Ad hoc Sensor Networks

(1)

Development of Complex Curricula for Molecular Bionics and Infobionics Programs within a consortial* framework**

Consortium leader

PETER PAZMANY CATHOLIC UNIVERSITY

Consortium members

SEMMELWEIS UNIVERSITY, DIALOG CAMPUS PUBLISHER

The Project has been realised with the support of the European Union and has been co-financed by the European Social Fund ***

**Molekuláris bionika és Infobionika Szakok tananyagának komplex fejlesztése konzorciumi keretben

***A projekt az Európai Unió támogatásával, az Európai Szociális Alap társfinanszírozásával valósul meg.

(2)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 2

Ad hoc Sensor Networks

Communication protocols for wireless sensor networks

Érzékelő mobilhálózatok

Kommunikációs protokollok vezeték nélküli érzékelő hálózatok számára

Dr. Levendovszky János

(3)

Lecture 8 review

• The IEEE 802 Family of Standards

• Cellular Systems (recall)

• Wireless LANs alias Wifi (recall)

• Bluetooth

• ZigBee

• Summary of

– 3G standards

– WLAN standards – WPAN standards

(4)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 4

Outline

• Technological motivations

• The role of routing

• Operation of the sensor networks

• Lifespan estimation

• Packet forwarding in WSNs

– Traditional protocols

– Overall Energy Reliable Algorithm (OERA) – Bottleneck Energy Reliable Algorithm (BERA)

• Numerical results

(5)

Sensor networks: data collection and distribution at a global scale

Challenges:

– Limited energy resources (limited life span)

– Limited computational capabilities – Dynamic topology

Motivations

Traditional protocols are not applicable!

? !

Protocol optimization to maximize life span, we need

energy aware protocols !!

(6)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 6

Network structure and packet forwarding in WSN

(7)

Protocol design

Comm. protocol (packet forwarding

algorithm)

Energy- and localization

parameters Statistical traffic models

MAX LIFETIME

Result: So far results only for the case of deterministic traffic load (see Haenggi, Johansson …etc.) ! Extension to statistical load !

Methodology: Large deviation theory, graph

theory and combinatorial optimization

What is inside this box ???

(our contribution)

QoS requirements (reliabilty, delay, etc.)

(8)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 8

Routing performance metrics

• Energy Efficiency/Lifetime: The lifetime can either be defined as the time till the first node goes flat or as the time till a pre- defined κ portion of the nodes go flat. Furthermore, the performance of a protocol can be quantified by the overall remaining energy when the network goes flat. If this remaining energy is high than the protocol did not perform well, as a lot of energy is wasted.

• Reliability: it determines the probability of successful reception of the packets at the BS.

• Latency (or Delay): the time interval between sensing and data

collection at the BS. In the case of multihop routing it is often

associated with the number of hops needed to reach the BS.

(9)

Operational of the sensor networks

• The WSN contains nodes which communicate each other by radio communication. Information is sent in the forms of packets and the task of the routing protocol is to ensure reliable packet transfer to the BS. We assume the following properties:

– there is only a single, stationary BS on a fixed location (in certain applications there can more than one BS with mobility);

– the BS is energy abundant as it can be recharged or connected to an energy supply network, and, as a result, the BS is able to reach each node of the WSN (even the furthest ones) by radio communication;

– the nodes are also assumed to be stationary;

– there may be some nodes which do not have enough power to reach the BS directly, hence multihop packet transfer is in use, where packet forwarding is determined by an addressing mechanism;

(10)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 10

Operational of the sensor networks (cont’)

– the Medium Access Control (MAC) interface discovers the neighbors and measures the LQI (Link Quality Indicator) to the neighbors;

– the direction of communication is Node-to-BS (the data acquired by the nodes must be collected by the BS)

– if necessary, the nodes can organize themselves into a hierarchy where a node at a given level of the hierarchy receives packets from nodes at a lower level of the hierarchy;

– the reporting model is either Query Driven (the BS requests data from a certain node) or Event Driven (the node sends a packet if an event occurs, e.g. the measured temperature exceeds a certain threshold);

(11)

Operational of the sensor networks (cont’)

– the radio propagation is described by the Rayleigh model, implying that the reliability of each packet hop (the probability of correct packet reception over a single hop) is

– where θ is the receiver sensitivity threshold, γ is the exponent of fading attenuation and σ² denotes the energy of noise. One must note that this formula connects the reliability of packet transfer P^(r) over distance d with the required energy g (for the sake of notational simplicity this relationship will be denoted by P^(r) =Ψ(g).

