Task: Network map discovery
Envisioned applications:
-
Network management tool; graphical representation of
connectivity
- Location discovery
Assumptions:
- Local neighbor discovery collects localized link
states to neighbors
- Some nodes may have GPS receivers
- There is some other mechanism that can be used to
assess relative distance to neighbors
- hop count? signal strength?
Potential problems:
- signal strength is flaky wrt obstructions, etc.
- ORL people had problems getting a good reading at
low power levels
- Idea: send message twice, second time at reduced power.
Nodes at the edge of transmission range will not hear
second message
- time of flight? ultrasound?
First order solution:
- Flood link state information to all neighbors
- Program to generate data for testing purposes, for modelling
larger numbers of nodes, initially very simplistic, later
more accurate radio models, etc?
Potential refinements:
- Limit flooding distance
- Filter link state to remove transients and unimportant
variations in distance measurements
- can we make the link state/distance info converge for
fixed node networks?
- what is the incremental impact of mobility?
- partition nodes into neighborhoods around gps nodes
- neighbors negotiate to disambiguate position information.
for example, suppose two nodes deduce their radial distances
from different gps nodes. if they can approximate their
relative distance from each other they can reduce their
locus of possible positions from a circle to two points.
(in 3-d it reduces from a sphere to some kind of oval)
- three non-colinear nodes that determine their relative locations
from each other determine a triangle with fixed angles.
however, relative to absolute coordinates, rotation, and
translation are unknown, and there is a mirror image
ambiguity. However progressive determination of adjacent
triangles can determine relative distance between any two nodes.
Input from Niru??