EM-MAC: A Dynamic Multichannel Energy-Efficient
MAC Protocol for Wireless Sensor Networks
Lei Tang1, Yanjun Sun2, Omer Gurewitz3, and David B. Johnson1
1Department
of Computer Science, Rice University, USA
and Applications R&D Center, Texas Instruments, USA
3Department of Communication Systems Engineering, Ben Gurion University, Israel
2Systems
ACM MobiHoc 2011 (Best Paper Award)
Outline





Introduction
System Model and Assumptions
EM-MAC Protocol Design
Evaluation on MICAz Motes
Conclusion
Introduction
 Energy efficiency is crucial in wireless sensor networks (WSNs).
 Many energy-efficient wireless MAC protocols have been proposed that
utilize duty cycling.
 In real world, WSN MAC protocols must operate under a number of
significant challenges.



collision or contention
interference from wireless transmissions by other types of devices, such as
Wi-Fi nodes, or jamming attack
time synchronization
Goals
 This paper proposes a new multichannel energy-efficient MAC
Protocol, called EM-MAC (Efficient Multichannel MAC).
 EM-MAC




is a distributed protocol.
is a asynchronous duty-cycling MAC protocol.
utilizes the available multiple orthogonal radio channels common with
many types of wireless devices.
avoids jamming attack
System Model and Assumptions
 Each node equips with only one transceiver.
 The network is a multichannel environment.
 Nodes do not require to synchronize their clocks.
 There is no common control channel.
EM-MAC
 Overview
k
A
A
A
A
i
j
A
A
A
EM-MAC
 Overview
1
B
DATA
ACK
S: 2
B
DATA
ACK
B
DATA
ACK
3
1
B
DATA
ACK
R: 2
3
B
EM-MAC
 How to choose the set of switching channels?
 How to avoid using congested channels and to be robust
against wireless interference and jamming?
 How to address the rendezvous problem in an asynchronous
environment?
How to choose the set of switching channels?
Switching Channel Selection
 Linear Congruential Generator (LCG)
X n1  aXn  c modm
m>0 is a modulus, a is a multiplier, c is an increment,
Xn is the current seed, and Xn+1 is the next seed.
D. E. Knuth,A:“The
Congruential
The Art of
1, 4, Linear
8, 11, 14,
2, 7, 12, 13,Method,”
15, 1, 2, …
Computer Programming, Third Edition, Volume 2:
Seminumerical Algorithm, pages 10-26. Addison-Wesley,
A: 1, 4, 8, 4, 8, 1, 1, 4, 8, 4, 4, 8, 1, 4, …
1997.
How to avoid using congested channels and
to be robust against wireless interference and
jamming?
Basic Concept
A: 1, 4, 8, 11, 14, 1, 4, 8, 11, 14, 1, 4, …
 Blacklisted Channel
Blacklist: 4, 8
A: 1, 11, 14, 1, 11, 14, 1, 11, …
1
4
8
11
14
A
A
A
A
A
A
A
A
A
A
Dynamic Channel Selection
 Detecting Channel Conditions
 Multichannel Rendezvous with Blacklisted Channels
Detecting Channel Conditions
 Badness metric (which is a non-negative metric)
k
S: i
j
B
DATA
ACK
B
DATA
ACK
k
R: i
j
Both S and R decrease that channel’s badness metric by 1.
Detecting Channel Conditions
 Badness metric (which is a non-negative metric)
k
S: i
j
B
DATA
B
DATA
B
DATA
k
R: i
j
B
B
B
R increases that channel’s badness metric by 2.
R
S
R
Detecting Channel Conditions
S
 Badness metric (which is a non-negative metric)
k
S: i
j
B
DATA
B
DATA
DATA
DATA
k
R: i
j
ACK
DATA
ACK
DATA
ACK
S increases that channel’s badness metric by 2.
Detecting Channel Conditions
 Badness metric (which is a non-negative metric)
k
Random backoff
S: i
j
DATA
B
B
DATA
B
k
R: i
j
DATA
retransmission
B
DATA
R increases that channel’s badness metric by 2.
Detecting Channel Conditions
 Badness metric (which is a non-negative metric)
k
S:
Random backoff
i
j
DATA
B
B
DATA
B
k
R: i
j
DATA
retransmission
B
DATA
S increases that channel’s badness metric by 2.
ACK
Multichannel Rendezvous with
Blacklisted Channels
 For each node
 When the badness metric of a channel is above a threshold Cbad, the
channel will be added to the node’s blacklist.
 If the next pseudorandomly chosen channel is on the node’s
channel blacklist, the node stays on its current channel.
 The maximum number of channels allowed on a node’s blacklist is
the total number of available channels minus 1.
 To enable potential senders to learn its blacklisted channels, a node
R represents its blacklisted channels using a bitmap and embeds it
in its wake-up beacons.
How to address the rendezvous problem in an
asynchronous environment?
Adaptive Time Modeling
 tr=kts+b (y=kx+b)




ts: the current time of Sender
tr: the current time of Receiver
k: clock rate difference
b: initial clock difference
 If k = 1, we have tr= ts+b.
 S and R have the same clock rate but have a time difference of b.
Adaptive Time Modeling
1
B
DATA
ACK
S: 2
B
DATA
ACK
B
DATA
ACK
3
1
B
DATA
ACK
R: 2
3
1
tr=ts+b (assume
t r1  ktk=1)
s b
S : 2
t r  kts2  b
b=tr–ts
wake-up advance time
Exponential Chase Algorithm
k
B
DATA
ACK
B
DATA
ACK
S: i
j
k
R: i
j
B
B
Evaluation on MICAz Motes
 The proposed EM-MAC is implemented on MICAz motes running
TinyOS.
 Each wake-up interval for a node using EM-MAC was pseudorandomly
chosen between 500 ms and 1500 ms.
 On average, a sender generates one new data packet every second.
 Cbad is configured as 15.
 The sender wake-up advance time here is configured as 20 ms,
 The random number generator resetting interval is configured as 300 s.
Duty Cycles of Sender and Destination
Multichannel: McMAC and Y-MAC
Single Channel: PW-MAC, WiseMAC, RI-MAC, and X-MAC
Delivery Ratio and Delivery Latency
Multichannel: McMAC and Y-MAC
Single Channel: PW-MAC, WiseMAC, RI-MAC, and X-MAC
Performances of Rendezvous and with
Large Clock Rate Difference
Performance with Wireless Interference
and Jamming
Conclusion
 EM-MAC uses no control channel and enables a node to dynamically
select the channels it switches among based on the channel conditions it
senses.
 By effectively utilizing multiple orthogonal radio channels, EM-MAC is
able to avoid using channels that are currently heavily loaded or are
otherwise undesirable such as due to interference or jamming.
Thank You ~