Geographic Routing without Location Information Ananth Rao, Sylvia Ratnasamy, Christos Papadimitriou

1
Geographic Routing without Location
InformationAnanth Rao, Sylvia Ratnasamy,
Christos Papadimitriou, Scott Shenker, Ion
StoicainMobiCom 2003

• Presented by
• Sanjeev Dwivedi

2
Sensor Networks Gentle Intro

• Small Form Factor, Low Resources
• Dense Deployment
• Relatively Inexpensive
• Unorganized deployment
• Low Overhead, Scalable protocols needed

On Demand Behavior
3
Scalability The Achilles Heel

• Routing means keeping and updating state
• Dense deployment gt High Overhead
• Ideally, should be able to work with Local State
  only.
• It is a hard problem, Need additional primitives

4
Solution Geographic Routing

• Need to keep state only about nearby nodes gt
  Save Messaging overhead.
• Data can be forwarded based on Locations instead
  of Identifiers
• Highly Scalable Routing table size does not
  depend on the size of the network.
• Few Geographic routing protocols proposed GPSR.
• But, Not all nodes might have location
  information.

5
Geo. Routing w/o Location!

• Assign Virtual Coordinates to nodes (these need
  not give accurate representations of the
  underlying geography but, in order to serve as a
  basis of routing, they must reflect the
  underlying connectivity
• gt
• Construct virtual coordinates using only local
  connectivity information.

6
Routing the data packets

• Greedy Forwarding Forward to the node which is
  closest to destination than the current node.
• Stop if you are the destination.
• Dead-End If no node is closer to destination
  than current node, try expanding ring search,
  until a closer node is found or TTL exceeds.

7
Virtual Coordinate Construction

• Based on Iterative Relaxation Techniques
• Presented in Three Steps
• (a) Perimeter nodes know that they are
  perimeter nodes and know coordinates
• (b) Perimeter nodes know they are perimeter
  nodes
• (c) Perimeter nodes dont know coordinates or
  that they are perimeter nodes

8
Evaluation Topology Criteria
3200 Nodes, Uniformly Distributed in a
200x200 area. 8 unit range, 16 Neighbors on Avg.
-Success Rate -Average Path Length
9
Perimeter Nodes know Location
10
100 Iterations
10 iterations
1000 Iterations
11
Perimeter Nodes Dont know Location

• Perimeter Nodes flood HELLO messages and measure
  distances to each other. The distances are stored
  in a Perimeter Vector
• Flood Perimeter Vector in the Network
• Perimeter Nodes select coordinates so as to
  minimize the sum

12
Setting Up Coordinates for Interior Nodes

• Interior Nodes can start with a static location,
  but it takes a long time to converge
• If information from the Perimeter Vector Exchange
  is used, within one iteration, 99.2 success rate
  is achieved.

13
Dealing with Message Losses etc.

• Message Losses and Node failures can lead to
  different perimeter nodes computing inconsistent
  coordinates
• Soln Use two bootstrapping nodes.

14
Bootstrapping Nodes

• Flood Beacons in Beginning
• Perimeter Nodes include that information in
  coordinate computation
• Perimeter Nodes compute CG, which acts as origin.
  Bootstrap nodes act as Axes.
• CG is resilient to Incomplete Information.

15
Perimeter Nodes dont know they are on the
Perimeter

• Bootstrap Nodes flood the network
• Perimeter Criterion if a node is farthest away,
  among all its two-hop neighbors from the first
  bootstrap node, then the node decides that it is
  on the perimeter.
• False Positives are relatively low and have
  little effect on the eventual outcome.

16
Performance Evaluation Model

• Losses Drop packets with p probability
• Mobility Random waypoint
• Obstacles Straight Walls parallel to Axes
• Irregular Shapes and Voids have been also
  included.

17
Success Rate Iterations
18
Path Length Virtual vs. Actual Coordinates
Almost No Difference!
19
Routing Load Virtual vs. True
20
Success Rate vs. Network Size at Different
Densities
200x200 network
250x250
21
Success Rate vs. Loss Rate
22
Success Rate vs. Obstacles