Mobile and Vehicular Network Lab
Vehicular Networks –
What can communications do for vehicles?
Prof. Michael Tsai
2015/6/5
Mobile and Vehicular Network Lab
Future cars
• GM EN-‐V video
2
Mobile and Vehicular Network Lab
Vehicle Talks.
• Safety related
– Warn you about an accident ahead (emergency brake)
– Tell you a component in your car is about to fail
• Energy related
– Provide routing to avoid a traffic jam
– Smart traffic light (or, “no traffic light”)
• Information and entertainment (“Infotainment”)
– Advertisement
– Electronic toll collection
– Social networking for passengers – Internet connection
3
Mobile and Vehicular Network Lab
How do they talk?
• “Blackbox” (On-‐Board Unit, OBU) with wireless radios
• Direct Vehicle-‐to-‐Vehicle (V2V) Communications
– IEEE 802.11p: designed for wireless access in vehicular environments (WAVE)
– IEEE 1609: Higher layer protocols and operations – Dedicated Short Range Communications (DSRC):
U.S. standard for vehicular communications
• Via Pre-‐deployed Infrastructure (V2I)
– Cellular Networks, e.g., Long Term Evolution (LTE)
– Road-‐side Unit (RSU) with V2V radio
4
Mobile and Vehicular Network Lab
Vehicular Networks
5
Mobile and Vehicular Network Lab
DSRC/WAVE/802.11p
• PHY layer almost identical to IEEE 802.11a
• OFDM using BPSK/QPSK/16 QAM/64 QAM
• Reduced inter symbol interference (multipath effects and Doppler shift)
– Doubled timing parameters (double the time durations) – Channel bandwidth (10 MHz instead of 20 MHz)
• Reduced throughput (3 ... 27 Mbit/s instead of 6 ... 54 Mbit/
s)
• Communication range of up to 1000 m
• Vehicles’ velocity up to 200 km/h
– MAC layer with extensions to IEEE 802.11a – Randomized MAC address
– QoS (Priorities, see IEEE 802.11e, ...)
• Support for multi-‐channel and multi-‐radio
• New ad hoc mode 6
Mobile and Vehicular Network Lab
DSRC/WAVE/802.11p
7
(a) WAVE protocol stack
(b) DSRC channel allocation
Mobile and Vehicular Network Lab
Change the world (on the road)
• What we will talk about today:
0. Google driver-‐less car (autonomous vehicle) 1. Increase highway capacity (efficient)
2. Avoid scooter accidents (safe)
3. Business perspective (whether it will happen)
8
Mobile and Vehicular Network Lab
Blinds can drive
• Google car video
9
Mobile and Vehicular Network Lab
What is in Google Car
Cost:
• Estimated total:
NT$ 9 M
• 3D LIDAR (laser scanner): over NT
$2 M
• Video cameras: a few hundred NT$
• Radar sensors: a few thousands NT
$
• Processor &
others
10
Mobile and Vehicular Network Lab
Some Statistics
• In April 2014, the team announced that they have completed over 300,000 autonomous-‐driving miles (500,000 km) accident-‐free.
• 4 U.S. States, Nevada, Florida, Michigan and
California, have passed laws that permit driverless cars on the road as of December 2013.
11
Mobile and Vehicular Network Lab
But Google Driverless Cars do not talk (to other cars).
