Communication
through Silence in Wireless
Sensor Networks
Puzzle










Logic class of 15 students
First day, teacher poses a challenge
Students will be isolated in 15 different rooms
Switch-Room with 2 switches (not connected to anything)
Teacher will pick one student at random, walk student to switch-room,
have student flip one switch, and bring back student to isolation room
Teacher will then pick another student at random, and repeat the process
Students cannot communicate with each other in any way
Both switches initially OFF
If students can figure out when ALL students have visited switch room a
nd tell the teacher, students get the entire semester off with an A grade!
If not (they don’t tell or they infer incorrectly), they will have to do the
class
2
Energy Based Transmission (EbT)

Traditional communication strategy that conveys information
between the sender and the receiver using energy only

Procedure:





Sender interprets the information to be transmitted as a bit stream
Sender transmits the bit stream
Receiver receives the bit stream
Information delivered!
The energy consumption for EbT is keb where k is the length of
the bit stream and eb is the amount of energy used to transmit
one bit
97
97 !
Example: 1 0 0 0 0 1 1
1 0 0 0 0 1 1
1 0 0 0 0 1 1
It takes 20eb for EbT to send a raw data packet of 20 bits without
considering the overhead
3
Communication through Silence (CtS)
 A new communication strategy that conveys information using silent
periods in tandem with small amount of energy
 Procedure:
Sender interprets the information to be transmitted as a data value
Sender transmits start signal
Receiver starts to count from 0 upon receiving start signal
Sender and receiver are synchronized in counting clock
Sender transmits stop signal when receiver counts up to the desired
value
 Information delivered!





 The energy consumption for CtS is always 2eb irrespective of the
value being sent.
96
97!
2
1
97
95
….
3
97
Example:
STOP
START
It takes 2eb for CtS to send a raw data packet of 20 bits without
considering the overhead
 10X improvement in energy consumption!
4
Energy - Throughput Tradeoff
0.6
0.3
0.5
0.25
Throughput Ratio
Ratio of Bits Transmitted
The CtS strategy incurs exponential throughput decrease compared to EbT
0.4
0.3
0.2
0.2
0.15
0.1
0.05
0.1
0
0
3
5
7
9
11
13
3
5
Data Frame Size (bit)
7
9
11
13
Data Frame Size (bit)
The energy consumed for
The time taken to transmit a packet
transmitting each bit of data
of s bits is 2s using Cts
 The throughput of CtS
decreases inverse linearly with
data frame size
decreases as s/ 2s
Example:10Mbps data rate, 10 bit data frame size
Energy improvement: 5 times
Effective throughput: 10Kbps
5
Optimization Strategies Overview
 Exploit the unique characteristics of CtS to improve
upon the considerably low throughput performance of
basic CtS.
 Assumptions:





The start and stop signals are uniquely addressable
Each start/stop signals occupies one bit
The communication channel is lossless
The sender and the receiver clocks are perfectly synchronized
There are no counting errors
6
Multiplexing
 Two or more contending sender-receiver pairs can transmit at the
same time as along as the start/stop signals do not overlap
10 time slots!
Basic CtS
4
A-B
start
A-B
stop
C-D
start
2
C-D
stop
6 time slots !
CtS with
Multiplexing
2
A-B
start
C-D
start
C-D
stop
A-B
stop
 Enabled by the typical long silent intervals between start and stop
signals in CtS
 Not possible in EbT since a sender-receiver pair has to occupy the
channel exclusively during transmissions
7
Cascading
 Consecutive data values that are monotonically increasing or
decreasing can be sent in a combined “signal train” consist of one
start signal, multiple intermediate signals and one stop signal instead
of multiple consecutive start/stop signals
12 time slots!
Basic CtS
A-B
start 1
3
A-B
stop1
5
A-B
start 2
A-B
stop 2
7 time slots !
CtS with
Cascading
A-B
start
3
A-B
interme
diate
A-B
stop
 The same counting process can be used for multiple packets at the
same time for CtS
 Not possible in EbT since overlapped data packets confuse the
receiver
8
Fast-forwarding
 For multihop CtS communication, a relay node can forward start/
intermediate signals before the stop signal is received
18 timeslots !
Basic CtS
[t7 t11]
[t1 t5]
S
[t13 t17]
[t2 t6]
D
R2
R1
[t8 t12]
[t14 t18]
10 timeslots !
CtS with
Fastforwarding
[t1 t5]
[t3 t7]
S
[t5 t9]
[t2 t6]
D
R2
R1
[t4 t8]
[t6 t10]
 Enabled by the separable signals of each packet and possibility of
simultaneous counting in CtS
 Not possible in EbT because a packet has to be received in full before
forwarding to ensure content integrity
9
Integrated Operation and Summary of Benefits
0.9
Throughput (bit/bit slot)
0.8
0.7
0.6
CtSmfc
EbT
0.5
0.4
0.3
0.2
0.1
0
0
10
20
30
40
Path Length (hop)
50
60
7000
6000
Energy (Eb/bit)
 A multi-hop sensor network
Summary:
 Sink located at the center of
The energy-delay trade-off of basic
the network
CtS strategy
can be alleviated by
Average
degree = 10
  Proper
combination
of
 optimization
A randomlystrategies
chosen sensor
 Adaptation
sends dataoftoparameters
sink via an
h-hop path
5000
CtSmfc
EbT
4000
3000
2000
1000
0
0
10
20
30
40
50
60
Path Length (hop)
 The throughput of integrated
CtS is up to twice that of
EbT
 The energy consumption of
EbT increases faster with
path length than integrated
CtS
10
Challenges Overview
Appropriate protocols tailored to the CtS paradigm from the
physical layer to MAC layer, and possibly higher layers
PHY layer requirements:
Synchronization overhead for
carrier based transmission
Baseband modulation
Limited energy supply and low
transciever complexity
Impulse radio
Addresses embedded in start/stop
signals
Pulse train technology
11
Challenge: Framing (1)
 Determining the length of the messages the transmitter will send to
the receiver
 The length of the CtS frame determines the amount of delay taken for
the transfer of that information.
 Issues:
 Energy throughput tradeoff
• Should be small enough such that the delay required for transmission is not
prohibitive:
e.g. A CtS packet size of 100 bytes incurs delay as high as 2800 bit slots!
• Should be large enough to ensure acceptable energy saving
 Contention Resolution
• Large frame size reduces the chances of collisions when simple contention
resolution mechanisms are employed
12
Challenge: Framing (2)
 A simple solution:
 Set the CtS delay frame size to be 256-65536 bit slot, which
translates to a CtS data frame size of 8-16 bits, based on
empirical evaluation
 Translates into 10Kbps (4X energy improvement) and 100bps (
8X energy improvement) respectively for a 10Mbps
raw data rate network.
 Drawbacks of this solution:
 The optimal frame size is yet to be determined within the
given range
 Need an adaptive scheme to vary the frame size according to
network conditions
13
Challenge: Addressing (1)
 Identify the sender and the intended receiver in a CtS frame
 For a shared media, a node needs to know if the received packet
is for itself, and where this packet comes from
 Given the limited CtS frame size, embedding the sender and
receiver ID in the frame will be too much an overhead
 Issues:
 How to identify the sender and receiver in an efficient way
 How to reduce the addressing overhead
14
Challenge: Addressing (2)
 A simple solution:
 Local addressing scheme
• Record the original global source and destination addresses only in the
“regular” link layer frame handled by CtS
• In the CtS frames, locally computed addresses that distinguish only
between sensors in a neighborhood will be encoded
• Minimum coloring problem
 Modulate signals with the address information
• Impulse radio
• Pulse train
• Translate the code into pulse-shift or phase shift pattern
 Drawbacks:
 Global coordination is required to find the minimum coloring number
 Update of local IDs may be needed to accommodate network
dynamics
15
Challenges: Sequencing

