Broadcast Reception Rates and Effects of
Priority Access in 802.11-Based Vehicular
Ad-Hoc Networks
Marc Torrent-Moreno, University of Karlsruhe
Daniel Jiang, DaimlerChrysler RTNA, Inc.
Hannes Hartenstein, University of Karlsruhe
VANET‘04
Marc Torrent-Moreno
Universität Karlsruhe (TH)
Problem statement
Broadcasts
emergency
message
What is the
probability of
reception for this
car?
But:
»
»
»
All vehicles send data → contending for the channel
Hidden terminal problem
Channel characteristics
VANET‘04
Marc Torrent-Moreno - 2
Universität Karlsruhe (TH)
Channel modeling: the standard way …
Unit Disc Graph Model
If no interference:
• Cars inside communication
c.r.
range receive the packet
• Cars outside communication
range do NOT receive the
packet
VANET‘04
Marc Torrent-Moreno - 3
Universität Karlsruhe (TH)
In reality …
Reality
If no interference:
• Cars inside “communication
range” CAN receive the packet
“c.r.”
• Cars outside “communication
range” MAY receive the packet
Reception Probability (d)?
VANET‘04
Marc Torrent-Moreno - 4
Universität Karlsruhe (TH)
Structure of this talk
»
VANET – Key Issues
»
802.11-Priority Access applied to VANETs
»
Simulation Setup
»
Results: Basic Scenario (understanding the mechanism)
»
Results: Dynamic Scenario (getting closer to reality)
»
Summary
»
Future Work
VANET‘04
Marc Torrent-Moreno - 5
Universität Karlsruhe (TH)
VANET – Key Issues
VANET differs from usual ad hoc network by its vehicular
environment, distributions, movement and applications
– Key usage: safety applications → broadcast
– There is a deployment problem but eventually there will be
saturation conditions (high penetration rates)
»
What is the probability of reception of a broadcast message at
an arbitrary distance to the sender?
»
How to give priority access and an improved reception rate
for important warnings?
»
How are the above two results affected by signal strength
fluctuations caused by radio channel fading?
VANET‘04
Marc Torrent-Moreno - 6
Universität Karlsruhe (TH)
IEEE 802.11
»
»
Why 802.11? Interests in U.S.A. (DSRC/WAVE), Europe (Fleetnet/NOW)
»
»
DIFS = SIFS + 2*SlotTime
»
Broadcast → no RTS/CTS, no ACK, CW = CWmin
Access method: CSMA/CA
(SIFS = 32us, SlotTime = 13us)
CWmin = 15
VANET‘04
Marc Torrent-Moreno - 7
Universität Karlsruhe (TH)
Priority Access
»
Mechanism of EDCA of IEEE 802.11e
»
Two main changes:
– AIFSD[AC] = SIFS + AIFS[AC]*SlotTime
– CWmin[AC]
VANET‘04
AC
CWmin[AC]
AIFS
0
CWmin
2
1
CWmin
1
2
(CWmin+1)/2 -1
1
3
(CWmin+1)/4 -1
1
Marc Torrent-Moreno - 8
Universität Karlsruhe (TH)
Simulation Set Up
»
Tool: Network Simulator ns-2
»
Changes to original code:
– Some bug fixing at 802.11 module
– Adjust values to 802.11a at 5.9GHz with 10MHz Channels
(DSRC/WAVE)
– Changes in both MAC and Physical Layer
– Nakagami radio propagation model
VANET‘04
Marc Torrent-Moreno - 9
Universität Karlsruhe (TH)
Basic Scenario (static and two-ray ground)
»
»
How CSMA/CA works in a saturated broadcast scenario?
How AIFS and CWmin affect broadcast performance?
- 8 lanes with 75 static cars
- 20m between cars
- 500 bytes packets
- 10 packets/s
- 200m comm. range
- “Receiver” at 100m
4m
1500 m
VANET‘04
Marc Torrent-Moreno - 10
Universität Karlsruhe (TH)
Basic Scenario – Results
Result
Analysis
AIFS/CW
RcvPkts
SntBT1
RcvBT1
SntBT0
RcvBT0
2/15
27,7%
69.3%
22.6%
5.8%
67.2%
2/7
31.1%
66.8%
22.1%
12.0%
66.6%
1/7
54.5%
46.4%
71.1%
11.9%
76.4%
1/3
59.4%
45.8%
71.0%
26.6%
78.9%
• Car with highest priority: best probability of reception
»
Why is a packet not received?
