Directed Diffusion for Wireless Sensor Networking

1
Directed Diffusion for Wireless Sensor Networking

• C. Intanagonwiwat, R. Govindan, D. Estrin, et.
  al.IEEE/ACM Transactions on Networking, Feb 2003
• ??? ???

2
Contents

• Introduction
• Directed Diffusion
• Naming
• Interests and Gradients
• Data Propagation
• Reinforcement
• Performance Evaluation
• Implementation
• Conclusion

3
Introduction (1/2)

• Wireless sensor networks
• Sensing devices with communication capability
• Event monitoring
• Enemy detection, aircraft interiors, large
  industrial plants
• Data-centric communication
• Data is named by attribute-value
• Different form IP-style communication
• End-to-end delivery service
• e.g.) How many pedestrians do you observe in the
  geographical region X?

A sensor field
Sources
Event
Sink Node
4
Introduction (2/2)

• Data-centric communication (cont.)
• Human operators query (task) is diffused
• Sensors begin collecting information about query
• Information returns along the reverse path
• Intermediate nodes aggregate the data
• Combing reports from sensors
• Challenges
• Scalability
• Energy efficiency
• Robustness / Fault tolerance in outdoor areas
• Efficient routing

5
Directed Diffusion

• Directed diffusion consists of
• Interest - Query which specifies what a user
  wants
• Data - Collected information
• Gradient
• Direction and data-rate
• Events start flowing towards the originators of
  interests
• Reinforcement
• After the sink starts receiving events, it
  reinforces at least one neighbor to draw down
  higher quality events

6
Data Naming

• Expressing an Interest
• Using attribute-value pairs
• E.g.,
• Other interest-expressing schemes possible
• E.g., hierarchical (different problem)

Type Wheeled vehicle // detect vehicle
location Interval 20 ms // send events
every 20ms Duration 10 s // Send for next
10 s rect -100,100, 200,400 // from
sensors in this area
7
Interests and Gradients

• Interest propagation
• The sink broadcasts an interest
• Exploratory interest with low data-rate
• Neighbors update interest-cache and forwards it
• Flooding
• Geographic routing
• Use cached data to direct interests
• Gradient establishment
• Gradient set up to upstream neighbor
• Low data-rate gradient
• Few packets per unit time needed

8
Exploratory Gradient
Exploratory Request Gradient
Event
Bidirectional gradients established on all links
through flooding
9
Data Propagation

• A sensor node that detects a target
• Searches its interest cache
• Computes the highest requested data-rate among
  all its outgoing gradients
• Data message is unicast individually
• A node that receives a data message
• Finds a matching interest entry in its cache
• Checks the data cache for loop prevention
• Re-sends the data to neighbors

10
Reinforcement (1/4)

• Positive reinforcement
• Sink selects the neighboring node
• Original interest message but with high data-rate
• Neighboring node must also reinforce at least one
  neighbor
• Low-delay path is selected
• Exploratory gradients still exist
• useful for faults

Event
A sensor field
Sink
11
Reinforcement (2/4)

• Path establishment for multiple sources and sinks
• Node reinforce all neighbors from which new
  events were recently received

12
Reinforcement (3/4)

• Path failure and recovery
• Link failure detected by reduced rate, data loss
• Choose next best link
• Negatively reinforce lossy link
• Either send interest with base (exploratory) data
  rate or allow neighbors cache to expire over
  time

Link A-M lossy A reinforces B B reinforces C C
reinforces D or A negative reinforces
M M negative reinforces D
Event
D
M
Src
A
C
Sink
B
13
Reinforcement (4/4)

• Using negative reinforcement
• Path Truncation
• Loop removal
• For resource saving
• B negative reinforces D, D negative reinforces E,
  

14
Performance Evaluation (1/6)

• Environment
• Animal tracking instance of directed diffusion
• ns-2 simulator
• Ranging from 50 to 250 nodes in 160mx160m
• 1.6 Mbps 802.11 MAC layer with small modification
• Random 5 sources in 70mx70m, random 5 sinks
• Metric
• Average dissipated energy
• Per node energy dissipation / events seen by
  sinks
• Average Delay
• Latency of event transmission to reception at
  sink
• Distinct event delivery ratio
• Ratio of events sent to events received by
  sink
• Compared with
• flooding and omniscient multicast

15
Performance Evaluation (2/6)

• Average dissipated energy

0.018
0.016
Flooding
0.014
0.012
0.01
0.008
(Joules/Node/Received Event)
Omniscient Multicast
Average Dissipated Energy
0.006
Diffusion
0.004
Due to the data-aggregation Nodes suppress
duplicate location estimates
0.002
0
0
50
100
150
200
250
300
Network Size
16
Performance Evaluation (3/6)

• Average delay

0.35
0.3
Flooding
0.25
0.2
Average Delay (secs)
0.15
0.1
Omniscient Multicast
Diffusion
0.05
0
Uncongested sensor network Reinforcement rules
find the low delay path
0
50
100
150
200
250
300
Network Size
17
Performance Evaluation (5/6)

• Impact of dynamics (Distinct event delivery
  ratio)

18
Performance Evaluation (5/6)

• Impact of negative reinforcement

Diffusion Without Negative Reinforcement
Prune off higher latency path
Diffusion With Negative Reinforcement
19
Performance Evaluation (6/6)

• Impact of duplicate suppression

Diffusion Without Suppression
Negative reinforcement
Diffusion With Suppression
Suppress identical data sent
20
Conclusion

• Directed diffusion, a paradigm proposed for event
  monitoring sensor networks
• Energy efficiency achievable
• Diffusion mechanism resilient to fault tolerance
• Network addressing is data centric
• Notion of gradient (exploratory and reinforced)
• Implementation on Motes
• Network API Publish/subscribe paradigm
• Filter API Aggregate the data, generate
  reinforcements