Exploiting Mobility of Adhoc Wireless Networks

1
Exploiting Mobility of Ad-hoc Wireless Networks

• Presented by Andjela Ilic

2
Problems treated in the past

• routing, physical layer
• Diversity of time, frequency, space

3
Nobody focused on capacity tillGupta Cumar

• The Capacity of Wireless Networks capacity
  constrained by the mutual interference of
  concurrent transmissions
• Multiuser diversity

4
A couple of ways to observe capacity

• In old-fashioned manner (Shannon)C BW ln( 1
  S/N )
• Or, using link rate? BW ln ( 1 P/(NI ))
• Or, regarding performance measuringC Thr /
  TX_Rate, whereC - max throughput for a certain
  topologyThr no. of bps successfully
  transmittedTX_Rate normalization by
  corresponding transmission rate

5
Assumption 1

• Random Network -n nodes are iid uniformly
  distributed in a disk of unit area
• -Each node has a randomly chosen destination
  for sending ?(n) bits/sec
• -Nodes are homogeneous all transmissions use
  the same power

6
Assumption 2

• Physical Model-successful transmission from Xi
  to Xj occurs if
• P / (Xi Xj)a/ N P / S(Xk Xj)a ß

7
Assumption 3

• n nodes, n S-D pairs,every node is src, dst and
  possible relay, at the same time

8
In fixed model (static nodes)

• as n increases -gtgt ?(n) per node decreases
  approx as 1/vn, even when optimal scheduling,
  routing and relaying is used (Gupta Cumar)
• Main perf. limitation is that long-range
  communication between many node pairs is
  infeasible, due to interference- throughput is
  interference limited
• comm. allowed just between neighbors, at d1/vn.
  
• Too much relaying-gtgtuseful throughput small

9
Some definitions

• Throughput time avg of bps that can be
  transmitted by each node to its destination
• Feasible throughput - throughput ?(n) bps for
  every node is feasible if it is possible to
  schedule transmissions, such that (with buffering
  and multihop fashion) every node can send ?(n)
  bps to its destination.

10
Main result of Gupta Cumar

• Lim P?(n) cR / v(nlogn) is feasible1
• Lim P?(n) cR / v(n) is feasible0
• i.e. within the factor of vlogn, ?(n) per node
  goes to 0 like R / v(n)

11
BUT

• When mobility is introduced into this model,it
  is shown that per node throughput can be kept
  constant, as n increases.

12
Mobility, but not relaying

• We assume that nodes will be near, from time to
  time
• Lim P?(n) cn -1/(1 a/2) R is feasible0
• Available throughput per S-D pair goes to 0 as n
  -1/(1 a/2)
• If we constrain comm. to nearest neighbors, there
  are just small no. of nodes close enough to
  communicate- throughput is distance limited

13
Third solution

• To overcome interference limitation- find a way
  for nodes to comm. only locally
• To overcome distance limitation-network need to
  have enough no. of S-D pairs
• Solution must do relaying!

14
How it works

• the packets of every S distributed across the
  network so each S-D pair always has packets to
  send.
• nodes independent, stationary and ergodic, it is
  sufficient to relay only once!
• achievable total throughput is n
• Lim P?(n) cR is feasible1

15
And

• each relay node always has packets to transmit to
  its nearest neighbor in steady state, the average
  throughput per S-D pair is limited by
  interference and achieves constant rate
• A two-phase approach is useful
• -Phase 1 Scheduling of pkt transmissions from
  sources to relays or destination -Phase 2
  Scheduling of pkt transmissions from relays or
  source to destinations

16
(No Transcript)
17
Data pkts relayed, FTP, LAR1
18
Data pkts relayed, FTP, LAR11
19
IP pkts traversed over node, CBR
20
Data pkts relayed/node, CBR
21
Receiver E2E delay, CBR
22
Special thanks

• to all of you Patient Listeners