Using Eqs. (25) and (26), we get the expression of reliability as X
p X
p
Esc;tot ¼ Esc Si;0;0 þ Esc Si;Lmax ;0 : (34)
’p ð1  r Þ i¼1 i¼1
R¼1  r
ð1  Þ (27) The first part of the above expression accounts for CCA1 and
¼ 1  ’p ð1 þ  þ 2 þ . . . þ r1 Þ  r ; the latter part corresponds to CCA2 . The energy consump-
tion when the sensor node is in the idle state is given as
where,
X
p LX
max X
r
’ ¼ a þ ð1  aÞb (28) Ei;tot ¼Ei ð1  Pcoll ÞðLs  Lp  LACK Þ Si;j;k Þ
i¼1 j¼Lmin k¼1
 ¼ Pcoll ð1  ’ Þ:p
(29)
X
p LX
max X
r
0
Let L be the payload of the critical sensor node, for which þ Ei Pcoll ðLc  Lp Þ Si;j;k :
i¼1 j¼Lmin k¼1
reliability of the critical sensor node is maximum. Partially
differentiating Eq. (27) with respect to L0 , we have (35)
! ! The above equation sums up the energy consumption after
@R Xr
@ Xr
@’
¼  r þ ’
r1 p
i i1
 p’p1
 i
: successful packet delivery and the energy consumption due
@L 0 @L 0 @L0
i¼1 i¼0 to packet drop owing to collision.
(30) In case of successful packet transmission, the sensor node
uses Lp slot length for transmitting the data packet and con-
Again, using Eq. (28), we obtain sumes Etx Lp amount of energy in each state. Therefore,
@’ @a @b X
p LX
max X
r
¼ ð1  bÞ 0 þ ð1  aÞ 0 : (31)
@L0 @L @L Etx;tot ¼ Etx Lp Si;j;k : (36)
i¼1 j¼Lmin k¼1
From Eqs. (3) and (29), we have
Similarly, Erx LACK amount of energy is spent in each state
@ @’ to receive the ACK frame. Thus,
¼  ð1  ð1  tÞN1 ð1  aÞð1  bÞÞp’p1 0
@L0
@L
@’ X
p LX
max X
r
þ ð1  ’p Þð1  tÞN2 ð1  tÞ 0 (32) Erx;tot ¼ Erx LACK ð1  Pcoll Þ Si;j;k : (37)
@L
 i¼1 j¼Lmin k¼1
@t
 ð1  aÞð1  bÞðN  1Þ 0 :
@L Adding the Eqs. (34), (35), (36), and (37), we get the required
expression for energy consumption as follows:
As, the carrier sensing probability t is independent of pay-
load of the critical node L0 , we conclude @L @t
0 ¼ 0. Again, Etot ¼ Esc;tot þ Ei;tot þ Etx;tot þ Erx;tot
@a @b


using Eqs. (8) and (9) we have, @L0 ¼ N , and @L0 ¼ cN3 . There-
c1
p 1 1
fore, using Eqs. (31) and (32), we obtain the expression for ¼ Px tð1  ða þ ð1  aÞbÞ Þ Esc 1þ
1b 1a
@R 0
@L0 as a function of payload (L ) of the critical sensor node. 1  Pcoll
r
For maximum reliability of the critical body sensor node þ Ei ðð1  Pcoll ÞðLs  Lp  LACK Þ
1  Pcoll

