Energy Adaptive MAC Protocol for Wireless Sensor

Networks with RF Energy Transfer

Jaeho Kim Jaeho Kim, Jang-Won Lee
U-embedded Convergence Research Center, Dept. of Electrical & Electronic Eng.,
Korea Electronics Technology Institute, Yonsei University,
Gyeonggi-do, Korea Seoul, Korea
Email: [email protected] Email: {jaehokim, jangwon}@yonsei.ac.kr

Abstract—Recently, various energy harvesting techniques from harvested from ambient environments is unreliable and un-
ambient environments were proposed as alternative methods controllable, since environmental energy is typically space-
for powering sensor nodes, which convert the ambient energy dependent and time-varying. In addition, energy harvesting
from environments into electricity to power sensor nodes. How-
ever, those techniques are not applicable to the wireless sensor techniques are not applicable to the WSN in the environment
networks (WSNs) in the environment with no ambient energy with no ambient energy source. To overcome these problems,
source. To overcome this problem, an RF energy transfer method RF energy transfer methods are proposed to power wireless
was proposed to power wireless sensor nodes. However, the sensor nodes [3]–[5]. The RF energy transfer system consists
RF energy transfer method also has a problem of unfairness of an energy emitting node and energy harvesting nodes
among sensor nodes due to the significant difference between
their energy harvesting rates according to their positions. In located within the tens of meters. The RF energy transfer
this paper, we propose a medium access control (MAC) protocol system has good characteristics for the WSN in terms of
for WSNs based on RF energy transfer. The proposed MAC reliability, controllability, and predictability, and also has the
protocol adaptively manages the duty cycle of sensor nodes advantage of being able to supply energy to many sensor nodes
according to their the amount of harvested energy as well as the simultaneously. Hence, the characteristics of the RF energy
contention time of sensor nodes considering fairness among them.
Through simulations, we show that our protocol can achieve a transfer system is suitable for the WSN with many nodes
high degree of fairness, while maintaining duty cycle of sensor and low power consumption. However we need a practical
nodes appropriately according to the amount of their harvested MAC protocol to efficiently utilize a small amount of energy
energy. harvested from RF energy transfer and it should be designed
Index Terms—wireless energy transfer, energy harvesting, considering the differences between RF energy transfer system
wireless sensor network, medium access control
and ambient energy harvesting system as follows:
I. I NTRODUCTION • The amount of energy harvested in the RF energy transfer
system is stable over time.
Wireless sensor networking is emerging technology with
• The amount of energy harvested by each of energy har-
wide range of potential applications, e.g. environment monitor-
vesting nodes is significantly different, highly depending
ing, smart spaces, medical systems, and military applications.
on the distance between the energy emitting node and the
Nodes in a wireless sensor network (WSN) are typically
energy harvesting node in the RF energy transfer system.
battery-powered. However, in many cases, it is difficult to
replace or recharge the exhausted battery. This is a reason To the best of our knowledge, almost all existing research
why the primary objective in WSN design is maximizing node on MAC protocols for the WSN mainly considered the battery-
and network lifetime, leaving the other performance metrics as powered sensor network, and the MAC protocol for WSN
secondary objectives. Hence, most of research on the WSN has with RF energy transfer, which is referred to as the RET-
focused on extending network lifetime by minimizing energy WSN in this paper, has not been actively studied. The MAC
usage. protocols for the battery-powered WSN can be classified
Recently, alternative methods for powering sensor nodes, into two categories: synchronous and asynchronous protocols.
which convert the ambient energy from environments into Synchronous approaches such as S-MAC [6], T-MAC [7],
electricity to power sensor nodes, are being actively studied and beacon enabled mode of IEEE 802.15.4 [8] synchronize
[1], [2]. This has motivated the search for alternative sources neighboring nodes in order to align their active or sleeping
of energy to power nodes in the WSN, especially for applica- periods. This approach greatly reduces idle listening time,
tions that require sensors to be installed for long durations but it is impractical in the RET-WSN, since energy harvest-
or embedded into structures where battery replacement is ing nodes are difficult to maintain synchronization due to
impractical. A variety of techniques are available for energy difference in the amount of energy harvested among nodes
harvesting in the WSN, including solar and wind powers, and a wake-up timer does not work properly when energy
thermoelectricity, and vibrational excitations. However, energy is exhausted. B-MAC [9], X-MAC [10], and RI-MAC [11],

