An Application-Specific Protocol Architecture for Wireless ... - MIT

An Application-Specific Protocol Architecture for Wireless Microsensor Networks Wendi Heinzelman Anantha Chandrakasan and Hari Balakrishnan Massachusetts Institute of Technology

Microsensor Networks B

C

f(A-G)

D

A G

F

E Base station

• Remote monitoring of the environment – Surveillance – Machine diagnostics

• Relevant parameters: – System lifetime (energy efficiency) – Quality – Latency

The MIT µ-AMPS Project • Develop energy-optimized solutions – Node architecture and radio hardware – OS, algorithms and protocols

• Assumptions: – Base station far from nodes – All nodes energy-constrained – Locally, data correlated Goal: Design protocols for high quality and energy and spectrum efficiency

Application-Specific Architecture Application Transport Network MAC Data-link Physical

General-Purpose Protocol Architectures • Direct transmission – Power scales as Rn – Bandwidth limitations

• Routing

R

A

– MTE (Min. Trans. Energy) Routing B C – Multi-step communication A’B’C if: dAB2+dBC2 µE eligible

Simulation Framework • Extensions to ns network simulator – Computation and communication energy models • Radio energy k bit packet

Eelec* k Transceiver

ε amp* k *

d2

Tx Amplifier

Transmitter

d Eelec* k

k bit packet

Transceiver

Receiver

• StrongARM processor beamforming [A. Wang]

– New node states – Medium access – LEACH, LEACH-C, MTE routing, static clustering

Simulation Parameters Base Station

100 meters

75 meters

Processing delay

50 µs

Bit rate

100 kbps

Radio electronics

50 nJ/bit

Transmit amplifier 100 pJ/bit/m 2 Beamforming cost 5 nJ/bit/signal Data size 100 meters

100 nodes

500 bytes

Average energy per round (J)

Optimum Number of Clusters k=11

k=9

k=1 k=7 k=3 k=5

Number of clusters (k)

• Too few clusters a cluster-head nodes far •

from sensors Too many clusters a not enough local signal processing

Analytical Optimum k clusters _ N/k nodes/cluster: ECH

N =α +β k

E non− CH

kopt =

1 =γ +δ k

N M 6 d toBS

dtoBS d2toCH α 1/k

N=100 M=100 75 < dtoBS < 185

• Simulation agrees with theory

_ 2 < kopt < 6

Nodes begin with limited energy

Amount of data received at BS

Data per Unit Energy LEACH-C

StaticClustering

LEACH

MTE Energy Dissipation (J)

• LEACH achieves order of magnitude more data per unit energy – 2 hops v. 10 hops average – Data aggregation successful

Number of nodes alive

Network Lifetime

StaticClustering

LEACH-C

LEACH MTE

Amount of data received at BS

• LEACH delivers over 10 times amount of •

data for any number of node deaths Rotating cluster-head effective

Summary • Microsensor network protocols must be designed for – Bandwidth efficiency – Energy efficiency – High quality

• Application-specific protocol architecture beneficial – Most efficient use of limited resources – Ideas extendable to other application spaces