RUGGeD: RoUting on finGerprint GraDients in Sensor Networks

1
RUGGeD RoUting on finGerprint GraDients in
Sensor Networks
Jabed Faruque, Ahmed Helmy
Wireless Networking Laboratory Department of
Electrical Engineering University of Southern
California faruque_at_usc.edu, helmy_at_usc.edu URL
http//nile.usc.edu, http//ceng.usc.edu/helmy
ICPS 2004 1
2
Introduction

• Sensor networks consist of sensor nodes with
• Limited Energy source
• Sensor devices
• Short range radio
• On-board processing capability

Mica2 mote and sensor board

• Use of Sensor networks is tightly coupled with
  physical phenomena
• May be most widely used for habitat and
  environment monitoring (e.g. temperature,
  humidity)
• For unattended and fine grained monitoring of
  natural phenomena
• Self configuration capability
• Also others e.g., for defense purpose

ICPS 2004 2
3
Motivation

• Every physical event produces a fingerprint in
  the environment, e.g.,
• Fire event increases temperature
• Nuclear leakage causes radiation

Many physical phenomena follow diffusion law
f(d) ? 1/d?, where d distance from the
source, ? diffusion parameter, depends on the
type of effect (e.g. for temperature ?
1, light ? 2)
ICPS 2004 3
4
Example (of diffusion) Isoseismal (intensity)
maps (North Palm Springs
earthquake of July 8, 1986 )
Ref. Southern California Earthquake Center.
(http//www.scec.org)
ICPS 2004 4
5
Why Using Natural Information Gradient is
Important?

• This natural information gradient is FREE
• Routing protocols can use it to forward query
  packet (greedily)
• - Locate event(s) e.g., fire, nuclear leakage.
• Can be extended for other notions of gradients
• - Example Time gradients can be used for mobile
  target tracking
• Existing approaches flooding, expanding ring
  search, random-walk, etc. do not utilize this
  information gradient

ICPS 2004 5
6
Challenges

• In real life, sensors are unable to detect or
  measure the events effect below certain
  threshold. So, diffusion curve has finite tail
• - Lack of sensitivity of sensor device(s)
• Erroneous reading of malfunctioning sensors
• - Due to calibration errors or obstacle- Cause
  local maxima or minima
• Environmental noise

ICPS 2004 6
7
Objective
Design an efficient algorithm to locate source(s)
in sensor networks, exploiting natural
information gradients i.e., the diffusion pattern
of the events effect - Gradient based- Fully
distributed- Robust to node or sensor failure or
malfunction- Capable of finding multiple sources
ICPS 2004 7
8
Related Work 1,2,3

• Traditional routing protocols for sensor
  networks are based on Flooding
  (directed-diffusion) or Random-walk (Rumor-
  routing, ACQUIRE, etc.)
• - Flooding based methods cause huge energy
  overhead
• - Random-walk increases latency and failure
  probability
• - Do not utilizes the natural information
  gradient
• Existing Information driven protocols 4,5 use
  single path approaches with/without look-ahead
  parameter
• - Use a proactive phase to prepare information
  repository
• Cause significant overhead at low query rate
• - Unable to handle local maxima or minima
• - Unable to find multiple sources
• - Robustness depends on the proactive phase and
  the look- ahead parameter

ICPS 2004 8
9
Protocol
? A node can exist in one of two modes/states -
flat-region mode - gradient-region mode ? A
node forwards the query to neighbors with its
information level ? To forward the query, each
node uses following algorithm 1.
Information gradient region follows greedy
approach - Forwards the query to the
neighbors if the information level about the
event improves 2. Unsmooth gradient
region use probabilistic forward based
on Simulated Annealing - Probabilistic
function is fp(x) 1/xa, where x hop count in
the information gradient region and a
depends on the diffusion parameter (? )
3. Use flooding for the flat (i.e., zero)
information region - Decrease latency to
reach gradient information region - Handles
query in the absence of events ? Query ID
prevents looping ? Once query is resolved, a
node uses the reverse path to reply
ICPS 2004 9
10
E
Q

• All neighbors (np) of Mx have less information,
  so they forward the query to their neighbors
  probabilistically

• All neighbors (ng) of Mn have more information,
  so they forward the query to their neighbors

ICPS 2004 10
11
Simulation Model

• Two different sensor network layouts 1.
  100 X 100 regular grid of 10000 nodes. Event
  located at (74,49) 2. 15 X 6 grid of 90
  nodes in 225 x 375 m2 sensor field with 50m
  communication radius. Grid points are
  perturbed by Gaussian noise (0,25)
• Diffusion parameter ? set to 0.8
• Two regions exist in each layout - Flat or
  zero information region - Gradient
  information region
• Malfunctioning nodes are uniformly
  distributed in both region
• Environmental noise is present in the
  gradient information region
• Malfunctioning nodes have arbitrary readings
  - For global maxima search, protocol uses a
  filter to prohibit replies from nodes
  having arbitrary high value