978-1-4577-1177-0/11/$26.00 ©2011 IEEE 89 ICUFN 2011

which are typical asynchronous MAC schemes to support low
power consumption, require extra overheads in the sender side
such as long preamble or extra beacon transmission.
MAC protocols for the WSN powered by ambient energy
harvesting have been studied in [2], [12]–[14]. In [2], [12],
[13], the authors presented adaptive duty-cycling mechanisms
to dynamically adapt to unstable energy harvesting from
ambient environments such as solar and wind. However,
adapting to unstable energy harvesting is not a key issue in the Fig. 1. RF energy transfer model for the RET-WSN.
RET-WSN, since the amount of energy harvested in the RF
energy transfer system is stable over time. In [14], the authors
proposed CSMA-based and polling-based protocols for the
WSN powered by ambient energy harvesting (WSN-HEAP)
to improve the network throughput. However, the protocols
were designed for sensor nodes with unstable energy source.
In addition, they did not consider the unfairness among energy
harvesting nodes due to the significant difference between
energy harvesting rates that depend on the distance between
the energy emitting node and the energy harvesting nodes in
the RET-WSN. Fig. 2. System architecture of the RET-WSN.
Recently effective schemes for RF energy harvesting and
the WSN with RF energy harvesting have been studied [3],
[15]–[17]. In [15] and [16], the effective RF energy harvesting assume that a slave node can transmit at most one packet in
technologies are proposed for low power consuming devices. an active state and a packet has a fixed length.
Powercast developed the first practical RF energy harvesters Fig. 1 shows energy flow for the RET-WSN. Here, Ptx is the
and transmitters for the WSN which are now commercially power level transmitted from the master node to slave nodes
available [3]. In [17] the authors developed the WSN powered and Pin,i is the power level received by slave node i. Given
by harvested energy from TV radio to show the effectiveness Ptx , Pin,i can be determined by the Friis transmission equation
of RF energy harvesting in the WSN. [18] under idealized conditions as
In this paper, we propose a MAC protocol with energy  2
adaptive duty cycle and energy adaptive contention window λ
Pin,i = ePtx Gt Gr , (1)
for the RET-WSN, called energy adaptive MAC (EA-MAC), 4πRi
and analyze its performance. where e is the energy harvesting efficiency of slave nodes, Gt
The rest of this paper is organized as follows. In Section and Gr are the antenna gains of transmitting and receiving
II, we introduce the architecture of the RET-WSN. In Section antenna, respectively, λ is the wavelength, and Ri is the
III, we describe our EA-MAC protocol for the RET-WSN and distance between the master node and slave node i.
the simulation results are presented in Section IV. A summary
Slave nodes harvest RF energy with their antenna and
of the contributions and some concluding remarks are given
convert it to the appropriate DC voltage with their rectifier. To
in Section V.
provide appropriate power for transmitting data, slave nodes
II. RET-WSN ARCHITECTURE store energy into their energy storage such as a capacitor. Fig.
2 shows the system architecture of the RET-WSN.
In this section, we introduce the system architecture of the
RET-WSN considered in this paper. For the RET-WSN, we
III. EA-MAC
consider a star topology that consists of a single master node
which has the capability of gathering data and slave nodes Note that the amount of energy harvested by slave nodes
which are able to sense data and transmit it to the master in the RET-WSN is significantly different, depending on their
node. The master node operates using main power and emits distance to the master node, as previously mentioned. If this
RF energy to power slave nodes and the slave nodes harvest issue is not treated appropriately, it may cause the unfairness
energy transferred by the master node, as in Fig. 1. The master among slave nodes. Since a slave node that is far away from
node always stays awake to receive the data from the slave the master node will have lower energy harvesting level than a
nodes and the slave nodes go to sleep and active states back slave node that is close to the master node, the former should
and forth, depending on the level of their remaining energy. have longer sleep time to save energy more than the latter.
In the active state, a slave node contends for the channel and Hence, we need to investigate a new duty-cycle management
transmits its data, if it acquires the channel. In the sleep state, scheme that adapts to energy harvesting and consumption
a slave node completely turns off its radio and processor to conditions and a method of achieving fairness among slave
save energy except for an interrupt routine to wake up. We nodes according to their energy harvesting condition. To this

