Context-Driven Disruption Tolerant Networking for
Vehicle Applications
Wai Chen
Director and Chief Scientist
Applied Research, Telcordia Technologies
C t t waichen@ieee.org
Contact:
i h @i
The Fully Networked Car
Geneva, 3-4 March 2010
1
Context-Driven Disruption Tolerant Networking for
Vehicle Applications
Wai Chen1, Jasmine Chennikara-Varghese1, Ratul Guha1, John Lee1, Ryokichi
Onishi2, Rama Vuyyuru2, Junichiro Fukuyama2
1Telcordia
Technologies
2Toyota InfoTechnology Center
March 3-4, 2010
2
3
Background and Motivation for DTN
o Dissemination of information among vehicles and RSUs with frequent
connectivity disruptions
o Current approaches
pp
adapted
p
from P2P content deliveryy networks
Approaches
Connection variants
• Vehicle-based Peer-to-Peer
• Fragmentation and forwarding
• Carry
C
and
d fforward
d operation
i
• Proactive configuration
• RSU-assisted Peer-to-Peer
• Backbone-assisted Peer-to-Peer
Challenges
• Long duration information download
RSU
• Dynamic Vehicle Position
• Frequent Disconnection of Links
RSU
The Fully Networked Car
Geneva, 3-4 March 2010
Some Related Work
4
Geographical Opportunistic Routing for Vehicular Networks
(GeOps)
o Approach
• Exploits geographic information from vehicle navigation system to deliver
message to particular location.
— Given the destination location, the sending vehicle (carrier) compute the
nearest point on its route to the destination.
— Based on nearest point it will calculate the minimum estimated delivery time to
th d
the
destination
ti ti b
by using
i navigation
i ti system.
t
Each vehicle will compute the minimum estimated delivery time
based on its route. And vehicle C will have minimum estimated
delivery time and choose to be carrier of message.
1
2
Ve c e C will
Vehicle
w choose
c oose aanother
ot e ca
carrier
e to ttransmit
a s t tthee message
essage to
the destination.
2
University College, London 2008
The Fully Networked Car
Geneva, 3-4 March 2010
1
Some Related Work
5
DTN Protocol
Destination
Applications
Choose forwarding that will take message close to
destination based on navigation system.
Fixed single
point
Not specified
Use Geo-graphic routing and switch to DTN mode
when network partitions are detected
GeoDTN Nav
GeoDTN+Nav
Determine delivery quality metric based on
navigation system to choose best neighbor
Fixed single
point
i t
Same as above
Fixed single
point
Not specified
(Packet routing)
GeOpps
VADD
Use predictable vehicle mobility to choose best
carry-and-forward neighbor.
Context-driven Use context information to optimize and
DTN
disseminate content
Mobile
Driving assistance
multiple points
o Context-driven DTN (C-DTN):
• Networking protocol operations driven by context
— In contrast to a generic mobile backbone to achieve end
end-to-end
to end networking
via multi-hop dissemination
The Fully Networked Car
Geneva, 3-4 March 2010
Design Approach of Context-driven DTN (C-DTN)
Context Identification
o Focus on individual vehicle application sessions and
requirements
o Leverage context to develop technology for disruption
tolerant network.
o Context consists of
• Temporal scope
— Content expiration, (Packet expiration)
• Spatial
p
scope
p
— Content position, direction, trajectory, area (region of interest,
posting area)
• Driving Attributes
— Vehicle information including trajectory (driving plan)
o
Used for estimation of future locations
— Communication pattern
o
End-to-end, end-to-multiple ends, end-to-region, etc.
— Etc.