• Based on the discussion above, the WSN is perceived as a 2D graph G(V,E,d), where V represents the set of wireless sensors E represents the radio links between node i and j operating with reliability P

_ij

and the distance d

_ij

from node i to j is covered by a transmission energy g

_ij

.

( )^r exp d ²

P g

γ σ

− Θ 

=  

 

(12)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 12

The model: WSN as a 2D graph

(13)

The model: packet forwarding over a 1D or linear model

From a 2D graph we often change to a 1D model, i.e. after the routing protocol has found the path to the BS the nodes participating in the packet transfer can be regarded as a one dimensional chain labeled by i = 1, ...,N.

Traffic state vector:

(14)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 14

Notations for the 1D model

• The topology is uniquely defined by a distance vector d=(d₁, ...,d_N) where d_i, i= 1, ...,N denotes the distance between node i and i−1, respectively.

• The initial battery power on each node is the same and denoted by C.

• The packets are generated in discrete time instant and the corresponding time variable is denoted by k.

• We assume that each node generates packets subject to an On/Off model, i.e.

packet generation occurs with probability P(y_i = 1) = p_i, whereas the node does not generate packet with probability P(y_i = 0) = 1− p_i.

• The traffic state of the network is represented by an N dimensional binary vector yє{0,1}^N and the corresponding probability of a traffic state is given as

assuming independence among the sensed quantities.

( )

¹

1

1 ⁱ

i

N y y

i i

i

p p ⁻

=

∏

−

(15)

Packet forwarding mechanisms

• Chain protocol: Each node transmits packet to its neighbor lying closer to the BS. In this way, each node consumes minimal energy being engaged with short range energy transmission. However, as each packet will go through the node being closest to the BS, the lifetime of this node is likely to determine the longevity of the whole network.

• Random shortcut protocol: Upon receiving or generating a packet node i can choose to forward the packet to its neighboring node being closer to the BS (labelled as i−1) with probability (1−a_i), or directly send the packet to the BS with probability a_i. In this way paths are randomized and not every packet will go through the node closest to the BS. As a result, a better energy balancing is expected.

• Single-hop protocol: Each node sends its packet directly to the BS. In this case, the farthest node is likely to determine the longevity of the network.

(16)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 16

Network characterization

Energy vector:

^g ⁼ ( ^g

¹

^,..., ^g

^N

)

Traffic stat. Vector:

^p ⁼ ( ^p

¹

^,..., ^p

^N

)

Initial energy C

Cost function f(y) (related to energy cons.) WSN network

( )

{ }

( )

: ( ) 1,1

( ) ( ) ( ); P ( ) ( ) ( )

N g C

E f f p f C f p

∈ − >

=

∑

> =

∑

y y y

y y y y y y

( )

²^N

O complexity large deviation theory

Network lifetime?

(17)

Lifespan estimation by using the Chernoff bound

• Let us assume that the chain protocol is in effect. The energy consumed by sending a packet generated on node i to the BS is given as

• and the average energy consumption up to time instant K is given as

• The lifespan of node is defined as

where α is a given reliability parameter of lifetime.

1 i

i j

j

G g

=

∑

( )

1 1

K N

i i

k j

G y k

= =

∑∑

α

1 1

: 1

K N

i i

k i

K P y G C e

N

−

= =

 

< =

 

 ∑ ∑

^ɶ



ɶ

(18)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 18

Lifespan estimation by using the Chernoff bound (cont’)

• By using the complementary probability:

• lifetime evaluation is cast as a tail estimation problem, where bounds like the Chernoff inequality can be used as follows:

where µ is the logmoment generating function:

and

α

1 1

K N

i i

k i

P y G C e

N

−

= =

 

> = −

 



∑ ∑

^ɶ 

(

^*

)