@R þ Pcoll ðLc  Lp Þ þ Etx Lp þ Erx LACK ð1  Pcoll ÞÞ ;
¼ 0: (33)
@L0
(38)
Solving Eq. (33), we obtain the range of payload for which
reliability of the critical sensor node is maximum, and it sat- where Ei , Etx , Erx , and Esc correspond to the average power
isfies the minimum reliability requirement of the selected consumption in the idle state, transmitting state, receiving
priority mode. Eq. (33) results in multiple solutions, satisfy- state, and sensing state, respectively.
ing the desired reliability criteria of different priority
modes. The minimum payload value for a particular mode 4.5 Throughput and Delay
may be obtained from the minimum value of these solu- Throughput is defined as the ratio of product of the slot
tions. Let us consider a WBAN with 20 sensor nodes, that is length equivalent to the delay due to successful transmission
N ¼ 20. Putting N ¼ 20 in Eq. (33), and ensuring that the of a packet ðLs Þ and packet size ðLp Þ, to the total delay in
@2 R
second order derivative is negative, i.e., @L 02 < 0, we have, terms of slot length ðLd Þ. The total slot length required to
L0min ¼ 366:25 and L0max ¼ 377:5. Hence, we need to tune the deliver a packet successfully depends on the probability of
payload of the critical sensor node in this range for maxi- successful delivery, and the slot length required to determine
mum reliability of the critical node. the successful packet delivery. The delay depends on the

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
1524 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 16, NO. 6, JUNE 2017

TABLE 1 and communicates using the 2.4 GHz ISM band. Simulation
Parameters for Simulation results are generated using Monte Carlo simulation where
the values of busy channel probabilities (a; b) are randomly
Parameters Values
generated. Using these values, we obtain the idle state S0;0;0 .
Maximum CSMA Backoffs 4 Accordingly, we calculate the throughput, reliability, colli-
Maximum Retry Limits 3
sion probability and other performance metrics. The system
Maximum Length of Payload 127 bytes
macMinBE 3 parameters used in our simulation are tabulated in Table 1.
macMaxBE 5 In this section, we compare the variation of reliability, colli-
Multiplicative constant to convert time 80 bits/slot sion probability, failure probability, power consumption,
length of frame to slot length average delay, and throughput against the offered payload
Data Rate 250 kbps corresponding to the General IEEE System [7] and the pro-
Symbol Rate 62.5 kbps posed AT-MAC system. General IEEE system refers to the
Length of ACK frame 88 bits
systems that follow IEEE 802.15.4 protocol but do not tune
MAC-frame payload in order to achieve maximum reliabil-
busy channel probability and the collision probability. Thus, ity, and also do not consider the severity of physiological
the probability of successful packet delivery is given as data during data transmission. The plots for the proposed
AT-MAC system are separately shown in terms of both
Ps Ls ¼ Px tð1  Pcoll
r
Þð1  ða þ ð1  aÞbÞp ÞLs ; (39) non-critical nodes and critical nodes. It is observed that the
plots corresponding to general IEEE systems and the non-
where Ps is the probability of successful packet delivery.
critical nodes in the AT-MAC system vary negligibly. The
Similarly, the delay is given as
reason is, our proposed AT-MAC algorithm always focuses
on the critical node at a certain time instant and tune it’s
Pd Ld ¼P1 tð1  ða þ ð1  aÞbÞp Þðð1  Pcoll
r
ÞLs þ Pcoll
r1
Lc Þ
MAC-frame payload.
þ ð1  tÞL0 ;
(40) 5.1 Simulations for Reliability
Fig. 4 illustrates the variation of reliability against the
where Pd denotes the probability that the packet is neither
offered load (frame arrival rate). For a WBAN system hav-
dropped nor transmitted. We define throughput as follows:
ing 20 sensor nodes, we compute the value of payload for
Ps Ls maximum reliability by ensuring that the reliability must
s¼ Lp : (41) not be less than the minimum reliability requirement for the
Pd Ld
high priority mode. The range of values of payload to
Hence, replacing Eqs. (39) and (40) in Eq. (41), we get the ensure the maximum reliability is obtained in the range
expression for throughput as (366.25, 377.5) in case of 20 sensor nodes, as mentioned ear-
lier in Section 4.3. Fig. 4a shows the variation of reliability
Px tð1  Pcoll
r
Þð1  ða þ ð1  aÞbÞp ÞLs Lp with the offered load, while payload assigned to the critical
s¼
Px tð1  ða þ ð1  aÞbÞp Þðð1  Pcoll
r ÞL Þ þ P r1 L Þ
s coll c sensor node (L0 ) = 366.25 bits and the other nodes in the net-
work are assigned payload as 800 bits. Solving Eq. (33), for
þ ð1  tÞL0 : (42) 30 nodes, we get the range of payload of the critical node as
(347.9, 359.1), and similarly for 40 nodes, L0 ranges between
5 PERFORMANCE EVALUATION 338.75 and 349.8 bits. The corresponding offered loads and
In this work we use MATLAB as the simulation tool along their effects on the reliability are isillustrated in Figs. 4b
with a physical layer modeling in it. For modeling the phys- and 4c, for 30 and 40 nodes, respectively.
ical layer in our code, we incorporate the noise figure (23 dB Inference: From Fig. 4, it is evident that the reliability of
for our device) and bandwidth, as per the IEEE 802.15.4 the critical node is enhanced significantly, compared to the
standard. Again, our system being low power enabled and general IEEE systems. Similarly the critical node also exhib-
having low sensor-LPU distance, we consider the path loss its better reliability than the other non-critical nodes in the
exponent to be 4. Simulations are performed for an intra- AT-MAC system. We see that as the offered load increases,
WBAN that follows star topology of IEEE 802.15.4 standard, the reliability decreases gradually, due to the channel access

