UAV-UAV NETWORK

advertisement
SATELLITE
COMMS
SURVEILLANCE
MISSION
SURVEILLANCE
MISSION
UAV-UAV NETWORK
AIR-TO-AIR
MISSION
STRIKE
MISSION
COMM/TASKING
Unmanned
Control Platform
COMM/TASKING
COMM/TASKING
RESUPPLY
MISSION
UAV-UGV NETWORK
FRIENDLY
GROUND CONTROL
(MOBILE)
Manned
Control Platform
Ad Hoc networks today
The Challenges of the battlefield
scenario
• Large scale – thousands of nodes
• Varying mobility patterns
– Group mobility
– Coordinated motion vs random motion
• Heterogeneous traffic
– Unicast; multicast
– Best effort; real time; reliable delivery;
• Unpredictable, hostile environment
– Jamming, interference; propagation effects
– Can jam your GPS
Scalable Routing using Group Motion
and Landmark Overlay
• Main assumption: nodes move in groups
• Landmark dynamically elected in each group
• Each node maintains two sets of routes:
– Routes to all neighbors up to K hops away
– Routes to all Landmarks
Landmark
Logical Subnet
Landmark Overlay Routing
 A packet to “local” destination is routed
directly using local tables
 A packet to remote destination is routed to
corresponding Landmark
 Destination has No landmark? Flood search
with AODV
• Benefits: low storage, low control O/H
Landmark
Logical Subnet
Delivery Ratio
LANMAR-DSDV
LANMAR-FSR
OLSR
FSR
DSDV
LANMAR-OLSR
LANMAR and the Mobile Backbone Network
(MBN)
• In the flat network, paths with many hops
(say > 6) degrade performance
• Solution: Mobile Backbone Network
• LANMAR extends “transparently” to the
MBN
Mobile Backbone Network
Future “commodity” ad hoc networks
Question: will commodity ad hoc nets look like the battlefield
networks?
• Military and civilian (disaster recovery) and hoc networks
are motivated by:
–
–
–
–
Instant deployment
Lack of infrastructure
Very specialized mission/function
Cost not most critical issue
• Future “commodity”, commercial ad hoc networks will not
grow in the same way:
– Cost is an issue (eg, ad hoc vs W-LAN vs 2.5 G)
– Connection to Internet is desirable (sometimes, a “must”)
– Multipurpose networking
• Enter “opportunistic ad hoc networking”
Opportunistic ad hoc networks
“Opportunistic extensions” of the infrastructure:
• Extend the home Wireless LAN with an ad hoc net to reach
the backyard
• Use ad hoc techniques to build infrastucture in remote
villages
• Daisy-chain friends to a 2.5G connection
• Interconnect patients in a hospital with an ad hoc LAN
(must be low power to avoid interference)
• Connect cars in the city with an urban ad hoc grid (for
driving safety, environment awareness, kid games, etc)
Urban “opportunistic” ad hoc networking
From Wireless to
Wired network
Via Multihop
Highway Network
Hot Spot
Hot Spot
The Urban Grid
Opportunistic urban grid networking is fueled by:
• Mesh networks
• DSRC – navigation safety network
• 2.5 and 3G presence in the vehicles
• Programmable, SDR radio technology
These complementary technologies are synergized
(benefit from each other) in the opportunistic ad
hoc networking model
Urban Grid Applications
•
•
•
•
•
Opportunistic packet piggybacking (eg, mules)
Car – 2 – car network games
The “sensor grid”
Fragmented, cooperative downloading
Recovery from massive infrastructure failures
(eg, natural disaster, power blackout, terrorist
attack)
Hot Spot
Hot Spot
STOP
Power
Blackout
Hot Spot
Hot Spot
STOP
Power
Blackout
P2P Collaborative Downloading
Hot Spot
P2P link
Network Link
Video File
Urban Sensor Grid
sensor
Technical Challenges in the Urban Grid
• Routing
– Geo forwarding (MIT GRID project)
– “last encounter” routing
– Group motion assisted routing
• Mobile sensor data retrieval
• Smooth handoff between media alternatives
Smooth handoff
• Suppose the car is receiving a live soccer game
from the mesh network (at 11Mbps)
• When the car gets out of range of the mesh
network it switches to 1xRTT (at 100Kbps) or to
UMTS (at 384 Kbps)
• The server or proxy in the Internet must be
alerted of the change in capacity, in order to
change video resolution or even video content.