12
Mobile and Vehicular Network Lab
Change the world (on the road)
• What we will talk about today:
0. Google driver-‐less car (autonomous vehicle) 1. Increase highway capacity (efficient)
2. Avoid scooter accidents (safe)
3. Business perspective (whether it will happen)
13
Mobile and Vehicular Network Lab
http://virginiaits.blogspot.com/
http://2howto.com/how-‐to-‐cope-‐with-‐ever-‐
rising-‐traffic-‐congestion-‐in-‐india/
http://www.csee.umbc.edu/courses/
undergraduate/FYS102D/streetsmart.pdf
Preventing Congestion
14
Mobile and Vehicular Network Lab
http://www.permatopia.com/
wetlands/traffic.html http://semesterinthesouth.blogspot.com/
2008_04_01_archive.html
http://english.peopledaily.com.cn/200701/05/
eng20070105_338437.html
Preventing Congestion
15
Mobile and Vehicular Network Lab
Increase Road Efficiency
• Wugu-Yangmei Overpass (just completed in 2013)
(for National Freeway No. 1)
– Cost: 88.2B TWD (3B USD) for 40 km of elevated road
- The most expensive road per km in Taiwan history – Divert approximately 25% of traffic from the original
highway
– Improve rush hour average speed:
- 40-50 kmph à 80-90 kmph
Mobile and Vehicular Network Lab
Increase Road Efficiency
• Maximum highway capacity per lane = 2,200 vehicle/hr
• At 100 kmph, the road surface utilization = 5%
• Automatic steering à reduce lane width
• Longitudinal control (adaptive cruise control)
à reduce the gap
• Drivers’ response delays cause stop and go disturbances
• Only possible through cooperation with communications
45.5 m
5 m
3. 5 m 1.8 m
Mobile and Vehicular Network Lab
Reduce Energy Consumption
• At highway speeds, HALF of energy is used to overcome aerodynamic drag
– Close-formation automated platoons can save 10% to 20% of total energy use
• Acceleration/deceleration cycles waste energy (and produce excess emissions)
– Automation can eliminate stop-and-go disturbances
• Again, only possible through
cooperation
Mobile and Vehicular Network Lab
Platooning
• Small gap between cars in a platoon
– Gentle impacts between vehicles in faulty cases
• Large gap between platoons
– Prevent severe impacts in faulty cases
• Platoons enable lane capacity to be doubled or tripled, while maintaining safety
– 6000 – 8000 cars/hour in 10-‐car platoons – 1500 heavy trucks/hour in 3-‐truck platoons
19
Mobile and Vehicular Network Lab
Back to “what’s wrong with Google Car”
• Cooperation Can Augment Sensing
– Autonomous vehicles are “deaf-‐mute”
– Cooperative vehicles can “talk” and “listen” as well as
“seeing”
• Communicate performance and condition data directly rather than sensing it indirectly
– Reduce uncertainties – Reduce filtering lags
– More sources of information available, including beyond line of sight
• Expand performance envelope – capacity and ride quality
20
Mobile and Vehicular Network Lab
Change the world (on the road)
• What we will talk about today:
0. Google driver-‐less car (autonomous vehicle) 1. Increase highway capacity (efficient)
2. Avoid scooter accidents (safe)
3. Business perspective (whether it will happen)
21
Mobile and Vehicular Network Lab
What is available in today’s car?
22
Mobile and Vehicular Network Lab
1. Crash – post crash 2. Crash is unavoidable 3. Crash may occur
4. Risk has appeared
5. Risk has not yet appeared
Risk
2 1 4 3
5
Larger Shield is SAFER!
For :
• Blind Spot Information System (BLIS)
• Adaptive Cruise Control (ACC)
• Forward collision warning
• Pedestrian detection
• Lane departure warning
We are here.
What can we do here?
The “Safety Shield”
5 4 3 2 1
4
23
Mobile and Vehicular Network Lab
Overcome Visual Obstacles
Conventional Approach:
Video camera and/or radar only Cooperative approach:
Sensors + Communications
Scooter A
1 2
3
4
1
2 4
3 Detects only risks with a LOS path Detects also risks
via comm. / nLOS path
3 4
Line-‐Of-‐Sight blocked by a bus
Line-‐Of-‐Sight blocked by the road corner
2 1 I’m merging into
the middle lane!
I’m turning right!
Be careful to the car on my left!
24
Mobile and Vehicular Network Lab
Reduce Uncertainty
• Measurements of location/velocity always have a level of uncertainty
• Self-‐information is the most accurate.
• Observation from nearby vehicles is more accurate than that from distant vehicles.
3 1
2
1
2
3
Location Uncertainty:
Possible Location
(e.g., vision-‐based localization)
25
Mobile and Vehicular Network Lab
Bring Down the Cost
Individual vehicle does not need to have total sensor coverage.
EXPENSIVE
Do we need every sensor in every car? Cost down. Energy saving.
CHEAP
26
Mobile and Vehicular Network Lab
1. Crash – post crash 2. Crash is unavoidable 3. Crash may occur
4. Risk has appeared
5. Risk has not yet appeared
Risk
2 1 4 3
5
• Blind Spot Information System (BLIS)
• Adaptive Cruise Control (ACC)
• Forward collision warning
• Pedestrian detection
• Lane departure warning
Next-‐generation Vehicle Safety
Vehicle-‐to-‐Vehicle technologies eliminate the “gap”, and make these solutions
applicable to as well!
5 4 3 2 1
5 27
Mobile and Vehicular Network Lab
Scooters in Taiwan
*http://www.epochtw.com http://upload.wikimedia.org 28
• In Taiwan, 68.35% of registered vehicles are scooters or motorcycles.