Providing a way for the receiver to determine the order of CtS frames
received

Enable the receiver to reconstruct higher layer frames in the presence
of CtS frames reordering

Issues:
 Sequence number results in significant overhead in CtS frames

A Simple solution:
 Deliver all the CtS frames belonging to the same higher layer frame back to back:
 no sequence number required!

Drawbacks:
 Introduce extra delay when CtS frame is being retransmitted
 Any completeness check can be performed only after the arrival of all CtS
frames
16
Challenge: Error Control (1)
 Preventing errors in the delivery of CtS frames
 Detecting errors and/or recover from errors in CtS frames
 Retransmission of CtS frames aggravate the already low CtS
throughput
 Issues
 Correctness of CtS requires:
• Precise delivery of the start/stop/intermediate signals
• Perfect synchronization of sender-receiver counting clocks
 Traditional error control techniques may not be applicable
• Applying error control coding to each CtS frame incurs high overhead
17
Challenge: Error Control (2)
 A Simple solution:
 Rely solely on error control only at the granularity of
the “regular” link layer frames
 Drawbacks:
 One erroneous CtS frame causes the corruption and
retransmission of the entire higher layer frame
 Extra coding may be required to identify the
erroneous CtS frame in a higher link layer frame
18
Challenge: Contention Resolution (1)
 Resolve contentions when the communication channel is shared
by multiple CtS senders and receivers
 Traditional contention resolution schemes can be highly resource
intensive for CtS
 Issues
 Only need to avoid overlapping of start/stop signals
 Approaches such as carrier-sensing are not applicable due to the
short duration of signals transmitted
 Exchange of control packets for contention resolution causes high
overhead
19
Challenge: Contention Resolution (2)
 A simple solution: ALOHA scheme
 Inherent low channel access probability of CtS
 ALOHA works best when the network load is less
than 20%
 Collisions of start/stop signals can be detected by
error control scheme and trigger retransmission
 Drawbacks:
 Not adaptive to network load
 Relies on error control strategies to detect errors
20
Other challenges
 Design effective PHY layer solutions tailored to CtS MAC l
ayer strategies and specific application environments
 Design effective routing layer solutions that reduce
control packet overhead and leverage the characteristics
of CtS
 Other brand new solutions may be needed for CtS:





Topology control
Synchronization
Sleep scheduling
Adaptive transmission strategies
And more ..
21
Related work & Conclusions
 Related work
 Timing channel
• Used for covert communication
• Intervals between data packets are interpreted as codes in an alphabet
• Not optimized for energy and throughput
 Digital Pulse Interval Modulation
• A modulation scheme
• Use intervals between pulses to encode information
• Typical frame length is 3-5 bits
 Conclusions
 Introduced the new paradigm of Communication through Silence for
wireless sensor networks
 Identified the energy-throughput trade-off of CtS
 Presented unique optimization strategies to improve upon the
throughput performance of CtS
 Discussed several research challenges related to the realization of CtS
22
Puzzle
 Teacher comes to class on Friday and tells stu
dents: “I will give you a surprise exam next we
ek”
 Class meets on M-F every day 9-10am
 Can the teacher accomplish giving a truly surp
rise exam?
23
The end!
Questions?
24