The Fully Networked Car
Geneva, 3-4 March 2010
6
Information Flow in System Components
Context-driven DTN
Content
RSU
Content querying
RSU
Context
generator
Content
querying
Infrastructure
Content generation
Content replying
Context handling
Content posting
Disruption Tolerant
Network
The Fully Networked Car
Geneva, 3-4 March 2010
7
Logical Stack Components of C-DTN
Applications
Context Handling Sub-layer
Context Generation
Module
Context Information
Module
Content Optimization
Sub-layer
Sub
layer
Detection & Condition
Module
DTN Forwarding
Module
DTN Sub-layer
Packet IN
The Fully Networked Car
Geneva, 3-4 March 2010
P k t OUT
Packet
8
Evaluation Scenarios
9
Overview
o
o
o
o
Scenario # 1
Content Source: static single node
Content Recipients: static intersections
Effect of variation in vehicle density and mobility
on performance
D
518
o Scenario # 2
o Content Source: multiple intersections
o Content Recipient: mobile vehicles
o Effect of variation in vehicle density and
destination mobility on performance
S
S
S
S
D
519
513
S
Sender
D
Destination locations
D
D
514
Carrying vehicle
Receiving Vehicle
S
501
Application Parameter
Value
Cumulative Packet Rate
1300 B pkts @ 1 packet/sec
Radio Range
100m
Duration of simulation
600 seconds
Speed distribution
Uniform [24-96] Kmph
The Fully Networked Car
Geneva, 3-4 March 2010
Scenario Evaluation of C-DTN
10
Packet Delivery Ratio and Average Delay
Scenario 2 PDR from Source 519
1
1
0.9
0.9
0.8
0.8
0.7
0.7
Delivery Ratio
Delivery Ratio
Scenario 1 PDR for #519
0.6
0.5
519 for 1 way; 1 lane
0.4
519 for 1 way;2 lane
0.3
519 for 2 way; 2 lane
0.6
0.5
0.4
0.3
0.2
0.2
0.1
0.1
0
519 for 1 way; 1 lane
519 for 1 way; 2 lane
519 for 2 way; 2 lane
0
0
5
10
15
20
25
0
5
10
Vehicle Density (vehicles per mile)
•
15
20
25
Vehicle Density (vehicles per mile)
Comparison for PDR at #519 (farthest from source)
•
Scenario 1 Delay for #519
Comparison of PDR at mobile vehicle 1 from source #519
Scenario 2 Delay for Source 519
400
400
350
519 for 1 way; 1 lane
519 for 1 way; 2 lane
519 for 2 way; 2 lane
300
350
300
2 0
250
Seconds
Seconds
250
200
150
200
519 for 1 way; 1 lane
519 for 1 way; 2 lane
519 for 2 way; 2 lane
150
100
100
50
50
0
0
0
5
•
10
15
20
Vehicle Density (Vehicles per mile)
The Fully Networked
Car
Comparison
delay2010
at #519 (farthest from source)
Geneva,
3-4for
March
25
0
5
10
15
20
25
Vehicle Density (vehicles per mile)
•
Comparison of delay at mobile vehicle 1 from source #519
Scenario Evaluation of C-DTN
Freshness
o
V hi l d
Vehicle
density
it = 5 vehicles
hi l per mile
il
2 way; 2 lane
2 way; 2 lane
Sent by source
•
•
Received at #519
DTNMP #519 (farthest from source) receives
from content source
Distinct instances of content reception
The Fully Networked Car
Geneva, 3-4 March 2010
•
•
Mobile destination receives from farthest
content source
Mobile destination receives content more
frequently due to correlated mobility
11
Summary
o Overviewed general DTN approaches
o Proposed a context
context-driven
driven DTN (C
(C-DTN)
DTN) that supports
networking in consideration of application context needs
o Evaluations to validate C-DTN dissemination performance
with realistic mobility
• Diversity in mobility pattern generally improves performance
• Under correlated mobility model, DTN dissemination performance
depends on a number of factors
The Fully Networked Car
Geneva, 3-4 March 2010
12
B k
Backup
Slides
Slid
The Fully Networked Car
Geneva, 3-4 March 2010
Some Related Work
14
GeoDTN+Nav: Hybrid Geographic and DTN Routing with Navigation
Assistance
o Approach
• Start forwarding in greedy forwarding (shortest path) mode and switch to DTN
when network is disconnected
— Proposed score function to detect the network disconnection.
• Compute delivery quality metric for each neighbor and choose the best quality
node to forward the packet.
— Ex.
Ex Compute d1,
d1 d2 and d3 for 3 neighbors and choose best (d1) node
node. (below)
Proposed Vehicle Navigation Interface
(VNI) for these vehicles.
d2
d1
d3
Univ. Of California, Los Angeles, 2008
The Fully Networked Car
Geneva, 3-4 March 2010
Some Related Work
15
Vehicle-Assisted Data Delivery (VADD) Protocol
o
Approach
• Estimates packet delivery delay on each roadway.
• Select forwarding direction using a priority list based on probabilistically
shortest delay path computed in a centralized manner.
• Source and destination assumed to be in a connected graph
Geographically shortest path
(traditional)
1
2
Penn State University, PA, 2005
The Fully Networked Car
Geneva, 3-4 March 2010
Fast speed wireless
communication path (VADD)