^*

1 1 1

1 exp µ ,

K N N

i i i i

k i i

P y G C s G s NC

N K

= = =

 

> ≤  − 

 



∑ ∑

^ɶ  

∑



ɶ

(

^,

)

^{: log}

( { }

^{sy G}ⁱ ⁱ

)

^{log 1}

(

^sGⁱ

)

i s Gi E e pi p ei

µ = = − +

( )

*

1

: arg min ,

N

i i

s i

s K s G sNC

µ K

=

 

−

 

 ^ɶ

∑

_ɶ 

(19)

Lifespan estimation by using the Chernoff bound (cont’)

• By using the estimation, one obtains

and the lifespan of the simple chain protocol can finally be estimated by the following formula:

(

^*

)

^*

1

exp , 1

N

i i

i

s G s NC e

K

µ ⁻α

=

 

− = −

 



∑

_ɶ 

( ) ( )

*

* 1

, log 1

N

i i

i

K s NC

s G e ^α

µ ⁻

=

− −

∑

ɶ

(20)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 20

Numerical results

C=100000

d_max=20 m

equidistant

(21)

Numerical results (cont’)

C=100000

d_max=20 m

equidistant

N=7

(22)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 22

Packet forwarding mechanisms: new challenges

• Energy plays a central role (nodes are with limited, non- rechargeable energy resources)

• Only the BS has unlimited energy resources

• Since the nodes have limited processing capabilities thus routing

can be done in a centralized manner by the BS and after setting

up the paths the nodes get notified by the BS about where to

forward packets

(23)

Packet forwarding mechanisms: traditional protocols

• Low-energy adaptive clustering hierarchy (LEACH): this protocol is a hierarchical routing algorithm which groups the nodes into clusters based on their energy levels. Each cluster elects a clusterhead by random selection, which collects the packets in the cluster and then sends them to the BS. Data fusion and aggregation takes place in the cluster. This protocol can considerably increase the lifetime. On the other hand, the dynamic selection of clusterheads and the need for announcing the results periodically entails a huge overhead.

• Directed Diffusion (DD): it is one of the most important data-

centric routing protocols. It uses attribute-value pairs for the

data and queries the sensors. Data aggregation in the network is

achieved by solving a minimum Steiner tree problem.

(24)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 24

Traditional protocols (cont’)

• Power-Efficient Gathering in Sensor Information Systems

(Pegasis): this protocol determines a chain of nodes instead of

clusters. Each node has only two neighbors. The path is

constructed by a greedy algorithm. At the end of the path is the

furthest node from the BS. Its neighbor is going to be the next

element in the path and these steps are repeated until we get to

the BS. Each node in the path aggregates the information

obtained from its neighbor and forward it towards an elected

leader in the chain which communicates the BS directly. The

leader is elected randomly in each round.

(25)

Traditional protocols (cont’)

• Power Efficient Data gathering and Aggregation Protocol (PEDAP): it calculates a minimum spanning tree by the Prim algorithm having the BS as the root node. The cost of a link in the case of a k-bit packet is

where E

_elec

is the transceiver consumption and E

_amp

denotes the transmitter energy, while d

_ij

is the distance between node i and node j.

Problem: a given probability of successful packet arrival is not guaranteed !!!

( ) ²

^elec ^amp ²

ij ij

g k = ⋅ E ⋅ + k E ⋅ ⋅ k d

(26)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 26

Overall Energy Reliability Algorithm(OERA)

• Objective: minimize the overall energy of the packet transfers subject to the energy constraints. (Constrained optimization)

{ }

{

^{1 2} ^{1 2}² ³ ¹

}

¹

opt opt

, ,....,

1 , ,....,

, : min

L l l

i i i i i iLL

L i i i i i

l

g g g

g

+ +

ℜ= =

ℑ=

ℜ ℑ ∑

( ) ( )

^opt¹

1

correct reception at BS (1 )

l l

i

i i j

P g

₊

ε

=

= ∏ Ψ ≥ − subject to

Reliability constraint

(27)

Overall Energy Reliability Algorithm(OERA) (cont’)

• Theorem: Relability constrainted energy efficient routing can be performed in polynomial time by the Bellman- Ford algorithm, where