Fig. 4. Variation of reliability versus offered load.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
MOULIK ET AL.: AT-MAC: ADAPTIVE MAC-FRAME PAYLOAD TUNING FOR RELIABLE COMMUNICATION IN WIRELESS BODY AREA... 1525

Fig. 5. Variation of collision probability versus offered load.

failure and buffer overflow. We also observe that as the The collision probability of the critical node is much less
number of nodes increases, the maximum reliability of the compared to the rest, and the rate of increment with the
critical sensor node decreases, as the collision probability, increase in number of nodes in the system is comparatively
and thus, the failure probability increases. From the results, small. It is because of the judicious choice of the payload of
it is evident that the proposed AT-MAC algorithm yields the critical node by the proposed AT-MAC algorithm.
average 70 percent improvement on the reliability of the
critical node. 5.3 Simulations for Failure Probability
As defined in Section 4, Definition 4, the failure probability
5.2 Simulations for Collision Probability of a sensor node is considered as a function of channel error
Fig. 5 shows the variation of collision probability with the and collision probability. Thus, similar to the results for col-
offered load, for 20, 30, and 40 sensor nodes. We see that the lision probability, the failure probability for the critical node
collision probability is much reduced for the critical node in is much reduced, compared to that of the general IEEE sys-
AT-MAC system, compared to the general IEEE system. tem, and the non-critical nodes of the proposed AT-MAC
Collision probability for the other non-critical nodes in the system, as illustrated in Fig. 6. We observe that the failure
AT-MAC system is approximately same with the general probability of the critical node in AT-MAC system lies
IEEE system, as the MAC-frame payloads of these sensor within 0:2-0:3. It varies from 0.4 to 0.9 in case of the general
nodes are not optimized. The proposed AT-MAC algorithm IEEE system, and the non-critical nodes in the AT-MAC
is able to bound the collision probability of the critical node system.
within a very little value (such as 0.1, in case of analyzing Inference: The chances of failure in end-to-end communica-
with maximum 40 sensor nodes). tion is more than the chances of collision between source and
Inference: Negligible collision probability exerts a positive destination, as failure also includes the factor of channel error.
effect to increase the reliability of a node, and it is already Comparison between Figs. 5 and 6 further validate the fact.
verified in Fig. 4. For a fixed number of sensor nodes, as the We also observe that, as the number of nodes increases, the
offered load increases, i.e., when the incoming frame arrival failure probability increases. Though the rate of increment is
rate is high, the transmission rate also increases, and hence, much less in case of the critical node in AT-MAC system. Due
the collision probability increases. However, the rate of to the judicious choice of the critical node and efficient optimi-
increment is much lower for the critical node in the pro- zation of its MAC-frame payload guided by the proposed AT-
posed AT-MAC system. In case of general IEEE systems, MAC algorithm, the failure probability of the critical node
the rate of increment of collision probability is approxi- reduces to the one-third of it’s previous value.
mately 6 times larger than that of the critical node in AT-
MAC systems. Again, as the number of nodes increases, the 5.4 Simulations for Power Consumption
collision probability increases for general IEEE systems and Fig. 7 depicts the variation of power consumption (in mW)
the proposed AT-MAC system. However, for the critical with the offered load. We observe that the power consump-
node the increment is negligible and mainly carried out by tion of the non-critical nodes in AT-MAC system is almost
the non-critical nodes in the system. same with the general IEEE standard. The critical sensor

