Energy–efficient Reliable Broadcast in Underwater Acoustic Networks

Paolo Casari and Albert F Harris IIIUniversity of Padova, Italy

University of Illinois at Urbana-Champaign

Standard network primitive Routing protocols Reprogramming of nodes

Standard techniques Push method

Each node sends broadcast out upon receiving

Optimization techniques Reduce number of sending

nodes Challenge

Very expensive Energy consumption Time

Underwater channel Bandwidth challenged Delay challenged Energy challenged

Underwater Reliable Broadcast

Techniques Forward error correction (FEC)

Mitigate error rate Combined short link / long link

communication Minimize energy

consumption/delay Metrics

Energy consumption Broadcast completion time

Three Important Underwater Channel Characteristics

Bandwidth Distance dependent AN factor

Attenuation Noise

Transmission power Signal-to-noise requirement AN factor

Delay Location in water Salinity and temperature of water

Noise is frequency dependent Four common components

Turbulence Shipping Wind Thermal

Underwater Attenuation-Noise

)(log10log10),(log10 fallkflA

Absorption factor (frequency dependent as O(f2))

Spreading loss(k=2 for spherical)

Absorption loss

Attenuation is both distance and frequency dependent

Dominant for high frequencies

Dominant for low frequencies

Bandwidth-Distance Relationship

Find frequency center Frequency with

minimal attenuation given the distance

Find bandwidth 3 dB definition for

example

Both the frequency center AND the bandwidth vary with distance between nodes

Transmit Power

Signal-to-noise ratio (SNR) Related to

Bandwidth (B(l)) Attenuation (A(l,f)) Noise (N(f))

Calculate needed transmit power (W) Distance between nodes SNR threshold

)(

)(

1

)(

),()(

)(

lB

lB

dffN

dfflAlBW

lSNR

Knee in curve appears at < 3 km

Underwater Acoustic Propagation Speed

Speed c ≈ O(T3)+O(T2S)+O(z2) Temperature (T) Salinity (S) Depth in water (z)

T is dependent on z Value Rate of change

Average speed in water 1,500 m/s

Varies by 20 ms over a depth of 4 km

Consider nodes 1 km apart

Thermocline

Towards Broadcasting

Leverage underwater properties Turn challenges into benefits

Bandwidth-distance relationship

Use new “pull” model Reduce the number of redundant

transmissions Use FEC

Reduce the need for retransmissions

Simple Reliable Broadcast (SRB)

Standard push method protocol Node begins broadcast Upon receiving broadcast

Re-broadcast message If broadcast is received incomplete

Wait for timeout Potential for some other neighbor to transmit needed

packet

Send retransmission request to neighbors

Single-band Reliable Broadcast (SBRB)

Problem Short links

Reduced coverage Nodes fail to overhear

broadcast Long links

Expensive Increase contention in the

network

Solution: Pull method Using high-power, long

links for notifications Using low-power short

links for data

Upon receiving a complete broadcast message Transmit notification on

long link Wait for transmission

requests Upon receiving a broadcast

request message Nodes with complete

broadcast contend for channel

Winning node broadcasts, other go back to listen mode

Dual-band Reliable Broadcast

Idea Instead of sending wasted data for

notification on long link, make use of the bits

Works like SBRB, except FEC data is sent over long link instead of

notification

Evaluation

Baseline: Simple Reliable Broadcast Each node re-broadcasts

using low-power short links SRB, without FEC FSRB, with FEC

Generate random topologies 5 km x 5 km x 5 km network Control maximum closest

neighbor distance (varied between 100 m and 2 km)

Vary number of nodes between 40 and 700

Three protocols Single-band Reliable

Broadcast SBRB, without FEC FSBRB, with FEC

Dual-band Reliable Broadcast

Pull Method Saves Energy

For a large range of network densities, both energy and time to broadcast completion are minimized

Conclusions

Reliable broadcast Standard network primitive required by

protocols and applications Leverage channel properties

Reduce redundant transmissions Leverage FEC

Reduce retransmissions

Future Directions

Enhancements Add more intelligent FEC

Fountain-style codes Reduce initial number of transmissions

further

MAC and routing work Implementation and deployments

Testbeds

Thank You

Albert Harris III aharris@cs.uiuc.edu http://mobius.cs.uiuc.edu/~aharris/