90
 Fig. 3. State transition diagram for EA-MAC.

end, in this paper, we propose a MAC protocol with energy-

adaptive duty cycle and energy-adaptive contention for the
RET-WSN.
A. Energy adaptive duty cycle management
As mentioned earlier, the master node always stays awake
to receive the data from slave nodes and slave nodes go to
sleep and active states back and forth, depending on the level
of their remaining energy. In the energy adaptive duty cycle
management mechanism of EA-MAC, to manage the duty
cycle of each slave node according to its harvested energy
level, the remaining energy level in its energy storage is used.
Initially, the slave node in the sleep state goes into the active
state to access the channel, i.e., the contention state, when
its harvested energy level reaches δ, which is an energy level
Fig. 4. CSMA/CA algorithm in EA-MAC.
threshold to transit from sleep state to active state. The value
of δ is determined such that the slave node has enough energy
to transmit one packet. If the slave node successfully gets the
The CSMA/CA algorithm works as follows. First, N Bi
channel, it goes into the transmit state and transmits its data
and BEi are initialized to 0 and minBE, respectively, where
packet. After packet transmission completes, the slave node
minBE is the minimum value of the backoff exponent. The
goes back to the sleep state. If the slave node fails to get the
slave node i waits for a random number of backoff slots 1
channel in the contention state, it goes back to the sleep state
in the range 0 to ⌈ωi × 2BE ⌉ − 1 and then performs a CCA
immediately. Once the slave node goes into the sleep state,
in order to check whether the channel is busy or not. If the
it remains in the sleep state until its energy level reaches
channel is assessed to be busy, it increases both N Bi and
the threshold δ again. In this way, the slave node repeats
BEi by one. If the value of N Bi is less than or equal to
transitions between sleep and active states according to its
maxCSMABackoffs 2 , it must return to the backoff procedure
energy harvesting condition. Fig. 3 depicts the state transition
for another random backoff, otherwise it declares the failure of
diagram for EA-MAC.
channel access and terminates the CSMA/CA algorithm, and
B. Energy adaptive contention algorithm then it goes into the sleep state. If the channel is assessed to
EA-MAC uses the CSMA/CA algorithm with the en- be idle during the CCA period, it goes into the transmit state
ergy adaptive contention algorithm based on the unslotted and transmits its data packet. Fig. 4 shows the procedure of
CSMA/CA algorithm in IEEE 802.15.4 [8]. The main dis- the CSMA/CA algorithm in EA-MAC.
tinguishing feature of our EA-MAC is that it has the energy
IV. P ERFORMANCE ANALYSIS
adaptive contention algorithm in which the backoff time of
each slave node is controlled by its energy harvesting rates. In this section, we evaluate the performance of our EA-
Each slave node i has three variables: N Bi , BEi , ωi for MAC in the OPNET simulator [19]. The parameters for the
the CSMA/CA algorithm of EA-MAC. First, N Bi is the RF energy transfer system, which are based on Powercast’s
number of clear channel assessment (CCA) performed in the TX91501 Powercaster and P2110 Powerharvester [3], are
CSMA/CA algorithm so far. Second, BEi is the backoff summarized in Table I. In addition, Table II summarizes
exponent, which is related to the maximum number of backoff the key parameters that we use for the EA-MAC in each
slots during which slave node i must wait before attempting sensor node. Most of these parameters are from IEEE 802.15.4
to assess the channel. Lastly, ωi is the weight factor. The standard [8] and the data sheet of Texas Instruments’ CC2430
weight factor is used to compensate the unfairness among [20].
slave nodes due to the significant difference between their
1 A backoff slot is the time period of a unit slot used for the backoff
energy harvesting rates and is calculated considering energy
procedure.
harvesting rate of the corresponding slave node and the average 2 maxCSMABackoffs is the maximum number of performing CCA before
energy harvesting rate of all slave nodes in the network. declare failure of channel access.

