1
Wireless Sensor Networks
Wireless Sensor Networks
Curt Schurgers
2
ECE 284
ECE 284
Wireless Sensor Networks
Wireless Sensor Networks
Networks of tiny autonomous data-gathering devices
Large-scale
●Dense local sampling, instead of distant more global view
●Small sensors have a limited range
Autonomous operation
●Network management and setup are intr actable for large-scale networks
●Self-configuration and adaptation
●Examples
Time synchronization
Self-localization
Mobile agents: instantiate functionality where needed
Data centric
●Interest in the data, not the sensor identity
●Tailor protocols to be data-centric rather than node-centric
Robust operation as a network
●Tiny cheap devices in often harsh environments
●Limited resources
UCB mote
The system is the
sensor network
3
ECE 284
ECE 284
Energy Efficiency
Energy Efficiency
Limited resources
●Energy and power: small form factor, thus small batteries and limited energy scavenging
possibilities
●Communication energy is important, since we are interested in the behavior of the entire
network
●Turning the radio off is one the main ways to be more energy efficient
Strategies: adaptation and application specific design
[CC2420] Lymberopoulos, D., A. Savvides, C.-C. Han, M. Srivastava, “XYZ: A motion-enabled power
aware sensor node platform for distributed sensor network applications,” IPSN’05 (SPOTS),2005.
[CC1000] http://etd.adm.unipi.it/theses/available/etd-05252004-154652/unrestricted/Chap4.pdf
[TR1000] A. Savvides, C.-C. Han, M. Srivastava, “Dynamic fine-grained localization in ad-hoc
networks of sensors,” MobiCom2001, Rome, Italy, pp. 166 –179, July 2001.
0
4
8
12
16
12.48 12.34
14.88
0.016
Tx Rx idle sleep
d ~2 0 meters, 2.4 kbps
(mW)
RFM TR1000
0
20
40
60
48 45
54
0.03
Tx Rx idle sleep
d ~ 50 meters, 38.4 kbps
ChipconCC1000
(mW)
0
20
40
60
80
65 65
59
0.9
Tx Rx idle sleep
d ~ 50 meters, 250 kbps
ChipconCC2420
(mW)
4
ECE 284
ECE 284
Localized Algorithms
Localized Algorithms
Centralized algorithms
●Global information
●Decisions are made in a central location and this information is propagated
●E.g. centralized routing scheme
Distributed algorithms
●Global information
●Decisions are made by the individual devices
●E.g. AODV
Localized algorithms
●Local information only
●Decisions are made by the individual devices
●E.g. geo-routing
Simulation of localized algorithms
●An infinite sensor field can be mimicked by ignoring statistics from ‘edge’ nodes
●The definition of ‘edge’ depends on how localized the algorithm is: e.g. only
information from 1-hop neighbors
edge
Scaling: ability for an algorithm to handle
increasingly large networks
●Localized > distributed > centralized
5
ECE 284
ECE 284
Network Density
Network Density
Possible definitions of network density
●Nodes per unit area
●Related to communication topology (number of neighbors)
Consider:
●Total network area A
●Abstract transmission range as R
●N nodes are distributed uniformly random
Probability
P(n)
of having
n
neighbors follows a binomial
distribution
For
N
→∞, keeping
λ
constant, this becomes a Poisson
distribution (with
λ
the average number of neighbors)
R
A
P(n)
λ
= 10
n
()( )
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛−
⋅−⋅= −−
n
N
PPnP nN
R
n
R
1
1)( 1
A
R
PR
2
⋅
=
π
A
R
N2
⋅
⋅=
π
λ
!
)( n
e
nP n
λ
λ
−
⋅
=
6
ECE 284
ECE 284
Network Connectivity
Network Connectivity
What density guarantees a fully connected
network, i.e. every node has a path to every
other node?
From simulations:
The critical density
λfully
for a fully connected
network can be approximated as:
λ
−
−=−= ePPconn 1)0(1
(
)
λ
−
⋅−≈= eNPP N
connpartly 1
(
)
(
)
partlypartly PN
−
−
=
1lnln
λ
(
)
(
)
partlyfully PP
−
⋅
≈
−
11
ε
e
≤
≤
ε
1
1
ˆˆ +≤≤
λλλ
fully
(
)
(
)
fully
PN −−= 1lnln
ˆ
λ
N = 10
N = 100
N = 1000
N = 10,000
N = 100,000
N = 1,000,000
λ
1 - Ppartly
Prob. of a node having
at least one neighbor
Prob. of each node having
at least one neighbor
Required density for a
given prob. of each node
having one neighbor
Prob. of at least one
isolated node
Pfully: Prob. of having a
fully connected network
Note: theoretical bound [Gup02]:
(
)
)(ln NcN
fully
+
=
λ
∞=⇔= ∞→∞→ )(lim1lim NcP N
fully
N