• Every 1.56 persons own a scooter.
Mobile and Vehicular Network Lab
Scooters Around the World
29
Country Number of
Motorcycle/Scooter Number of Population per Scooter
Taiwan 14,844,932 1.56
Malaysia 9,443,922 2.95
Vietnam 25,414,689 3.38
Thailand 17,229,814 3.88
Indonesia 52,433,132 4.47
Italy 9,425,098 6.39
Spain 4,958,879 9.26
Switzerland 806,577 9.60
Japan 12,477,417 10.22
Czech 903,346 11.61
Austria 712,092 11.75
China 100,004,714 13.35
Netherlands 1,228,058 13.46
Brazil 13,088,074 14.63
USA 7,929,724 38.72
United Kingdom 1,433,124 43.13
Mobile and Vehicular Network Lab
Scooter Accidents in Taiwan
• Scooter passengers account for most deaths/injuries
• The percentages and the actual numbers are both rising!
2003 2004 2005 2006 2007 2008 2009 2010 60
70 80 90 100
Year Percentage of Scooter Passenger Contribution (%)
1359 1361
1515
1793 1487 1310 1228 2089 2270 2548 1267
2708 2325 2152 2038 2110 119577 138720 157128 166389 174298 184616 201444 243089
Deaths occur within 24 hrs Deaths occur within 30 days Injuries
30
Mobile and Vehicular Network Lab
Safety Features in
Off-‐the-‐Shelf Vehicle Product
Car Scooter
Active Safety Feature 1. Anti-‐lock Breaking System (ABS)
2. Electronic Brake-‐Force Distribution (EBD)
3. Electronic Stability Control (ESC) / Traction Control System (TCS)
1. Anti-‐lock Breaking System (ABS)
2. Traction Control System (TCS)
3. Power mode map
Passive Safety Feature 1. Seat belts 2. Body frame 3. Head restraints
4. Front air bags (driver passenger)
5. Side curtain air bags
1. Full-‐face helmet 2. Leather suit (jacket,
gloves, pants) 3. Boots
Not common!
Open-‐face helmet only Not much development! 31
Mobile and Vehicular Network Lab
Advantages of Using a Smartphone
Useful Components
Multi-‐touch Screen
Powerful CPU
Useful Sensors:
GPS
Accelerometer
High initial market penetration rate
Gyro
• Lower effect cost for users
• 2012 smartphone market share:
• Existing Supporting Infrastructure
42% 44%
Good user experience
• Know about the user behaviors in other context as well
• Already familiar with the smartphone’s HCI
32
Camera
WiFi Cellular and DSRC (Qualcomm)
Mobile and Vehicular Network Lab
Classification of
Causes of Fatal Accidents
• Of the 3,209 deaths in accidents in 2010, top 4 causes of accidents are:
• Drive Under Influence (drunk driver) – 17.82%
(572)
• Did not yield to others – 16.45% (528)
• Lost of attention to the road – 15.39% (494)
• Violation of traffic signals – 7.35% (236)
33
What can be done to avoid these accidents?
(and save over 1,000 lives per year!)
Submission
doc.: IEEE 802.11-13/0541r1
Safety Applications – V2V
Forward Collision Avoidance FCA Emergency Electronic Brake Lights EEBL
Blind Spot Warning BSW
Lane Change Assist LCA
Do Not Pass Warning DNPW
Intersection Collision Warning ICA
Wrong Way Driver Warning WWDW Cooperative Adaptive Cruise Control CACC
Examples Follow:
Slide 34 John Kenney, Toyota Info Technology Center
May 2013
Submission
doc.: IEEE 802.11-13/0541r1
V2V Safety Use Case
If driver of approaching car does not stop, or slow appropriately, warning issued within car.
Slide 35 John Kenney, Toyota Info Technology Center
May 2013
DSRC communication
Stopped
Car Approaching
Car
Forward Collision Warning (FCW)
Submission
doc.: IEEE 802.11-13/0541r1
V2V Safety Use Case
High deceleration by car approaching jam. Trailing car Informed via DSRC within 100 msec.
Slide 36 John Kenney, Toyota Info Technology Center
May 2013
Emergency Electronic Brake Lights (EEBL)
Traffic Jam
Submission
doc.: IEEE 802.11-13/0541r1
V2V Safety Use Case
Slide 37 John Kenney, Toyota Info Technology Center
May 2013
Normal driving –
advisory indicator of car in blind spot Driver receives warning
when showing intent to change lanes
Blind Spot Warning (BSW)
Note: Specific timing, format, or decision logic for advisories and warnings will likely vary for each car manufacturer
Submission
doc.: IEEE 802.11-13/0541r1
V2V Safety Use Case
When showing intent to move to oncoming lane, driver receives warning if not safe to pass.