Fig. 6. Variation of failure probability versus offered load.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
1526 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 16, NO. 6, JUNE 2017

Fig. 7. Variation of power consumption versus offered load.

Fig. 8. Variation of average delay versus offered load.

Fig. 9. Variation of throughput versus offered load.

node in the AT-MAC system consumes more power, com- the critical node of AT-MAC system decreases initially upto
pared to the non-critical nodes and the general IEEE system. a certain offered load value, and beyond that it remains
Again, as the number of nodes increases, the power con- almost stable.
sumption of every individual sensor node decreases. Inference: Average delay primarily depends on the chan-
Inference: The approximate increment in the power con- nel access failure probability, which in turn is directly pro-
sumption of the critical node lies within 12-16 percent. portional to the packet arrival rate. Thus, normally the
However, with this little more expense in energy the critical delay increases with the increase of offered load. However,
node exhibits low collision and failure probability, and in case of the AT-MAC critical node we already observe in
thus, achieves maximum possible reliability. It is appropri- Fig. 6 that the rate of increase of failure probability with the
ate to consider this expense as a trade-off between between increase of offered load is almost negligible. Thus, the aver-
energy and reliability. Moreover, when the number of sen- age delay witnesses a different behavior in case of the criti-
sors increase, the probability of finding channel idle for cal node.
data transmission is less. As a consequence, the energy con-
sumption is less, as the share of transmission energy is max- 5.6 Simulations for Throughput
imum in the total amount of energy consumption. Fig. 9 depicts the variation of throughput with the offered
load. As the offered load increases, throughput for both the
5.5 Simulations for Average Delay general IEEE system and AT-MAC increases. However, in
The critical node in the proposed AT-MAC system attains case of general IEEE systems the throughput increase till it
significant achievement in reducing end-to-end delay. Fig. 8 reaches the congestion area. Again, with more number of
illustrates that the average delay increases with the incre- sensors the system throughput gets affected, especially in
ment in the offered load, in case of the general IEEE systems case of general IEEE systems. In all cases the proposed AT-
along with the non-critical nodes of AT-MAC. However, we MAC system yields better throughput than the general
observe an interesting phenomena that the average delay of IEEE systems.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
MOULIK ET AL.: AT-MAC: ADAPTIVE MAC-FRAME PAYLOAD TUNING FOR RELIABLE COMMUNICATION IN WIRELESS BODY AREA... 1527

