Hybrid-ARQ Based Intra-Cluster Geographic Relaying

1
Hybrid-ARQ BasedIntra-ClusterGeographic Relaying

• Matthew Valenti, Ph.D.
• Assistant Professor
• West Virginia University
• Morgantown, WV
• mvalenti_at_wvu.edu
• This work was supported by the Office of Naval
  Research
• under grant N00014-00-0655

Bin Zhao, Ph.D. Efficient Channel Coding Brooklyn
Heights, OH bzhao_at_eccincorp.com
2
Problem Statement

• Consider the following ad hoc network
• Questions
• How can the message be efficiently routed to the
  destination?
• What is the tradeoff between latency and energy
  efficiency?
• How to jointly implement error control, routing,
  and access control?
• Joint (cross-layer) solution is emphasized.

Source
Destination
Relays
3
Conventional Hybrid ARQ

• Consider a point-to-point link
• Hybrid-ARQ using incremental redundancy
• The data is encoded into a rate RM mother code
• Implemented using rate-compatible puncturing.
• Break the codeword into M distinct blocks
• Each block has rate R MRM
• Source begins by sending the first block.
• If destination does not signal with an ACK, the
  next block is sent.
• Process continues until source receives an ACK or
  all blocks sent.
• After mth transmission, effective rate is Rm
  R/m

Source
Destination
4
Info Theory of Hybrid-ARQ

• Throughput of hybrid-ARQ over block fading
  channels has been studied by Caire and Tuninetti
  (IT 2001).
• Let ?m denote the received SNR during the mth
  transmission
• The instantaneous capacity (mutual information)
  is
• The cumulative capacity is
• An outage occurs if

5
Conventional Approach Multihop

• Multihop picks from among several possible
  routes
• Creates the route from a cascade of
  point-to-point links
• Each point-to-point link could use hybrid-ARQ
• Drawbacks
• Routing tables need to be created and maintained.
• Not robust to changes in topology, interference,
  or channel.
• Routing ultimately relies on cascade of
  point-point links.
• Need to keep retrying over bad links.
• Spatial (MIMO) diversity not exploited.
• Wireless is broadcast-oriented, not
  link-oriented!
• The network could instead be interpreted as a
  large distributed array.

Source
Destination
Relays
6
Generalized Hybrid-ARQ

• Now consider a multi-terminal network
• Suppose the source attempts to communicate with
  the destination using hybrid-ARQ.
• After each ARQ transmission, some of the
  intermediate nodes could overhear the
  transmissions.
• Overhearing nodes that correctly decode could
  serve as relays.
• The ARQ retransmission could come from a relay
  instead of the source.
• Decode and forward relaying.

Source
Destination
Relays
7
HARBINGER

• Source broadcasts first packet, m1.
• Relays that can decode are added to the decoding
  set D.
• The source is also in D
• The next packet is sent by a node in D.
• The choice of which node depends on the protocol.
• Geographic-Relaying Pick the node in D closest
  to destination.
• The process continues until the destination can
  decode.
• We term this protocol HARBINGER
• Hybrid ARq-Based INtercluster GEographic
  Relaying.
• Energy-latency tradeoff can be analyzed by
  generalizing Caire and Tuninettis analysis.

8
HARBINGER Initialization
Source
Destination
Solid circles are in the decoding set D. Amount
of fill is proportional to the accumulated
entropy. Keep transmitting until Destination is
in D.
9
HARBINGER First Hop
Source
Destination
hop I
Solid circles are in the decoding set D. Amount
of fill is proportional to the accumulated
entropy. Keep transmitting until Destination is
in D.
10
HARBINGER Selecting theRelay for the Second Hop
Source
Destination
hop I
ACK /CTS
contention period
11
HARBINGER Second Hop
Source
Destination
Relay
hop II
Solid circles are in the decoding set D. Amount
of fill is proportional to the accumulated
entropy. Keep transmitting until Destination is
in D.
12
HARBINGER Third Hop
Relay
Source
Destination
hop IV
Solid circles are in the decoding set D. Amount
of fill is proportional to the accumulated
entropy. Keep transmitting until Destination is
in D.
13
HARBINGER Fourth Hop
Relay
Source
Destination
hop III
Solid circles are in the decoding set D. Amount
of fill is proportional to the accumulated
entropy. Keep transmitting until Destination is
in D.
14
HARBINGER Results
Topology Relays on straight line S-D separated
by 10 m Coding parameters Per-block rate R1 No
limit on M Code Combining Channel parameters n
3 path loss exponent 2.4 GHz Rayleigh block
fading d0 1 m reference dist Monte Carlo
Integration
B. Zhao and M. C. Valenti. A block-fading
perspective on energy efficient random access
relay networks, to appear in JSAC special issue
on Wireless Ad Hoc Networks.
15
Discussion

