• No results found

Analysis of Energy Consumption for IEEE 802 15 4 MAC and S MAC Protocols in Wireless Sensor Network

N/A
N/A
Protected

Academic year: 2020

Share "Analysis of Energy Consumption for IEEE 802 15 4 MAC and S MAC Protocols in Wireless Sensor Network"

Copied!
6
0
0

Loading.... (view fulltext now)

Full text

(1)

Abstract— One of the major challenges of wireless sensor networks (WSN) is how to utilize efficiently battery energy resource to extend the lifetime of the entire network because of power constrained devices. Many MAC (Medium Access Control) protocols have been taken to decrease the energy consumption of sensor nodes such as IEEE 802.11, 802.15.4 MAC or sensor MAC (S-MAC) protocols. In this paper, we provide analyses performance of IEEE 802.15.4 MAC and S-MAC protocols in terms of energy consumption, energy efficiency and data packet delivery. Our simulation results show that S-MAC protocol with active and sleep cycle consumes energy less than about 10% and 15% compared to IEEE 802.15.4 beacon and non-beacon enabled mode respectively.

Index Terms— Wireless Sensor Networks, energy-efficient, S-MAC, IEEE 802.15.4, MAC.

I. INTRODUCTION

Wireless sensor network (WSN) includes hundreds or thousands of micro-sensor nodes which is deployed in various fields such as military, environment monitor, intelligent home and so on [1]. Sensor nodes have a small size low cost, processor abilities, RAM and resources. Especially, the battery power of sensor node is not recharged during the active time of network. Therefore, it is very important for considering energy consumption of MAC protocols in order to save energy and prolongs the lifetime of network. MAC protocols are designed at MAC sub-layer in data link layer in OSI model, which is responsible for controlling medium access so that the nodes in network can communicate with other nodes available without occurring collision. Besides, the energy efficient also is one of the most important for designing MAC protocols in order to extend the life of the network as long as possible.

In WSN, a sensor node consumes energy in idle listening of the channel, transmission, reception, sleep state or transition state, in which idle listening is one of the most significant sources of energy consumption in sensor nodes. In order to limit the problem of idle listening, currently, many MAC protocols has proposed by researchers for this problem as well as evaluated about energy efficient of that, such as

Manuscript received March, 2020.

Nguyen Duy Tan, Faculty of Information Technology, Hung Yen University of Technology and Education, Hungyen, Vietnam, e-mail: [email protected],

Pham Ngoc Hung, Faculty of Information Technology, Hung Yen University of Technology and Education, Hungyen, Vietnam, e-mail:

IEEE 802.11, IEEE 802.15.4 MAC protocol or Sensor-MAC (S-MAC) protocols [1, 2].

Bengheni [3] et al. has compared energy consumption of asynchronous MAC protocols in wireless sensor networks: BMAC, XMAC and RIMAC that use a duty-cycle to reduce idle listening, which is cause of waste energy.

In [4] the authors have analyzed the performance of S-MAC, which operate at different duty cycles and estimate the parameters required to achieve any desired throughput, data rate and energy consumption.

Recently, Kobayashi [5] et al. proposes a method energy saving by combining the IEEE 802.11with IEEE 802.15.4. If the big data packets is are sent, the method will use IEEE 802.11 for high throughput. On the other hand, if the size of data packet is small, IEEE 802.15.4 is selected to decrease energy consumption. Up until now, there have been many analysis and evaluation of energy efficient MAC protocols in WSN such as simulation and performance evaluation of energy efficient S-MAC and RI-MAC protocols [6], simulation and analysis of energy consumption for S-MAC and T-MAC protocols [7], energy consumption in mobile ad hoc networks [8]. However, none of the above evaluations consider the energy efficient between S-MAC and the IEEE 802.15.4 MAC protocols, which is worked at beacon and non-beacon enabled mode.

In this paper, we focus on the performance evaluation of the S-MAC and 802.15.4 MAC protocols to help the development of power saving schemes in WSN. Our simulation results show that the energy consumption of S-MAC with active and sleep cycle consumes energy less than about 10% and 15% compared to IEEE 802.15.4 MAC protocols at beacon and non-beacon enabled mode, respectively, in case of large network (50 nodes deployed in 500m×500m area). Besides the energy consumption of idle listening state is the most in comparison with transmission and reception state in all protocols.