decreases. Likewise, higher the value of macMaxFrame

Retries or r, the source get more chance to retry packet
transmission, which failed in previous attempts. Either of
these circumstances increase the reliability of packet deliv-
ery at the destination node. However, it is evident from
Fig. 10, that the prposed AT-MAC model is more reliable
than the Park’s model, especially when the critical node
uses its optimized payload range.
Apart from the better reliability achievement, the pro-
posed AT-MAC model also yields slightly less delay, and
less power consumption than the Park’s model, as illus-
Fig. 10. Reliability comparison. trated in Fig. 11. Though, the average throughput of the
Park’s system is trivially higher than the proposed AT-
Inference: The involvement of more sensor nodes increases MAC system. However, the minor less throughput of the
the failure probability, and the probabilities of finding chan- proposed AT-MAC system is considerable as the benefits
nel busy. Thus, it also affects the system throughput. How- we get in return are manyfold, including its efficient identi-
ever, in case of AT-MAC system it does not affect much, and fication of the critical node and maximizing its data trans-
the reason is the optimized MAC-frame payload tuning of mission reliability significantly.
the critical node. The aim of the proposed algorithm is to We also compare the proposed solution with the solution
maximize reliability with satisfying the minimum reliability provided by Rodrigues et al. [20]. The latter work primarily
requirement of the selected mode, as discussed earlier. It focuses on mobile application development in order to setup
takes care of the number of sensors within the optimization, completely functional health-monitoring platforms that can
and achieves maximum reliability. Thus, we observe that be accessed via different smart-phones. On the contrary, AT-
there is not much effect of the number of sensors on the MAC primarily emphasizes on the theoretical modeling of
related attributes such as—the reliability, collision probabil- IEEE 802.15.4 with the help of the three-dimensional Markov
ity, failure probability of the critical node, and as well as on chain to optimize QoS parameters. However, still we com-
the average throughput of the AT-MAC system, in which the pare the different analytical attributes that are common
critical node with its optimized payload becomes an impor- between the above mentioned works and the proposed AT-
tant contributor. MAC algorithm. Rodrigues et al. [20] considered the LPU as
a mobile device, whereas, we use a DELL Inspiron N4050 lap-
top with Intel Core i5 2nd gen processor with 4 GB DDR3
6 PERFORMANCE COMPARISON
RAM as the LPU. In our simulations, the connection time or
In this section, we compare the performance of the pro- the delay time is obtained in the range of 30-50 ms. The proc-
posed AT-MAC algorithm with one of the most recent essing time lies in the range of 7-10 ms. The connection time
works of Park et al. [19], as both of these works deal with a indicates the time required to fetch and transfer information
common goal to increase the reliability of communication. from body sensors to the LPU. The processing time is an indi-
The proposed AT-MAC algorithm yields much better reli- cator of the time required for conversion of the packet data
ability for the critical node, with respect to the model pro- into useful health information within the LPU. The amount
posed by Park et al. Even the average reliability of the of corrupted data in our simulation is observed to be 4.31 per-
proposed AT-MAC system is slightly higher than the aver- cent, on an average. Corrupted data indicates the amount of
age reliability of the Park’s model. Fig. 10 depicts the com- data corrupted due to inherent channel errors, and is calcu-
parative analysis of the effects of different macMaxCSMA lated as the difference between the probability of failure and
Backoffs and macMaxFrameRetries on reliability of packet the probability of collision. Hence, comparing the proposed
delivery. According to the slotted CSMA/CA algorithm, for work with the work of Rodirigues et al. [20], we observe that
high value of macMaxCSMABackoffs or m, the source gets the amount of corrupted data is nearly equal, whereas, the
more opportunity to sense the channel as idle, and thus the connection and processing times are much less in case of the
probability of discarding a packet due to busy channel proposed system.