Approach:
• The server continuously monitors the “min
capacity” on the path to destination
Packet Pair Dispersion
• Previous Min Cap estimate based on:
– Packet Pair transmission
– “packet dispersion” measurement
• Packet Dispersion = separation of the Pair
Previous Work
• Packet Pairs
– Send a large number of PPs; draw capacity distribution
– highest mode of the capacity distribution is the CAPACITY
estimate
– However, PP capacity estimate can be “multimodal” (Paxton)
• Dovrolis’ proposed approach
– First send packet pair
– If multimodal, send packet trains
• Still not satisfactory:
– Most techniques too complicated, time/bw-consuming,
inaccurate and sensitive to choice of parameters
– Never tested on wireless
Previous work bugged down by cross traffic!
• Cross-traffic (CT) serviced between PP packets
T
Cross
Traffic
T’ > T
Narrow
Link
• This leads to under-estimation of Capacity
Key Observation
( Rohit Kapoor, Phd 2003, UCLA)
• Expansion or Compression errors occur
when one of the two PP packets is queued
• Thus, the estimate is correct if there is NO
queueing delay
• Approach:
– Carry out several Packet Pair trials
– Monitor the sum of delays of PP packets
– When the sum is minimum the estimate is the
Bandwidth!
Cap probe
• PP sample provides two types of information
– Dispersion of packets
– Delay sum of packets
• CAP probe combines the two:
– From dispersion, compute capacity estimate for each pair
– From delay sum -> find min sum and thus correct capacity
Minimum Delay Sum (sec)
0.005
0.0045
0.004
0.0035
0.003
Capacity
0.0025
0.002
0.0015
0.001
0.0005
0
0
1.6
3.2
4.8
6.4
Bandw idth Estim ates (Mbps)
8
Simulations
• 6-hop path: capacities {10, 7.5, 5.5, 4, 6, 8} Mbps
• PP pkt size = 200 bytes, CT pkt size = 1000 bytes
• Persistent TCP Cross-Traffic
Bandwidth Estimate
Frequency
0.01
Cross Traffic Rate
0.009
1Mbps
2Mbps
4Mbps
0.008
0.007
1
1Mbps
2Mbps
4Mbps
0.7
0.005
0.004
0.6
0.5
0.4
0.003
0.3
0.002
0.2
0.001
0.1
0
0
1.6
3.2
4.8
6.4
Bandw idth Estim ate (Mbps)
Cross Traffic Rate
0.8
0.006
0
Over-Estimation
0.9
Frequency
Min Delay Sums (sec)
Minimum Delay Sums
8
0
1.6
3.2
4.8
6.4
Bandw idth Estim ate (Mbps)
8
Laptop1
PING Source/
Destination
Wireless Measurements
802.11b
Access Point
Internet
802.11b
Connectivity
Laptop2
Cross-Traffic
• Results for 802.11b, with Bluetooth TCP cross-traffic
• Note: max 802.11b throughput (with IP) = 6Mbps
Experiment No.
1
2
3
4
5
Capacity
Estimated by
CapProbe (kbps)
5526.68
5364.46
5522.26
5369.15
5409.85
Capacity Estimated
by strongest mode
(kbps)
4955.02
462.8
4631.76
5046.62
449.73
Conclusions
• Early commercial ad hoc nets will be opportunistic
extensions of the infrastructure
• The first large scale example will be the urban grid
• New urban grid applications and protocols are emerging,
very different from the classic ad hoc network problems
• A recurring problem will be the integration of wired and
wireless paths; example: vertical handoff (and capacity
estimation)
• Rapid transition from operational to emergency (fully ad
hoc) mode will be critical.
• Another challenge will be the effective use of the “mobile”
(vehicle and pedestrian) sensor platform
The End
Thank You!
Download