91
 TABLE I
PARAMETERS FOR RF ENERGY TRANSFER SYSTEM
25 2 meter
3 meter
Parameter Value Unit 4 meter
20
Frequency (C/λ) 915 MHz 5 meter

throughput (Kbps)
Energy harvesting efficiency (e) 0.58 – 6 meter
7 meter
Transmitter antenna linear gain (Gt ) 10 dBi 15
8 meter
Receiver antenna linear gain (Gr ) 6 dBi 9 meter
10 10 meter

TABLE II
PARAMETERS FOR EA-MAC 5

Parameter Value Unit 0

0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4 8.2 9 9.8 10.6 11.4 12.2 13 13.8 14.6
Frequency 2.401 GHz transmitted power (W)
Data rate 250 Kbps
A frame size 128 bits
Fig. 6. The throughput of EA-MAC/EAC.
Power consumption of contention state 31 mA
Power consumption of transmit state 29 mA
Power consumption of sleep state 0.5 µA 1
Operating voltage 3.0 V
minBE 3 – 0.9
EA−MAC
maxCSMABackoffs 4 – EA−MAC/EAC
A backoff slot 0.32 msec 0.8

fairness index
Active threshold (δ) 30 mJ
0.7

0.6

25 2 meter 0.5

3 meter
4 meter 0.4
20
5 meter
throughput (Kbps)

6 meter 0.3
0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4 8.2 9 9.8 10.6 11.4 12.2 13 13.8 14.6
15 7 meter transmitted power (W)
8 meter
9 meter
10 10 meter Fig. 7. Comparison of Fairness indices in EA-MAC and EA-MAC/EAC.

0
nodes by the master node. From Fig. 5 and Fig. 6, we can
0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4 8.2 9 9.8 10.6 11.4 12.2 13 13.8 14.6
transmitted power (W) observe that when the energy adaptive contention algorithm
is applied, the throughput of the slave nodes having relatively
Fig. 5. The throughput of EA-MAC. lower energy harvesting rate is increased, while the throughput
of the slave nodes having relatively higher energy harvesting
rate is decreased, which implies that we can improve the
In our simulation, we deploy one master node and 9 slave degree of fairness among slave nodes by using the energy
nodes, each of which is located at the place from 2 meters adaptive contention algorithm. Especially, the throughput of
to 10 meters away from the master node with the interval of the slave node at 10 meters is increased about 240 %, while
one meter. We compare the throughput of slave nodes with the throughput of the slave node at 2 meters is decreased about
varying transmitted power for RF energy transfer from the 70 %.
master node. We also compare the performance of the EA- To show the degree of fairness among slave nodes more
MAC with energy adaptive contention algorithm and that of clearly, in Fig. 7 we provide Jain’s fairness index [21], which
the EA-MAC without it. We call the former as EA-MAC/EAC is widely used to measure the degree of fairness in many
and the latter as EA-MAC. resource allocation schemes in communication networks and
Fig. 5 shows the throughput of EA-MAC. From the results, is given by
we observe that a slave node that is close to the master node
( ni=1 Si )2

achieves much higher throughput than a slave node that is far I = n , (2)
away from the master node, since each slave node adaptively n i=1 Si2
controls the duty cycle according to the amount of energy where n is the number of slave nodes, Si is the throughput of
harvested by it. slave node i. The fairness index I has a value between zero
Fig. 6 shows the throughput of EA-MAC/EAC. In EA- and one and in general, as the degree of fairness increases,
MAC/EAC, slave nodes use the weight factor to compensate the value of I also increases. As the figure shows, EA-
the unfairness due to the significant difference between energy MAC/EAC achieves a higher degree of fairness than EA-MAC,
harvesting rates of the slave nodes. The weight factor, ωi , of as expected.
slave node i is calculated by the ratio of its harvested energy to The improvement of the degree of fairness in EA-
the average harvested energy of all slave nodes in the network. MAC/EAC comes from the adaptive control of the contention
The average harvested energy can be informed to each slave time from the energy adaptive contention algorithm. To show