Fig. 11. Comparison with Park’s model with 20 sensor nodes.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
1528 IEEE TRANSACTIONS ON MOBILE COMPUTING, VOL. 16, NO. 6, JUNE 2017

7 CONCLUSION X
p X
p1
Si;Lmax ;0 ¼ Px tð1  aÞ ða þ ð1  aÞbÞx
In this paper, we present an adaptively-tuned MAC algorithm i¼1 x¼0
for enhancing the reliability of a critical sensor node, by opti- 1  ða þ ð1  aÞbÞp
mizing the MAC-frame payload of the critical node. Based on ¼ Px t : (48)
1b
the AT-MAC algorithm, we propose a Markov chain-based
analytical model for slotted CSMA/CA of IEEE 802.15.4 with
retry limits and backoffs. Numerical results show that the pro- By adding Eqs. (44), (45), (46), (47), and (48), and equat-
posed scheme is efficient and ensures maximum reliability of ing them to 1 using Eq. (43), we get the desired expres-
the critical sensor node. In case of medical emergency, the crit- sion for S0;0;0 . &
ical sensor node in a WBAN-assisted health monitoring sys-
tem, exploits the benefits of the proposed AT-MAC system REFERENCES
with less failure probability and delay. [1] D. Tobon, T. Falk, and M. Maier, “Context awareness in WBANs:
In the future, we wish to extend this work with a multi- A survey on medical and non-medical applications,” IEEE Wire-
less Commun., vol. 20, no. 4, pp. 30–37, Aug. 2013.
objective optimization approach, where we plan to maxi- [2] B. Liu, Z. Yan, and C. W. Chen, “MAC protocol in wireless body
mize reliability and minimize energy consumption of a criti- area networks for E-health: Challenges and a context-aware
cal node simultaneously. On the other hand, we performed design,” IEEE Wireless Commun., vol. 20, no. 4, pp. 64–72, Aug.
the Markov chain formulation and analysis for a saturated 2013.
[3] X. Liu, Y. Zheng, M. W. Phyu, F. Endru, V. Navaneethan, and
network. However, in case of an unsaturated network the B. Zhao, “An ultra-low power ECG acquisition and monitoring
analysis is different. Therefore, in the future, we also wish ASIC system for WBAN applications,” IEEE J. Emerg. Sel. Topics
to extend this work for unsaturated networks. Circuits Syst., vol. 2, no. 1, pp. 60–70, Mar. 2012.
[4] M. Chen, S. Gonzalez, A. Vasilakos, H. Cao, and V. C. M. Leung,
“Body area networks: A survey,” Mobile Netw. Appl., vol. 16,
APPENDIX pp. 171–193, Apr. 2011.
[5] C.-S. Lin and P.-J. Chuang, “Energy-efficient two-hop extension
Proof of Proposition 2. From the proposed Markov chain protocol for wireless body area networks,” IET Wireless Sensor
model, by the normalization condition, we have Syst., vol. 3, pp. 37–56, Mar. 2013.
[6] E. Reusens, et al., “Characterization of on-body communication