II. PROTOCOLSDESCRIPTION

In this section, we briefly describe the IEEE 802.15.4 MAC and S-MAC protocols, which are used in our analysis. A. IEEE 802.15.4 Medium Access Control

IEEE 802.15.4 standard is designed for low-rate and low-power applications [9, 10]. In physical layer, it can work at three operational frequency bands: 868 MHz, 915 MHz, and 2.4 GHz bands. There are 27 sub-channels defined in IEEE 802.15.4 standard, which consists of 16 sub-channels in

Analysis of Energy Consumption for IEEE

802.15.4 MAC and S-MAC Protocols in

Wireless Sensor Network

(2)

sub-channel in the 868 MHz band. In MAC sub-layer, IEEE 802.15.4 standard can operate in two modes: a beacon enabled mode and non-beacon enabled mode.

In a non-beacon enabled mode, IEEE 802.15.4 uses un-slotted Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) mechanism to control channel access and maintain network activities. In a beacon-enabled mode, the network is managed by a coordinator device, which regularly transmits a beacon frame to other devices to synchronize, identify the PAN and describe the structure of the super-frames. The structure of the super-frame is defined by the coordinator, which consists of an active and inactive periods, as shown in Figure 1. The active periods of each super-frame is divided into three parts: a beacon, a contention access period (CAP) and a contention-free period (CFP). In the CAP periods, the beacon frame is transmitted at the start of time slot 0 without using CSMA/CA algorithm. Any device wants to communicate during the CAP between two beacons, it shall contend with other devices to get channel by using a slotted CSMA/CA mechanism. After CAP, CFP always appears at the end of the active period of super-frame, which starts at a time slot boundary immediately following the CAP. The PAN coordinator may allocate up to seven of guaranteed time slots (GTS), and a GTS can occupy more than one time slot.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

SD = aBaseSuperframeDuration 2SO Symbols

CAP CFP

BI = aBaseSuperframeDuration 2BOSymbols

Beacon

 

Beacon Time slots

BP

Inactive Period

ACK Data Frame

Data Frame IFS t IFS

Acknowledged transmission ack

Unacknowledged transmission

GTS duration

GTS duration Tdata Toverhead

Tdata Toverhead

[image:2.595.317.557.238.695.2]

Active Period GTS GTS

Figure 1. Super-frame structure of IEEE 802.15.4

According to IEEE 802.15.4 standard [9, 10], the structure of the super-frame is illustrated by the values of Beacon Order (BO) and Super-frame Order (SO). The length of the super-frame referred to as Beacon Interval (BI), and the length of its active periods is defined through the Super-frame Duration (SD) parameter as follows:

. 14 0

, 2 *

14 0

, 2 *

   

  

BO SO ion

rframeDuat aBaseSuppe

SD

BO ion

frameDurat aBaseSuper

BI

SO BO

(1)

WhereaBaseSuperframeDurationdescribes the minimum duration of super-frame corresponding to SO = 0 and this duration is fixed by 960 symbols, 1 symbol is equivalent to 4 bits and equal 16 µs, assuming 250 kbps in the 2.4 GHz frequency band [10]. In addition, in order to save energy, all the nodes will go into a sleep mode during the long inactive period in schedule.

Slotted and Un-slotted CSMA/CA mechanism

Slotted CSMA/CA: This algorithm is used in beacon enabled modes for channel access, each device maintains three variables for each transmission attempt: Number of Backoffs (NB), which indicates the number of times the CSMA/CA algorithm required to back-off while attempt accessing the channel; its value is initialized to 0 for each new transmission attempt as illustrated in Figure 2. Contention Window (CW) represents the number of back-off periods that need to execute Clear Channel Assessment (CCA) procedure before the transmission can start. Back-off Exponent (BE) used to compute the Back-off Delay (BD) before assessing the channel, which is a random number between 0 and (2BE -1),BEis initialized tomacMinBE.

BE=min(2, ) macMinBE NB=0, CW=2

True

Channel idle?

False True

NB> macMaxCSMA

Backoffs?

CW=CW-1 Battery life

extension?

