On the performance of geographical routing in the presence of localization errors

1
On the performance of geographical routing in the
presence of localization errors
Rahul C. Shah, Jan Rabaey University of
California, Berkeley
Adam Wolisz Technical University, Berlin
2
Geographic routing

• Nodes know
• Their own location
• Neighbors location
• Destination location
• Use the neighbor closest to the destination

destination
Current node
3
Routing around voids

• Geographic routing protocols differ on the
  recovery mechanism
• Common mechanisms
• Face routing
• Flooding

destination
Current node
4
Geographic routing protocols

• Ad hoc networks
• Greedy perimeter stateless routing (GPSR)
• Location aided routing (LAR)
• Distance routing effect algorithm for mobility
  (DREAM)
• Sensor networks
• Geographic random forwarding (GeRaF)
• Opportunistic routing
• Geographic hash tables (GHT)

5
Properties of geographic routing

• Advantages
• Very little knowledge of the network needed
• Routing table size O(log N) where N is the
  number of nodes in the network
• Disadvantages
• Need separate mechanisms to handle voids/
  obstacles
• Flooding
• Face routing
• Requires a localization mechanism

6
Localization errors
Radial uncertainty model
Actual position
?Rmax
Perceived position equally likely in circle
independent errors. Rmax is the maximum radio
range.

• Nodes inside forwarding region perceive they are
  outside
• Nodes outside forwarding region perceive they
  are inside

7
Using second order neighborhood info
Second order neighbors Neighbors of
neighbors Use only those neighbors with (second
order) neighbors closer to the destination
First order neighbors
Second order neighbors
8
Simulation model

• Created in Omnet
• Nodes uniform randomly placed in 100m x 100m area
• Interference and collision effects not modeled
• Transmit and receive power was measured
• Metrics
• Packet delivery ratio
• Route discovery cost (power)
• Protocols simulated
• With and without flooding
• With and without second order routing

9
Analytical results
No second order routing

• ? node density
• ? localization error
• ? area of forwarding region

10
Simulation and analytical results
Packet delivery ratio ()
Number of neighbors per node
11
Goodput with localization errors
Packet delivery ratio ()
Location error (Percentage of radio range)
No flooding is used
If flooding is used
12
Second order info improves goodput
Packet delivery ratio ()
Packet delivery ratio ()
Location error (Percentage of radio range)
Location error (Percentage of radio range)
Without using second order neighborhood
information
With using second order neighborhood information
No flooding is used
13
When obstacles are present
Packet delivery ratio ()
Packet delivery ratio ()
Location error (Percentage of radio range)
Location error (Percentage of radio range)
Without using second order neighborhood
information
With using second order neighborhood information
No flooding is used
14
Power consumption
Power consumed per node (mW)
Power consumed per node (mW)
Location error (Percentage of radio range)
Location error (Percentage of radio range)
Without obstacles in the network
With obstacles in the network
Flooding is used
15
Conclusions

• Geographic routing can tolerate
• Errors up to 20 radio range
• Second order routing improves tolerance to about
  40
• Flooding is very expensive to route around voids
• Quenching the flood would improve the performance
• GPSR is better, but it cannot guarantee packet
  delivery even if a route exists
• Second order routing shortens path lengths by
  about 10 (flooding is used)
• Distributed localization algorithms produce
  correlated errors important to analyze
  performance in those cases