p
1 X LX max X X
r p channel and energy efficient topology design for wireless body
Si;j;k þ Sps ;0;r þ Si;0;0 þ S0;0;r area networks,” IEEE Trans. Inf. Technol. Biomedicine, vol. 13, no. 6,
Px t i¼1 j¼L k¼1 i¼1
pp. 933–945, Nov. 2009.
min
 (43) [7] IEEE Standard for Local and Metropolitan Area Networks-Part 15.4:
Xp
Low-Rate Wireless Personal AreaNetworks (LR-WPANs)-Amendment
þ Si;Lmax ;0 ¼ 1: 6: TV White Space Between 54 MHz and 862 MHz Physical Layer,
i¼1 IEEE Std 802.15.4m-2014, pp. 1–118, Apr. 2014.
[8] A. Koubaa, M. Alves, and E. Tovar, “A comprehensive simulation
Using Eqs. (12), (13), (14), (15), (16), (17), (18), (19), (20), study of slotted CSMA/CA for IEEE 802.15.4 wireless sensor
networks,” in Proc. IEEE Int. Workshop Factory Commun. Syst.,
(21), and (22) and owing to chain regularities, we obtain 2006, pp. 183–192.
[9] J. Zheng and M. J. Lee, “A comprehensive performance study of
X
p LX
max X
r IEEE 802.15.4,” Sensor Network Operations. Hoboken, NJ, USA:
Si;j;k Wiley, 2006, pp. 218–237.
i¼1 j¼Lmin k¼1 [10] V. Vishnevsky, A. Lyakhov, A. Safonov, S. Mo, and A. Gelman,
“Study of beaconing in multihop wireless PAN with distributed
X
r1 X
p1
control,” IEEE Trans. Mobile Comput., vol. 7, no. 1, pp. 113–126,
¼ Px tð1  aÞð1  bÞ Pcoll ða þ ð1  aÞbÞu S0;0;0 Jan. 2008.
x¼0 u¼0 [11] A. C. Pang and H. W. Tseng, “Dynamic backoff for wireless
1  Pcoll
r
personal networks,” in Proc. IEEE Global Telecommun. Conf.,
¼ Px t ð1  ða þ ð1  aÞbÞp ÞS0;0;0 Nov./Dec. 2004, pp. 1580–1584.
1  Pcoll
[12] J. G. Ko, Y. H. Cho, and H. Kim, “Performance evaluation of IEEE
(44) 802.15.4 MAC with different backoff ranges in wireless sensor
networks,” in Proc. IEEE Int. Conf. Commun. Syst., Oct. 2006, pp. 1–5.
[13] Z. Fang, B. Bensaou, and Y. Wang, “Performance evaluation of a
Sps ;0;r ¼ Px tð1  Pcoll Þð1  aÞð1  bÞ
fair backoff algorithm for IEEE 802.11 DFWMAC,” in Proc. 3rd
X
r1 X
p1 ACM Int. Symp. Mobile Ad Hoc Netw. Comput., Jun. 2002, pp. 48–57.
u
Pcoll ða þ ð1  aÞbÞx S0;0;0 (45) [14] G. Bianchi, “Performance analysis of the IEEE 802.11 distributed
u¼0 x¼0 coordination function,” IEEE J. Sel. Areas Commun., vol. 18, no. 3,
pp. 535–547, Mar. 2000.
¼ Px tð1  Pcoll
r
Þð1  ða þ ð1  aÞbÞp ÞS0;0;0 [15] J. Misic, V. Misic, and S. Shafi, “Performance of IEEE 802.15.4 bea-
con enabled PAN with uplink transmissions in non-saturation
X
p X
p1 mode-access delay for finite buffers,” in Proc. 1st Int. Conf. Broad-
Si;0;0 ¼ Px t ða þ ð1  aÞbÞx S0;0;0 band Netw., Oct. 2004, pp. 416–425.
[16] P. K. Sahoo and J.-P. Sheu, “Modeling IEEE 802.15.4 based wire-
i¼1 x¼0 (46)

 less sensor network with packet retry limits,” in Proc. 5th ACM
1  ða þ ð1  aÞbÞp Symp. Performance Evaluation Wireless Ad Hoc Sensor Ubiquitous
¼ Px t S0;0;0 Netw., 2008, pp. 63–70.
ð1  aÞð1  bÞ
[17] S. Pollin, et al., “Performance analysis of slotted carrier sense IEEE
802.15.4 medium access layer,” IEEE Trans. Wireless Commun.,
X
p1
vol. 7, no. 9, pp. 3359–3371, Sep. 2008.
S0;0;r ¼ Px tðð1  aÞð1  bÞPcoll
r
ða þ ð1  aÞbÞx [18] P. Park, C. Fischione, and K. H. Johansson, “Adaptive IEEE
x¼0 (47) 802.15.4 protocol for energy efficient, reliable and timely
þ ða þ ð1  aÞbÞp ÞS0;0;0 communications,” in Proc. 9th ACM/IEEE Int. Conf. Inf. Process.
Sensor Netw., Apr. 2010, pp. 327–338.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.
MOULIK ET AL.: AT-MAC: ADAPTIVE MAC-FRAME PAYLOAD TUNING FOR RELIABLE COMMUNICATION IN WIRELESS BODY AREA... 1529