– the optimal energy:

– the redefined shortest path problem:

where

opt , 1

1

: min

l l

L

i i l

w +

ℜ =

ℜ

∑

( )

1

opt

1 2 ...

i il l L l

g w w w w

+ = + + + ⋅

(

¹

)

1

2 ,

, ln 1

k k

i i i i

w d

γ σ

+ ε

+

= Θ

− −

(28)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 28

Botlleneck Energy Reliability Algorithm(BERA)

• Open problem: So far we have only optimized the overall energy needed to transfer a packet from destination to the Base Station. The current energy state of the WSN has not yet been taken into account

• Different objective: maximize the minimum residual energy level of the bottlenek node (make the energy distribution uniform over the network) under the reliability constraints.

{1 2 }

{ (

¹

) }

opt

: max

, ,....,

min ( )

l l l

L

i i i

i i i l

C k G

ℜ= +

ℜ −

residual energy after packet transfer

(29)

Numerical results: topologies

Random topology with 100 sensor nodes.

Grid topology with 100 sensor nodes.

(30)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 30

Numerical results: a comparative analysis of the lifespan

Time to the earliest death. Time to the latest death.

(31)

Numerical results: a comparative analysis of the lifespan

(α)

(32)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 32

Numerical results: load balancing

(33)

Numerical results: load balancing (cont’)

(34)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 34

Numerical results: delay

(35)

Numerical results: impact of fading parameter est.

γ_est

(36)

2011.11.27.. TÁMOP – 4.1.2-08/2/A/KMR-2009-0006 36

• The Chernoff bound based method can be extended to estimate the lifespan of any arbitrary protocol.

• Traditional algorithms do not provide reliable packet forwarding

• Our solution for minimum energy routing with reliability constraint (OERA and BERA) can be mapped into additive graph optimization problem thus solution is available in polynomial time by the BF algorithm

Ad hoc Sensor Networks

Ad hoc Sensor Networks

Communication protocols for wireless sensor networks

Dr. Levendovszky János

Lecture 8 review

• The IEEE 802 Family of Standards

• Cellular Systems (recall)

• Wireless LANs alias Wifi (recall)

• Bluetooth

• ZigBee

• Summary of

Outline

• Technological motivations

• The role of routing

• Operation of the sensor networks

• Lifespan estimation

• Packet forwarding in WSNs

• Numerical results

Motivations

? !

Network structure and packet forwarding in WSN

Protocol design

Comm. protocol (packet forwarding

algorithm)

Routing performance metrics

• Reliability: it determines the probability of successful reception of the packets at the BS.

• Latency (or Delay): the time interval between sensing and data

collection at the BS. In the case of multihop routing it is often

associated with the number of hops needed to reach the BS.

Operational of the sensor networks

• The WSN contains nodes which communicate each other by radio communication. Information is sent in the forms of packets and the task of the routing protocol is to ensure reliable packet transfer to the BS. We assume the following properties:

Operational of the sensor networks (cont’)

Operational of the sensor networks (cont’)

• Based on the discussion above, the WSN is perceived as a 2D graph G(V,E,d), where V represents the set of wireless sensors E represents the radio links between node i and j operating with reliability P

and the distance d

from node i to j is covered by a transmission energy g

.

The model: WSN as a 2D graph

The model: packet forwarding over a 1D or linear model

From a 2D graph we often change to a 1D model, i.e. after the routing protocol has found the path to the BS the nodes participating in the packet transfer can be regarded as a one dimensional chain labeled by i = 1, ...,N.

Notations for the 1D model

( )

∏

Packet forwarding mechanisms

Network characterization

g = ( g

,..., g

)

p = ( p

,..., p

)

( )

( )

∑

∑

( )

Lifespan estimation by using the Chernoff bound

∑

( )

∑∑

: 1

K P y G C e

N

 

< =

 

 ∑ ∑



ɶ

Lifespan estimation by using the Chernoff bound (cont’)

∑ ∑

(

)

∑ ∑

∑

(

)

( { }

)

(

^g ⁼ ( ^g

^,..., ^g

^p ⁼ ( ^p

^,..., ^p

( ) ²