92
 −3
x 10
8 60

2 meter EA−MAC
7
3 meter 50 EA−MAC/EAC

network throughput (Kbps)

4 meter
contention time (Sec) 6
5 meter
40
6 meter
5
7 meter
8 meter 30
4
9 meter
3 10 meter 20

2
10

1
0
0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4 8.2 9 9.8 10.6 11.4 12.2 13 13.8 14.6
0
0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4 8.2 9 9.8 10.6 11.4 12.2 13 13.8 14.6 transmitted power (W)
transmitted power (W)

Fig. 10. Comparison of network throughputs in EA-MAC and EA-

Fig. 8. The contention time of EA-MAC. MAC/EAC.

−3
x 10
8

7
2 meter fairness among slave nodes. Our simulation results showed
3 meter

6
4 meter that EA-MAC with the energy adaptive contention algorithm
contention time (Sec)

5 meter
6 meter
can significantly improve the degree of fairness compared with
5
7 meter EA-MAC without the energy adaptive contention algorithm.
4 8 meter
9 meter
3 10 meter VI. ACKNOWLEDGMENTS
2
This work was financially supported by the grant from
1
the strategic technology development program (Project No.
0
0.2 1 1.8 2.6 3.4 4.2 5 5.8 6.6 7.4 8.2 9 9.8 10.6 11.4 12.2 13 13.8 14.6 10033869) of the Ministry of Knowledge Economy (MKE) of
transmitted power (W)
Korea and by the KCC(Korea Communications Commission),
Fig. 9. The contention time of EA-MAC/EAC.
Korea, under the R&D program supervised by the KCA(Korea
Communications Agency) (KCA-2011-(11913-04004)).