False

BE=macMinBE

Locate Backoff periods boundary

Wait for random(0 and 2 -1)*Unit BackoffPeroids

Perform CCA on Backoff periods boundary

CW=2, NB=NB+1 min(

BE= BE+1,amacMinBE)

CW=0? False

True

True

False

BE Start

Have data to send? True Receive beacon, command frame? Sleep to next beacon,

command frame?

False

Transmit data and wait for ACK

Receive ACK within macACKwaitDuration

time?

True

False

Data frame is dropped

Figure 2. Operations of IEEE 802.15.4 in the beacon enabled mode with slotted CSMA/CA algorithm

[image:2.595.50.290.366.563.2]
(3)

random period of time and then try again. The macMaxCSMABackoffparameter is maximum number of the CSMA-CA algorithm, which it will attempt before declaring a channel access failure (macMaxCSMABackoff = 5) as shown in Figure 3.

False True

NB> macMaxCSMA

Backoff? BE=macMinBE, NB=0

NB=NB+1, BE=min(BE+1,macMinBE)

True False

Start

Wait for random(0 and 2 -1)*Unit BackoffPeroids

Perform CCA

BE

Channel idle?

Data frame is dropped

Have data to send? True Receive command

frame Sleep to next

command frame?

Transmit data and wait for ACK

Receive ACK within

macACKwaitDuration

time?

True

[image:3.595.54.284.120.434.2]

False False

Figure 3. Operations of IEEE 802.15.4 in non-beacon enabled mode with slotted CSMA/CA algorithm

Data Transmission Models

There are two types of data transmission model in IEEE 802.15.4 networks: data transmission from coordinator to device and data transmission from device to coordinator.

Data transmission from device to coordinator: In a beacon enabled mode, when a device has data to send to the coordinator, it first listens for beacon frame to synchronize with the coordinator as in the super-frame structure.

Network

Device Coordinator

Beacon Data

Acknowledgment Network

Device Coordinator

(a) (b)

Data

[image:3.595.305.549.206.330.2]

Acknowledgment

Figure 4. Data transmission to the coordinator in (a) beacon enabled and (b) non-beacon enabled mode

Then it transmits its data frame, using slotted CSMA/CA algorithm as shown in Figure 4(a). After receiving the frame, the coordinator sends back an acknowledgment frame to confirm the successful reception of the data. On the other hand, in a non-beacon enabled mode, when a device has data

to send to the coordinator, it simply transmits its data frame by using un-slotted CSMA/CA algorithm, as shown in Figure 4(b).

Data transfer from the coordinator:In a beacon enabled mode, when the coordinator wants to send data to a device, firstly it will inform in the beacon frame that there is data frame to be transmitted. The device periodically listens to beacon frame and know that the coordinator has data to transmit; it transmits the MAC command frame to request data, using slotted CSMA/CA algorithm. The device sends an acknowledgment frame to notify the successful reception as in Figure 5(a).

Network

Device Coordinator Beacon

Data

Acknowledgment

(a) (b)

Data request

Acknowledgment

Network

Device Coordinator

Data

Acknowledgment Data request

[image:3.595.47.289.593.682.2]

Acknowledgment

Figure 5. Data transmission from the coordinator in (a) beacon enabled and (b) non-beacon enabled mode

On the other hand, in non-beacon enabled mode, when coordinator wants to transfer data to a device, it makes connect and requests the data frame. The device accepts a proposal requesting by transmitting a MAC command frame, using un-slotted CSMA/CA. Both coordinator and device will transmit an acknowledgment frame to authenticate the successful reception of the data frame.

B. Sensor-MAC (S-MAC)

S-MAC is a medium access control protocol base on contention-based random access that allows nodes directly communicate to each other in network. S-MAC is designed to reduce energy consumption from all the sources for wireless sensor networks that we can identify to cause energy waste (collision, idle listening, overhearing and control overhead) by using fixed listen and sleep duty cycle called a time frame, in which nodes periodically transition between a listen state and a sleep state to reduce energy consumption in idle listening channel.

A time frame in S-MAC contains two parts: one for a listening period and the other for a sleeping period as shown in Figure 6 follow:

(4)

Sleep time

SYNC DATA

Sender node ID Next sleep time

[image:4.595.64.276.48.133.2]

Sleep Listen Listen

Figure 6. Listen and sleep period of S-MAC

In order to synchronize the time of listen and sleep period among nodes in network, the nodes regularly share their information about schedule table by broadcasting SYNC message, which is very small and consist of node ID (identification), the next sleep time...

