Techniques for Improving Opportunistic Sensor Networking Performance

1
Techniques for Improving Opportunistic Sensor
Networking Performance

• Shane Eisenman Nic Lane, Andrew
  Campbell

Columbia University
Dartmouth College
2
Large-scale Sensing requires...
3
Application Query Assignment (i.e., Tasking)
Local Tasking
4
Application Query Assignment
Remote Tasking
Internet
RAN 1
Sensing Query
RAN 2
John Q. Public
5
I have the query. Now what?

• Query Sensor type,
• target location,
• sensing deadline, ...

No sensor --gt ?
6
Improving Success Probability

• Sensor Sharing
• Sensor Substitution

7
Modeling Query Success Probability

• P(success) P(reaching tasking point)
• x P(reaching
  target)
• x
  P(having right sensor)
• ...before the sensing deadline.

8
N x N grid representation of a neighborhood
9
Torus Topology (N10)
Move N,E,S,W
10
Campus Topology (N10)
Move according to real-world connectivity
11
Results

• Randomly assign sensors to mobile nodes
• (20 choose r)
• Calculate the sensing success probability for
  various values of number of nodes, number of
  sensors per node, and sensing deadline

12
Comparing Campus and Torus
Fixed num sensor types3.
The topology of the campus limits node mobility,
requiring a longer deadline to reach the same
success probability.
13
Comparing Campus and Torus
Fixed sensing deadline50.
Qualitative trends are very similar between Torus
and Campus, despite the very different mobility
patterns.
14
Sensor Sharing
P(success) P(reaching tasking point)
x P(reaching target)
x ( P(having right sensor)
P(sensor
sharing) ) ...before the sensing deadline.
15
Benefit of Sensor Sharing
Fixed num sensor types3.
Sharing improves success probability up to 16.
As the deadline increases the improvement
diminishes.
16
Benefit of Sensor Sharing
Fixed sensing deadline50.
Sharing improves success probability up to 70.
Increasing number of sensor types decreases
possible improvement.
17
Sensor Substitution
P(success) P(reaching tasking point)
x P(reaching target)
x ( P(having right sensor)
P(sensor
substitution) ) ...before the sensing deadline.
18
Benefit of Sensor Substitution
Fixed num sensor types3. Fixed num mobile
sensors20.
Substitution improves success probability up to
155. Increasing deadline decreases possible
improvement.
19
Benefit of Sensor Substitution
Fixed sensing deadline50. Fixed substitution
prob0.02.
Substitution improves success probability up to
36. Increasing num. sensor types decreases
improvement.
20
Combined Sharing and Substitution
P(success) P(reaching tasking point)
x P(reaching target)
x ( P(having right sensor)
P(sensor
sharing)
P(sensor substitution)) ...before the sensing
deadline.
21
Combined Sharing and Substitution
Fixed num sensor types3. Fixed num mobile
sensors20.
Up to 270 gains in success probability
(sharing-only 16, substitution-only 160 with
same parameterization).
22
Combined Sharing and Substitution
Fixed sensing deadline50. Fixed substitution
prob0.02.
Up to 140 gains in success probability
(sharing-only 70, substitution-only 35 with
same parameterization).
23
Summary

• Opportunistic people-centric sensing is the
  future for very large-scale sensing.
• Sensor Sharing and Sensor Substitution are
  composable techniques to increase the probability
  of query success.
• Help to provide a level of abstraction between
  locally sensing resources and application
  requirements.

24
For more information...

• http//METROSENSE.cs.dartmouth.edu