ICPS 2004 11
12
Query Types

• Single-value query - Search for a specific
  value and have a single response
• Global Maxima search (only sensor layout 1 is
  used) - Search for the maximum value of
  information in the system - Intermediate nodes
  suppress non-promising replies
• Multiple Events detection (only sensor layout 1
  is used) - Search for multiple events of the
  same type

ICPS 2004 12
13
Single-value query- effect of flat information
region nodes(3 environmental noise and 15
malfunctioning nodes)
- With increase of flat region - Flooding
overhead becomes dominant increasing energy
consumption - Malfunctioning nodes cause query
to switch to gradient mode erroneously - Decrease
in a creates more paths, increasing
reachability and energy consumption
ICPS 2004 13
14
Single-value query- effect of the malfunctioning
nodes(3 environmental noise and 36 flat
information region nodes)

• With increase of malfunctioning nodes the
  protocol switches from the flat region mode to
  the gradient region mode rapidly - Reduces
  flooding overhead - Increases failure rate

ICPS 2004 14
15
Single-value query- route a query around the
sensors hole(3 environmental noise and 20
malfunctioning nodes)

• For smaller value of a (e.g., a 0.65),
  reachability is above 98 even at the presence of
  55 flat information region

• For the probabilistic function fp(x) 1/xa, a lt
  ? is recommended, but close to ? gives optimal
  trade-off between reachability and overhead

ICPS 2004 15
16
Global Maxima Search-effect of flat information
region nodes(3 environmental noise and 15
malfunctioning nodes)
(without Filter)
(with Filter)

• Average energy dissipation reduces significantly
  due to use of the simple filter

ICPS 2004 16
17
Multiple Events Detection-effect of flat
information region nodes(3 environmental noise
and 15 malfunctioning nodes)

• With the increase of number of sources, some
  plateaux regions are created in the resultant
  gradient information region that require more
  probabilistic forwarding
• - for five or more sources, a 0.35 is a good
  setting in the simulated scenario

ICPS 2004 17
18
Conclusion

• Developed a multiple-path exploration protocol
  to discover events in sensor networks efficiently
• The protocol is fully reactive, effectively
  exploits the natural information gradients and
  controls the instantiation of multiple paths
  probabilistically
• The performance of the probabilistic function is
  closely tied to the diffusion parameter
• Three different problems were studied
• Single-value, Global maximum, Multiple events
• Obtained high success rate to route around the
  sensors hole, with proper setting of the
  probability function parameters
• More efficient than existing approaches

ICPS 2004 18
19
On-going and Future work

• Establish analytical relationship between
  diffusion pattern and the probabilistic
  forwarding function
• Develop protocol for target tracking and target
  counting using the multiple path exploration
  mechanisms

ICPS 2004 19
20
Backup Slides
21
Environment Model

• f(di) f(di) fEN(f(di)),
• fEN(f(di)) ? fmax - f(di)
• where,
• di distance of the location from peak
  information point (i.e., the event)
• f(di) gradient information of the location with
  environmental noise,
• fmax peak information,
• f(di) gradient information without
  environmental noise.
• The proportional constant is considered 0.03 to
  model the environmental for our protocol, i.e.,
  3 environmental noise is considered

22
Filtering of Malfunctioning Nodes

• Let distance of sensors S1 and S2 from the
  events location are d and d1 hops with readings
  R1 and R2

In our simulations ? 0.8
We use the filter
23
Reply Suppression Mechanism
Intermediate nodes suppress the non-promising
replies
24
(No Transcript)
25
References
1 C. Intanagonwiwat, R. Govindan and D. Estrin,
Directed Diffusion A Scalable and Robust
Communication Paradigm for Sensor Networks,
MobiCom 2000. 2 D. Braginsky and D. Estrin,
Rumor Routing Algorithm for Sensor Networks",
WSNA 2002. 3 N. Sadagopan, B. Krishnamachari,
and A. Helmy, Active Query Forwarding in Sensor
Networks (ACQUIRE)", SNPA 2003. 4 M. Chu, H.
Haussecker, and F. Zhao, Scalable
Information-Driven Sensor Querying and Routing
for ad hoc Heterogeneous Sensor Networks", Int'l
J. High Performance Computing Applications,
16(3)90-110, Fall 2002. 5 J. Liu, F. Zhao,
and D. Petrovic, Information-Directed Routing
in Ad Hoc Sensor Networks", WSNA 2003.
ICPS 2004 20