Efficient Geographic Routing in Multihop Wireless Networks

1
Efficient Geographic Routing in Multihop
Wireless Networks

• Seungjoon Lee, Bobby Bhattacharjee, Suman
  Banerjee
• MobiHoc, 2005
• Kang Mi-kyung
• (mkkang_at_camars.kaist.ac.kr)

2
Contents

• Introduction
• Overview of NADV
• NADV
• New link metric for geographic routing
• Link cost types and estimation
• Simulation and results
• Conclusion

3
Introduction

• Geographic routing
• Location information for packet delivery
• Next hop node selection based on neighborhood and
  destination information
• No route establishment, no per-destination state
• Popular strategy for geographic routing
• To the neighbor geographically closest to the
  destination
• Route around voids problem
• No neighbor closer to the destination than the
  current node

d
v
w
x
y
z
4
Overview of NADV

• NADV (normalized advance)
• New link metric for geographic routing
• Optimal trade-off between proximity and link
  cost
• Adaptive routing
• Efficient routing
• Support a variety of different cost types
• Different routing strategies depending on system
  objectives, message priority or applications

5
New link metric for geographic routing(1/3)

• ADV (advance) background
• Current node S greedily select the neighbor
  closest to destination T
• Minimization the hop count between source and
  destination
• Advance (ADV) of n
• Amount of decrease in distance by a neighbor n
• Demerit
• No consideration of link cost
• Use of poor quality links, unnecessarily high
  communication cost

D(x) distance from node x to T
Large advance
Good link quality
vs
6
New link metric for geographic routing(2/3)

• NADV (normalized advance)
• Normalized advance of neighbor n
• -gt Expected advance per transmission

Psucc(n) probability of success in transmitting
data to n
7
New link metric for geographic routing(3/3)

• Optimality of NADV in an idealized environment
• Link costs along the found path by NADV is
  minimum
• Assumptions
• We can find a node at an arbitrary point
• Link cost is an unknown increasing convex
  function of distance
• Process
• DIST distance from source S and destination T
  (relatively large)
• Optimal path straight line between S and T
• Sum of link costs minimized when all links have
  the same distance
• Optimal interval

T
S
8
Link cost types and estimation (1/7)

• Wireless integration sublayer extension (WISE)
• Three types of link cost
• Packet error rate
• Link delay
• Energy consumption

• For efficient link cost estimation
• Additional control messages available
• -gt WISE extract relevant link cost info.
• Otherwise
• -gt WISE exploits MAC-specific info.

9
Link cost types and estimation (2/7)

• Packet error rate (PER)
• Four PER estimation methods for
• Using probe messages
• Using signal-to-noise ratio
• Neighborhood monitoring
• Self monitoring

10
Link cost types and estimation (3/7)

• PER estimation 1 Using probe messages
• Link error probability
• Probe message
• Reception ratio from periodic message exchange
• Adjusting PER depending on the data packet length
• l-bit probe messages
• Infer bit error rate from observed PER(l)
• L-bit data frame

Observed and estimated PERs for five
experiments with varying distance
11
Link cost types and estimation (4/7)

• PER estimation 2 Using signal-to noise ratio
  (SNR)
• Bit error rate
• PER estimation 3 Neighborhood monitoring
• Passive monitoring to infer link PERs
• Node A monitor frames sent by neighbors
• Using the MAC sequence number, A count frames
  missed from neighbor B
• A infer PER of link from B to A
• A needs to inform B of the PER estimation

12
Link cost types and estimation (5/7)

• PER estimation 4 Self monitoring
• Condition
• Additional control messages not possible
• Modification of beacon message format not
  possible
• Technique
• Node transmits a data frame to neighbor n
• Mac-layer informs the WISE whether transmission
  was successful or not
• F1 (fail), F0 (success)
• PER of wireless link to neighbor n

13
Link cost types and estimation (6/7)

• Delay
• Two types of link delay
• Medium time
• Time spent in sending a packet over the link
• WISE can easily retrieve the current value of
  transmission rate from the MAC layer and
  calculate the necessary medium time to the
  neighbor
• Total delay
• Time from the packet insertion into the interface
  queue until the notification of successful
  transmission
• Queuing delay, backoff time out, contention
  period, retransmissions due to errors or
  collisions

14
Link cost types and estimation (7/7)

• Power consumption
• Assumptions
• Control mechanism for transmission power
  adjustment to save battery
• Appropriate transmission power level Ptx
• WISE retrieve Ptx and calculate actual system
  power consumption Cpower

15
Simulation model (1/2)

• Environment
• Ns-2 simulation
• Node placement
• Uniformly at random on a 1000m by 1000m square
• Static source at (50, 500), destination at (50D,
  500)
• Geographic routing simulation code for GPSR
• IEEE 802.11b standard for the underlying MAC
  layer protocol
• Error model
• Random packet error model
• Performance of NADV in the presence of randomness
  in packet errors
• Blacklisting
• Fixed threshold
• Node excludes neighbors with low-quality link
  based on the threshold

16
Simulation model (2/2)

• Power consumption model
• Transmission power for successful reception at a
  receiver
• Transmission power
• Energy each packet forwarding consumes

17
Simulation results (1/4)

• Experiments with perfect estimation of link
  errors
• NADV vs ADV
• Data transmission overhead of ADV increases
  abruptly
• NADV vs blacklisting
• Blacklisting different threshold values lead to
  best results
• NADV adapt to the changing network

• Number of retransmission
• Network bandwidth, resources
• Packet delay

18
Simulation results (2/4)

• Experiments using proposed PER estimation
  techniques
• Changing noise power
• Adaptiveness of PER estimation schemes

• Start with high noise
• After 300 sec, low noise
• After 700 sec, medium noise
• ( ) Packet delivery ratio

19
Simulation results (3/4)

• Using power consumption as link cost

Average power consumption with different
schemes
20
Simulation results (4/4)

• Experiments with generic cost
• Link cost
• Experiment scenario
• Source and destination are 900meters apart
• Source starts to send data packets after 10
  seconds
• At 30 seconds, environment of some part of the
  network changes
• We randomly select 50 of links and increase
  their link costs by 50

r uniformly distributed random number d
distance between two nodes R maximum
transmission range
Average path quality of each scheme before and
after the link cost change
21
Conclusion

• NADV
• New link metric for geographic routing in
  multihop wireless networks
• Adaptive, general and useful for various link
  cost types
• Combination of NADV and cost estimation
  techniques outperforms the current geographic
  routing schemes
• NADV finds paths whose cost is close to the
  optimum