Slide 38 John Kenney, Toyota Info Technology Center
May 2013
Do Not Pass Warning (DNPW)
Oncoming traffic
Submission
doc.: IEEE 802.11-13/0541r1
V2V Safety Use Case
If intersecting trajectories are indicated, driver is warned.
Slide 39 John Kenney, Toyota Info Technology Center
May 2013
Building: Leads to Non-Line Of Sight (NLOS) communication
Intersection Collision Warning (ICA)
Submission
doc.: IEEE 802.11-13/0541r1
Safety Applications – V2I
Applications enabled by SPaT:
Red Light Running RLR
Left Turn Assist LTA
Right Turn Assist RTA
Pedestrian Signal Assist PED-SIG
Applications enabled by Signal Request Message (bi-directional communication):
Emergency Vehicle Preempt PREEMPT
Transit Signal Priority TSP
Freight Signal Priority FSP
Rail Crossing RCA
Examples follow:
Slide 40 John Kenney, Toyota Info Technology Center
May 2013
Submission
doc.: IEEE 802.11-13/0541r1
Safety Use Cases: SPaT and MAP
Slide 41 John Kenney, Toyota Info Technology Center
May 2013
Messages sent from RSE:
• Signal Phase and Timing (SPaT) - dynamic
• MAP (intersection geometric description) - static RSE may also send GPS Correction Data
Application Types:
RLR LTA RTA TSP FSP
PED-SIG
SPaT can also enable non -safety applications like “Green Wave”
Freight Yard Vehicle Without DSRC
Transit Vehicle
Submission
doc.: IEEE 802.11-13/0541r1
Safety Use Case: Work Zone Warning
Slide 42
Grass Divider
up to 1100 ft range
Work Zone Warning Com. Zone
Work Zone
Traffic Cones
RSU
In-Vehicle Display and Annunciation
ZONE AHEAD WORK
May 2013
John Kenney, Toyota Info Technology Center
Submission
doc.: IEEE 802.11-13/0541r1
V2I Safety Use Case: Road Hazard Warning
Slide 43 John Kenney, Toyota Info Technology Center
May 2013
Median
Dynamic Message Sign and Multi-App RSU
Road Condition Warning Com. Zone
Road Sensor Station
Bridge
ICE
Up to 650 ft forward of the Hazard
90 m (300 ft) range
Variations include:
Road Condition (ice), Curve Speed
Low Bridge Roll-over
Roadway Weather (RWIS) In Vehicle Signage Accident Ahead Rock slide, etc.
Submission
doc.: IEEE 802.11-13/0541r1
V2I Safety Use Case: (PREEMPT)
(also used for Transit/Freight Priority)
Slide 44
Emergency Vehicle
~ ~
RSE
up to 1000 m (3281 ft)
~ ~
~ ~
Preempt Transaction
1. DSRC OBE-to-RSE: Vehicle Host Preemption Request 2. DSRC RSE-to-OBE: ACK
3. Emergency Vehicle Host Displays Preempt-ACK within vehicle
DSRC Transaction occurs on Ch. 184 at high power.
OBE
May 2013
John Kenney, Toyota Info Technology Center
Submission
doc.: IEEE 802.11-13/0541r1
V2I Safety Use Case: Standardized Tolls
Slide 45
Open Road Example
Capture Zone
RSE-Equipped Gantry
30 m (98 ft) May 2013
John Kenney, Toyota Info Technology Center
Submission
doc.: IEEE 802.11-13/0541r1
V2I Safety Use Case: RR Grade Crossing
May 2013
John Kenney, Toyota Info Technology Center Slide 46
Train 20-40 sec.
distant
Conventional
RR Grade Crossing Equipped with RSE RSE warning range
increased compared to conventional equipment Can also be used at non -signalized crossings Range up
to 1100 ft
RR Warning Sign
Train 20-40 sec.