– Hidden node
– Car inside range chooses the same time slot to send
»
For error analysis new parameters were needed: SntBTx, RcvBTx
VANET‘04
Marc Torrent-Moreno - 11
Universität Karlsruhe (TH)
Dynamic Scenario (mobile and TRG or Nakagami)
1
50m distance
100m distance
200m distance
Distribution
0.8
0.6
0.4
700 m
0.2
0
-92
-90
-88
-86
-84
-82
-80
-78
Received power (dBm)
-76
-74
-72
-70
Nakagami received power distribution
- 8 lanes, 2 direcctions (912 cars)
- Comm. Range: 100m, 200m
- Constant speed different lanes
- Packet size: 200B, 500B
- Distance between cars: 20m
- 10 pckts/s
VANET‘04
Marc Torrent-Moreno - 12
Universität Karlsruhe (TH)
Dynamic Scenario – Metrics
»
Safety Applications
Delay
Broadcast Scenario
Effectiveness
Metrics:
– Channel access time
– Probability of reception
VANET‘04
Marc Torrent-Moreno - 13
Universität Karlsruhe (TH)
Dynamic Scenario – Channel Access Time
Scenario
»
Channel Access Time
Com. Range
Pckt Size
Load Mbps
Priority
Non-Priority
100m
200 B
1.28
0.4 ms
0.9 ms
100m
500 B
3.2
1.6 ms
4.8 ms
200m
200 B
2.56
1.2 ms
3.9 ms
200m
500 B
6.4
3.6 ms
16.4 ms
200m (Nak)
500 B
6.4
9.0 ms
26.5 ms
Clearly the Priority node has always shorter Channel Access Time !
VANET‘04
Marc Torrent-Moreno - 14
Universität Karlsruhe (TH)
Dynamic Scenario – Probability of Reception
Nakagami
Two Ray Ground
1
1
priority
non-priority
0.8
Probability of Reception
Probability of Reception
0.8
priority
non-priority
0.6
0.4
0.2
0.6
0.4
0.2
0
0
0
50
100
150
200
250
Distance (m)
»
»
VANET‘04
0
50
100
150
200
250
Distance (m)
500 bytes packets
200 m intended communication range
Marc Torrent-Moreno - 15
Universität Karlsruhe (TH)
Effect of Realistic Propagation Model - Metrics
»
With Non-Deterministic Radio Propagation model:
– Channel Access Time
– Probability of Reception
(worse effect on priority node)
WHY !?
two-ray-ground
Nakagami
Probability of Reception
1
0.8
0.6
New metrics:
• Sensed Packets per second
• Channel Idle Time Ratio
0.4
0.2
0
0
VANET‘04
100
200
300
Distance (m)
400
500
600
Marc Torrent-Moreno - 16
Universität Karlsruhe (TH)
Effect of Realistic Propagation Model - Results
Scenario
Priority
Non-Priority
200m, 500B
TRG
Nak
TRG
Nak
Ch.Acc.Time
3.6 ms
9.0 ms
16.4 ms
26.5 ms
Sens. Pkts/s *
3325.2
3093.2
3324.6
3096.8
Ch. Idle Time
10.8%
4.4%
10.6%
4.4%
* Same average number of packets sent by all nodes
»
VANET‘04
Less packets sensed but less Channel Idle Time with
non-deterministic radio propagation model !!!
Marc Torrent-Moreno - 17
Universität Karlsruhe (TH)
Summary
»
Target: reliable reception of important messages in a saturated
broadcast network (VANET):
– Probability of reception and channel access time
– Prioritization
– Received power fluctuations
»
Results
– When saturation occurs, nodes experience severe performance degradation
(probability of reception can be as low as 20% at half com. range)
– Priority access provides improvement in both channel access time and
probability of reception (AIFS and CW can be adjusted to reduce channel
access time and collisions)
– Non-deterministic radio model degrades performance of both types of
nodes (worse effect on probability of reception of the prioritized ones)
»
Conclusions
– Need better understanding of distributed MAC on global scale
– Priority access methods as well as relay/repetition strategies are very
important aspects of future safety-critical applications
VANET‘04
Marc Torrent-Moreno - 18
Universität Karlsruhe (TH)
Future Work
»
More than 1 prioritized nodes
»
Relay/Repetition strategies
»
Topology control
»
Realistic radio models
»
Spatial and temporal correlations
»
Traffic modeling
VANET‘04
Marc Torrent-Moreno - 19
Universität Karlsruhe (TH)
Thank you very much for your attention !
VANET‘04
Marc Torrent-Moreno
Universität Karlsruhe (TH)