In network, a packet collision occurs when two or more nodes attempt to transmit packets into the medium over the network at the same time. Packet collisions can be the cause of wasting energy and decreasing performance network. To solve this problem, S-MAC uses traditional mechanisms like the IEEE 802.11 such as the exchange of RTS/CTS message and using ACK message to affirm a good data packet received. In addition, S-MAC combines the physical carrier sensing called Clear Channel Assessment (CCA) and virtual carrier sensing called Network Allocation Vector (NAV) to avoid collision and overhearing. NAV contains a value, only when this value is set to zero, packets can be transmitted.

C. Simulation Parameters

[image:4.595.370.548.300.476.2]

To evaluate the performance of IEEE 802.15.4 MAC and S-MAC protocols, we use the network simulator ns-2 (v.2.35) [11] with the parameters in the scenarios that are described in "ns-allinone-2.34/ns-2.34/wpan/p802_15_4const.h" file and Table I, [11, 12].

TABLE I: THEARRANGEMENT OFCHANNELS

Parameters Values

Topology area 500 m × 500 m

Numbers of nodes 50

Antenna type Omni Antenna

Routing protocol AODV

Packet size 128 bytes

Simulation time 500 seconds

Transmission range (m) 250

Traffic type CBR

Data rate 1 (kbps)

Initial energy 2 (Joules)

Idle power 712e-6 (Watt)

Receiving power 0.328 (Watt)

Transmission power 0.3132 (Watt)

Sleep power 144e-9 (Watt)

D. Performance Metrics 1) Energy Consumption

Energy consumption of a node: This denotes the relationship among energy dissipation to the total of data packets delivered by each node in the network. A node consists of energy consumption in different states as transmission (ETx), reception (ERx), idle listening (EIdle),

sleeping (ESlep), transition (ETrans) and Clear Channel

Accessment (ECCA) state that can be calculated as follows:

CCA CCA N

k trans trans

Sleep Sleep Idle Idle Nrx

j Rx j

Ntx

i Tx i

CCA Trans Sleep Idle Rx Tx Node T P k T P T P T P R Ps P R Ps P E E E E E E E trans            

   1 1 1 ) ( / / (2)

whereEx, PxandTxare the energy consumption (joules), the

power (watt) and the time interval of transceiver in state x (second) respectively.PsiandRare the size of length of the ith

packet of receiving or sending andRis the data transferring rate. Ntxand Nrx are total numbers of receiving or sending

packets.

ni Node

Network

E

i

E

1

(

)

(3)

where ENetwork are the energy consumption of all nodes in

network,nis number of nodes in network.

Energy consumption percentage in states of network is calculated as follows:

Network n i Tx network Tx

E

i

E

E

1

_

)

(

(%)

(4) Network n i Rx network Rx

E

i

E

E

1

_

)

(

(%)

(5) Network n i Idle network Idle

E

i

E

E

1

_

)

(

(%)

(6) Network n i CCA network CCA

E

i

E

E

1

_

)

(

(%)

(7)

Energy consumption percentage on MAC overhead of network is the ratio of energy consumption for transmitting and receiving short control message at MAC sub-layer, i.e. beacon frame (BCN), synchronous frame (SYNC), command (CM) frame and acknowledgement (ACK) of all nodes in network, which can be calculated as formula (8).

Network n

i TX RX for BCN SYNC ACK

overhead MAC E i E i E

E

1  _ _ _

_ ) ( ) ( (%) (8) 2) Throughput

Throughput of a connection express the total of data frames transported to destination node of one end-to-end connection (flow) in network during the simulation time. The throughput of the entire network expresses the sum of the throughput of connections and it is measured in bits per second (bit/s or bps) as formula (8).

) ( 8 * 1 2 1 bps t t Ps _of_ flow Throughput m i i

j

 (9) j k j ) t_of_ flow (Throughpu rk _of_ netwo Throughput

  1 (10)

where Psiis the size of length of the ithframe reaching the

[image:4.595.47.287.493.670.2]
(5)

source node and the time when last frame received by destination node, respectively.

3) Energy Efficiency

Energy efficiency is defined as the throughput achieved per unit of energy consumed, where the throughput represents the number of successfully delivered packets.