the effectiveness of the energy adaptive contention algorithm, R EFERENCES

we compare the contention times of slave nodes in EA-MAC [1] S. Sudevalayam and P. Kulkarni, “Energy harvesting sensor nodes:
and EA-MAC/EAC in Fig. 8 and Fig. 9, respectively. As Survey and implications,” IEEE Communications Surveys & Tutorials,
vol. PP, no. 99, pp. 1–19, 2010.
shown in Fig. 8, in EA-MAC, all slave nodes have almost [2] W. Seah, Z. Eu, and H. Tan, “Wireless sensor networks powered by
the same contention time, since in EA-MAC, all slave nodes ambient energy harvesting (WSN-HEAP)-Survey and challenges,” in
use the same equation to calculate the backoff time. However, Wireless VITAE 2009, May 2009, pp. 1–5.
as shown in Fig. in EA-MAC/EAC, as a slave node has a [3] Powercast Corporation, TX91501 User’s Manual & P2110’s Datasheet.
available online at: http://www.powercastco.com/resources/.
lower energy harvesting rate, it has a shorter contention time, [4] R. Selvakumaran, W. Liu, B. Soong, L. Ming, and Y. L. Sum, “Design
since it can have a shorter backoff time according to its weight of low power Rectenna for wireless power transfer,” in IEEE TENCON
factor. A shorter contention time of a slave node with a lower 2009, Jan. 2009, pp. 1–5.
[5] Z. Sim, R. Shuttleworth, M. Alexander, and B. Grieve, “Compact Patch
energy harvesting rate can helps to improve its throughput, Antenna Design for Outdoor RF Energy Harvesting in Wireless Sensor
which results in the improvement of the degree of fairness in Networks,” Progress In Electromagnetics Research, vol. 105, pp. 273–
EA-MAC/EAC. 294, 2010.
[6] W. Ye, J. Heidemann, and D. Estrin, “An energy-efficient MAC protocol
Finally, in Fig. 10, we compare network throughput for wireless sensor networks,” in IEEE INFOCOM 2002, vol. 3, Jun.
achieved by EA-MAC/EAC with that achieved by EA-MAC. 2002, pp. 1567–1576.
As the figure shows, the network throughput of EA-MAC/EAC [7] T. Van Dam and K. Langendoen, “An adaptive energy-efficient MAC
protocol for wireless sensor networks,” in ACM SenSys 2003, Nov. 2003,
is lower than that of EA-MAC. This implies that the improve- pp. 171–180.
ment of the degree of fairness in EA-MAC/EAC is achieved [8] Part 15.4: Wireless Medium Access Control (MAC) and Physical Layer
at the price of the decrement of the network throughput. (Phy) Specifications for Low-Rate Wireless Personal Area Networks (LR-
WPANS). IEEE 802.15.4. Std, 2006.
V. C ONCLUSION [9] J. Polastre, J. Hill, and D. Culler, “Versatile low power media access for
wireless sensor networks,” in ACM SenSys 2004, Nov. 2004, pp. 03–05.
In this paper, we proposed a MAC protocol with energy [10] M. Buettner, G. Yee, E. Anderson, and R. Han, “X-MAC: a short
adaptive duty cycle and energy adaptive contention window preamble MAC protocol for duty-cycled wireless sensor networks,” in
ACM SenSys 2006, Oct. 2006, pp. 307–320.
that can be used in wireless sensor networks based on RF [11] Y. Sun, O. Gurewitz, and D. Johnson, “RI-MAC: a receiver-initiated
energy transfer. The proposed MAC protocol adaptively man- asynchronous duty cycle MAC protocol for dynamic traffic loads in
ages not only the duty cycle of the slave node according wireless sensor networks,” in ACM SenSys 2008, Nov. 2008, pp. 1–14.
[12] J. Hsu, S. Zahedi, A. Kansal, M. Srivastava, and V. Raghunathan,
to its harvested energy level but also the contention time of “Adaptive duty cycling for energy harvesting systems,” in ACM LSLPED
the slave node by its energy harvesting rate to achieve the 2006, Oct. 2006, pp. 180–185.

93
[13] C. Vigorito, D. Ganesan, and A. Barto, “Adaptive control of duty cycling
in energy-harvesting wireless sensor networks,” in IEEE SECON 2007,
Jun. 2007, pp. 21–30.
[14] Z. Eu, W. Seah, and H. Tan, “A study of MAC schemes for wireless
sensor networks powered by ambient energy harvesting,” in WICON
2008, Nov. 2008, p. 78.
[15] H. Yan, J. Montero, A. Akhnoukh, L. de Vreede, and J. Burghartz, “An
integration scheme for RF power harvesting,” in SAFE 2005, Nov. 2005,
pp. 17–18.
[16] T. Ungan and L. Reindl, “Harvesting low ambient RF-sources for
autonomous measurement systems,” in IEEE IMTC 2008, May 2008,
pp. 62–65.
[17] H. Nishimoto, Y. Kawahara, and T. Asami, “Prototype implementation
of wireless sensor network using TV broadcast RF energy harvesting,”
in ACM Ubicomp 2010, Sep. 2010, pp. 373–374.
[18] T. Rappaport, Wireless communications. Prentice Hall PTR, 2002.
[19] OPNET modeler & wireless suit. available online at:
http://www.opnet.com/solutions/network rd/modeler wireless.html.
[20] B. Selvig, “Measuring power consumption with CC2430 & Z-Stack,”
Application Note AN053, Texas Instruments.
[21] M. Dianati, X. Shen, and S. Naik, “A new fairness index for radio
resource allocation in wireless networks,” in IEEE WCNC 2005, vol. 2,
Mar. 2005, pp. 712–717.

94