[19] P. Park, P. Marco, C. Fischione, and K. Johansson, “Modeling and Sudip Misra (SM’11) received the PhD degree in
optimization of the IEEE 802.15.4 protocol for reliable and timely computer science from Carleton University,
communications,” IEEE Trans. Parallel Distrib. Syst., vol. 24, no. 3, Ottawa, Canada. He is currently an associate pro-
pp. 550–564, Mar. 2013. fessor in the School of Information Technology,
[20] J. J. Rodrigues, O. R. Pereira, and P. A. Neves, “Biofeedback data Indian Institute of Technology Kharagpur. Prior to
visualization for body sensor networks,” J. Netw. Comput. Appl., this, he was associated with Cornell University,
vol. 34, no. 1, pp. 151–158, 2011. Yale University, Nortel Networks (Canada), and
[21] O. Pereira, J. M. Caldeira, and J. J. Rodrigues, “Body sensor net- the Government of Ontario (Canada). He has been
work mobile solutions for biofeedback monitoring,” Mobile Netw. a recipient of eight research paper awards in differ-
Appl., vol. 16, no. 6, pp. 713–732, 2011. ent conferences. He was awarded the IEEE Com-
[22] I. Anjum, N. Alam, M. A. Razzaque, M. M. Hassan, and Soc Asia Pacific Outstanding Young Researcher
A. Alamri, “Traffic priority and load adaptive MAC protocol for Award at IEEE GLOBECOM 2012. He was also awarded the Canadian
QOS provisioning in body sensor networks,” Int. J. Distrib. Sensor Governments prestigious NSERC post-doctoral fellowship and the Hum-
Netw., vol. 9, no. 1, Feb. 2013, Art. no. 205192. boldt research fellowship in Germany. He is a senior member of the IEEE.
[23] S. Moulik, S. Misra, C. Chakraborty, and M. Obaidat, “Prioritized
payload tuning mechanism for wireless body area network-based
healthcare systems,” in Proc. IEEE Global Commun. Conf., Debayan Das received the BE degree in elec-
Dec. 2014, pp. 2393–2398. tronics and telecommunication engineering from
[24] S. Misra, S. Moulik, and H. C. Chao, “A cooperative bargaining Jadavpur University, West Bengal, India, in 2015.
solution for priority-based data-rate tuning in a wireless body After graduation, he worked with xSi Semicon-
area network,” IEEE Trans. Wireless Commun., vol. 14, no. 5, ductors, India, as an analog design engineer. He
pp. 2769–2777, May 2015. will be joining the SPARC Lab, Purdue University,
[25] H. Cao, V. Leung, C. Chow, and H. Chan, “Enabling technologies in August 2016 as a doctoral student. His
for wireless body area networks: A survey and outlook,” IEEE research interests include mixed signal circuit
Commun. Mag., vol. 47, no. 12, pp. 84–93, Dec. 2009. design and wireless networks.
[26] E. Mamdani and S. Assilian, “An experiment in linguistic synthe-
sis with a fuzzy logic controller,” Int. J. Man-Mach. Studies,
vol. MMS-7, pp. 1–13, Jan. 1975.
" For more information on this or any other computing topic,
Soumen Moulik (S’14) received the BTech please visit our Digital Library at www.computer.org/publications/dlib.
degree in computer science and engineering from
the West Bengal University of Technology, Kolkata,
India, in 2010. He is presently working toward the
PhD degree in the School of Medical Science
and Technology, Indian Institute of Technology
Kharagpur, India. His current research interests
include ubiquitous healthcare systems, networking
and communication aspects of wireless body area
networks, and its association with cloud and big
data. He is a student member of the IEEE.

Authorized licensed use limited to: R V College of Engineering. Downloaded on July 04,2020 at 08:42:06 UTC from IEEE Xplore. Restrictions apply.