) (

_ _ _ ( )

_

joules n consumptio Energy

bps network of

Throughput efficiency

Energy  (11)

4) Packet Delivery Ratio (PDR):

PDR represents the ratio of data frames successfully received from all the sent data frames, which is computed as Equation (12) below:

Ns Nr

PDR (12)

where Nr and Ns are the number of frames received by destination node and the number of frames sent by source node, respectively.

III. RESULTS ANDANALYSIS

Figure 7 represents the power consumption of IEEE 802.15.4 enabled, non-enabled mode and S-MAC protocols during simulation time of nodes. It is clearly observable that the S-MAC protocol with active and sleep cycle has better performance in reducing energy consumption of nodes than IEEE 802.15.4 MAC protocols.

Number of Nodes

0 5 10 15 20 25 30 35 40 45 50 0

10

En

er

gy

C

on

su

m

pt

io

n

(J

ou

le

s)

30

20 40 50

60 S-MAC

[image:5.595.306.531.48.208.2]

802.15.4 non-Beacon 802.15.4 Beacon

Figure 7. Energy consumption during the simulation time

As illustrated in Figure 8 and 9, the energy consumption of nodes when we increase the number of end-to-end connections (flows) and the interval of transmission time in network. It seems that energy consumed increases as the number of nodes sent data packets increases and energy consumption decrease when the interval time increases with all MAC protocols. Therefore, it is the most energy consumption with IEEE 802.15.4 beacon enabled mode and S-MAC has the lowest consumption of energy.

Figure 10 shows the energy efficiency of protocols with number of sent nodes, we can observe that S-MAC consumes less amount of energy than the others protocols; IEEE 802.15.4 non-beacon enabled is better energy efficiency compared to 802.15.4 beacon enabled mode.

Number of End-to-End Connections

0 2 4 6 8 10

0 20 40 60 80 100 120

En

er

gy

C

on

su

m

pt

io

n

(%

)

1 3 5 7 9

S-MAC

802.15.4 non-Beacon 802.15.4 Beacon

Figure 8. Energy consumption per number of flows

0 20

En

er

gy

C

on

su

m

pt

io

n

(J

ou

le

s)

60

40 80 100

Interval (Sec)

3 7 11 15 19

1 5 9 13 17

S-MAC

[image:5.595.305.542.55.610.2]

802.15.4 non-Beacon 802.15.4 Beacon

Figure 9. Energy consumption per interval

Number of Nodes

En

er

gy

E

ffi

ci

en

t (

K

bp

s/J

ou

le

)

S-MAC

802.15.4 non-Beacon 802.15.4 Beacon

0 5 10 15 20 25 30 35 40 45 50

Figure 10. Energy efficiency

The percentage of energy consumption in different states of all nodes in network are illustrated in Figures 11 in which Idle listening state consumes more energy than other states in S-MAC because nodes have to listen medium to receive frames coming. Meanwhile, IEEE 802.15.4 has the highest power consumption in CCA because nodes usually turn on the radio component to sense medium, which is the most energy consumed component in sensor node, followed by Idle and sleep one is at least [13].

[image:5.595.311.542.310.586.2] [image:5.595.54.284.408.565.2]
(6)

of not sending beacon frame. However, S-MAC with good energy efficiency still consumes the least energy for MAC sub-layer overhead.

0 10 20 40 60

En

er

gy

C

on

su

m

pt

io

n

in

d

iff

er

en

t s

ta

te

s (

%

)

30 50

S-MAC 802.15.4 non-Beacon 802.15.4 Beacon Idle Listening

Sleeping TransmissionReception TransitionCCA

70

39.1

3

0.

43

29

.7

3

0.

03

2

30

.6

8

29

.8

9

9.

27

6.47

54

.3

8

28

.2

10

.0

5

7.

12

54.6

[image:6.595.59.291.100.274.2]

1

Figure 11. Percentage of energy consumption in several states

Number of Nodes

S-MAC

802.15.4 non-Beacon 802.15.4 Beacon

En

er

gy

C

on

su

m

pt

io

n

(%

)

[image:6.595.54.285.308.466.2]

0 5 10 15 20 25 30 35 40 45 50

Figure 12. Percentage of energy consumption due to MAC overhead

Number of End-to-End Connections

0 1 2 3 4 5 6 7 8 9 10

0 0.1 0.2 0.3 0.4 0.5

Pa

ck

et

D

el

iv

er

y

R

at

io

(%

)

S-MAC

802.15.4 non-Beacon 802.15.4 Beacon

Figure 13. Packet delivery ratio

We illustrate the packet delivery ratio for both three protocols in the number of end-to-end connections (flows). Based on results shown in Figure 13, we can obviously observe that the packet delivery ratio in the network in the S-MAC protocol is higher than about 70% compared to IEEE 802.15.4 MAC protocols.