• Advantages.
• Better energy-latency tradeoff than multihop.
• Nodes can transmit with significantly lower
  energy.
• System exploits momentarily good links to reduce
  delay.
• No need to maintain routing tables (reactive).
• Disadvantages.
• More receivers must listen to each broadcast.
• Reception consumes energy.
• Nodes within a cluster must remain quiet.
• Longer contention period in the MAC protocol.
• Results are intractable, must resort to
  simulation.
• Requires position estimates.
• These tradeoffs can be balanced by properly
  selecting the number of relays in a cluster.

16
Simplifying Assumptions

• Closed-form analysis is not tractable.
• Statistically variable channels.
• Nodes have memory for entire source-destination
  transaction.
• Possible changes in topology.
• Nodes could cycle on-and-off according to a sleep
  schedule.
• Analysis is possible under simplifying
  assumptions
• Channels are non-faded (AWGN).
• Nodes flush memory once a new relay is selected.
• Still maintain memory of ARQ packets from current
  transmitter.
• Topology is 2-D Poisson.
• Nodes cycle on-and-off according to a sleep
  schedule.

17
Versions of HARBINGER

• Consider a network with nodes that cycle on and
  off.
• Network coherence time time nodes are awake.
• Two main versions
• Fast HARBINGER
• After each ARQ transmission, nodes cycle in/out
  of sleep state.
• Coherence time ?? 1 block
• Slow HARBINGER
• Nodes only cycle in/out sleep state after entire
  codeword transmitted.
• Coherence time ?? M blocks
• Slow HARBINGER-A
• Tries to minimize latency
• Slow HARBINGER-B
• Tries to minimize energy consumption

18
GeRaF

• Geographic Random Forwarding (GeRaF)
• Zorzi and Rao (Trans Mobile Computing 2003)
• Node activity follows a sleep schedule.
• Common strategy for sensor networks.
• Source broadcasts over an AWGN channel.
• If one node is within range it becomes the
  designated relay.
• If multiple nodes, the one closest to destination
  becomes relay.
• Otherwise, source tries again later to see if a
  relay awoke.
• No ARQ or diversity combining effect.
• This is precisely HARBINGER with the simplifying
  assumptions and M1 (no ARQ)

19
Slow HARBINGER-A
Topology 2-D Poisson S-D separated by 10
m Coding parameters Per-block rate R1 Code
Combining Normalized power (Initial TX range is 1
m) Channel parameters n 3 path loss
exponent 2.4 GHz d0 1 m reference
dist Protocol picks the node that is closest
to the destination.
20
Slow HARBINGER-B
Topology 2-D Poisson S-D separated by 10
m Coding parameters Per-block rate R1 Code
Combining Normalized power (Initial TX range is 1
m) Channel parameters n 3 path loss
exponent 2.4 GHz d0 1 m reference
dist Protocol picks the best relay that can be
reached with fewest ARQ transmissions
21
Conclusions

• Wireless is a broadcast-oriented medium
• Link-oriented protocols do not exploit this.
• Ad hoc network can be viewed as a distributed
  MIMO system.
• Cooperative diversity (orthogonal relaying) can
  give a better tradeoff between energy and latency
  than traditional multihop.
• The number of participating relays should be
  carefully chosen.
• A cross-layer approach can yield significant
  gains
• Error control using hybrid-ARQ
• CSMA-style medium access control
• Position-based relaying
• Analytical results possible under simplified
  conditions.