IV. CONCLUSION

In this paper, we analyzed the energy consumption of nodes in wireless sensor network considering the interactions of the IEEE 802.15.4 MAC and S-MAC protocols. Our goal is to evaluate performance of MAC protocols which helps the development of power saving schemes in WSN. Our simulation results show that the energy consumption of S-MAC with active and sleep cycle is better than that of IEEE 802.15.4. IEEE 802.15.4 non-beacon enabled mode has energy efficient higher than IEEE 802.15.4 beacon enabled mode about 5% case of large network (50 nodes deployed in 00m×500m area). Besides Idle listening state consumes more energy than other states in S-MAC, IEEE 802.15.4 has the highest power consumption in CCA state because nodes usually turn on the radio component to sense medium, followed by Idle and sleep stases are at least in all states.

REFERENCES

[1] W. Dargie and C. Poellabauer, Fundamentals of Wireless Sensor Networks Theory and Practice, USA: John Wiley & Sons, 2010, ch. 6, pp. 125-161.

[2] D. Simaiya, U. Sharma and A. N. Tripathi, "Simulation and Performance Evaluation of Energy Efficient MAC Protocols for Wireless Sensor Networks," Tenth International Conference on Wireless and Optical Communications Networks (WOCN), pp. 1-5, July 2013.

[3] A. Bengheni, F. Didi and I. Bambrik, "Energy-Saving Comparison of Asynchronous MAC Protocols for Wireless Sensor Networks," International Conference on Mathematics and information Technology, pp. 263-268, December 2017.

[4] C. Wang, Y. Chen and Y. Hou, "The analysis and improvement of SMAC protocol for Wireless sensor networks,"IEEE 9th International Conference on Mobile Ad-hoc and Sensor Networks, pp. 437-441, December 2013.

[5] H. Kobayashi, S. Izumi and K. Takahashi, "Proposal of IEEE 802.11 Wake-Up Control Method Using IEEE 802.15.4 for Low Energy Consumption," the 5th International Conference on Business and Industrial Research (ICBIR), pp. 17-20, May 2018.

[6] S. V. Rao and S. S. Pillai, "Performance of Sensor-MAC in an Energy Harvesting Environment," Proceedings of IEEE International Conference on Circuits and Systems, pp. 157-161, December 2017. [7] R. Munadi, A. E. Sulistyorini, F. U. Fauzi and T. Adiprabowo,

"Simulation and Analysis of Energy Consumption for S-MAC and T-MAC Protocols on Wireless Sensor Network,"IEEE Asia Pacific Conference on Wireless and Mobile, pp.142-146, August 2015. [8] H. Xiao, D. M. Ibrahim and B. Christianson, "Energy Consumption in

Mobile Ad Hoc Networks," IEEE Wireless Communications and Networking Conference (WCNC), pp. 2599-2604, April 2014. [9] Y. K. Huang, S. W. Huang and A. C. Pang, "An Energy-Efficient MAC

Design for IEEE 802.15.4-Based Wireless Sensor Networks," IFIP International Federation for Information Processing, pp. 237-248, 2007.

[10] LAN/MAN Standards Committee, IEEE Standard for Local and metropolitan area networks-Part 15.4: Low-Rate Wireless Personal Area Networks (LR-WPANs), IEEE Computer Society, 2011. [11] VINT Project, “The network simulator - NS2,”

http://www.isi.edu/nsnam/ns, 1997 (accessed: March 5, 2020). [12] M. Errouidi, H. Moudni, H. Mouncif and A. Merbouha, "An Energy

Consumption Evaluation of Reactive and Proactive Routing Protocols in Mobile Ad-hoc Network," the 13th International Conference Computer Graphics, Imaging and Visualization, pp. 437-441, March 2016.

[image:6.595.56.283.499.667.2]

References

Related documents