distant
Mobile and Vehicular Network Lab
Red-‐light Running:
Infrastructure-‐based Solution
47
A LIDAR or a camera to detect Red-‐Light Runners
Broadcast to warn near-‐by vehicles
Mobile and Vehicular Network Lab
Red-‐light Running:
Smartphone-‐based Solution
48
Rider’s smartphone
detects possible red-‐light running
Broadcast to warn near-‐by vehicles
Mobile and Vehicular Network Lab
MD-‐SVM Results
0 10 20 30 40 50 49
50 55 60 65 70 75 80 85 90 95 100
False positive rate (%)
True positive rate (%)
MD-SVM: 15-45m
va vaG vaGt vaGtr
LADAR-vatr
0 10 20 30 40 50
50 55 60 65 70 75 80 85 90 95 100
False positive rate (%)
True positive rate (%)
MD-SVM: 20-45m
va vaG vaGt vaGtr
LADAR-vatr
90%,90% for 15-‐45m
85%,80% for 20-‐45m
Mobile and Vehicular Network Lab
NLOS BSM: Motivation
50
Implement V2V communications in the scenario
How can we do?
People usually honk the horn, but some problems could occur.
Mobile and Vehicular Network Lab
51
Environments
Location1 Location2 Location3
Mobile and Vehicular Network Lab
52
Minimum Beacon Rate to reach a 99% Successful Rate
V=45km/hr
Beacon size = 64 bytes
V=60km/hr
• nLoS links through the corner buildings often neglected
• Fairly usable! At (30m,30m), path loss ~ 85 dB!
• At 45 km/hr, 5 beacons/s can save 99% of collisions!
Mobile and Vehicular Network Lab
Challenge 1: Scalability
• A 100-m road segment can have:
– More than 600 scooters – More than 120 cars
• Up to 240 km/h relative speed
• Different routes for different vehicle
• Communication requirements:
– Very reliable link – Decent end-to-end
throughput
– Application-dependent delay
Mobile and Vehicular Network Lab
54
“Pairwise Point-‐to-‐Point Links”
for Vehicle-‐to-‐Vehicle and Vehicle-‐to-‐
Infrastructure
EE Times: Top 10 Automotive Agendas for 2014 & Beyond 7. Time to act on V2V,V2I
Mobile and Vehicular Network Lab
Scalability: Interference
Your Car
RF Communication Coverage
VLC Coverage
Utilizing visible light’s property to do
“Automatic neighbor filtering”:
• Zero communication overhead
• Zero delay
àAlways Scalable!
Line-‐of-‐Sight only propagation channel
long range + lower interference
Mobile and Vehicular Network Lab
56
Challenge 2: who is speaking?
My loca(on is (X3, Y3) and my
speed is 98 mph, and I am braking.
My loca(on is (X2, Y2) and my speed is 101 mph.
My loca(on is (X1, Y1) and my speed is 106 mph.
GPS Uncertainty = 5 -‐10 m
Mobile and Vehicular Network Lab
Change the world (on the road)
• What we will talk about today:
0. Google driver-‐less car (autonomous vehicle) 1. Increase highway capacity (efficient)
2. Avoid scooter accidents (safe)
3. Business perspective (whether it will happen)
57
Mobile and Vehicular Network Lab
Why do we need a business model?
• You have this cool V2V radio that talks to other vehicle in your car
• But you are ALONE.
• What would happen?
• Nothing…
58
Mobile and Vehicular Network Lab
Why do we need a business model?
• 300K-‐500K new cars annually in Taiwan
• 6M car population
• 10% minimum market penetration rate is required for most V2V applications
• Even if all new cars start to equip V2V, it takes 2 years to reach 10%.
59
No customer wants to buy something that has to wait for another 2 years to function normally.
à No real products.
Mobile and Vehicular Network Lab
Potential Issues: Cooperative Systems
1. Bootstrapping: what is the required initial market penetration?
– If no one is using it, then it is useless
Ø No incentive for the user to purchase the new system
– Possible solution:
• The system is bundled as a "side-‐feature". You get the main feature you want (ex. DriverCam), but also get this "nice-‐to-‐have" feature additionally.
60
Mobile and Vehicular Network Lab
2. Control point:
– How to make sure only paid users can use the system/service?
Ø V2V: no obvious methods (need close systems)
Ø V2I: the gateway (base stations & RSUs) can control who can go through
3. Security:
will you trust your neighboring vehicles?
61
Potential Issues: Cooperative Systems
Mobile and Vehicular Network Lab
Possible Business Model
How to obtain the profit?
1. Bundled as part of the vehicle
– New feature, thus higher price
– Does the user want to pay for the additional cost?
2. Smartphone: deployed as a paid app.
– Obtain profit when the user purchases the app
3. Monthly subscription for the user:
– the user is buying a service
4. Subsidized by the government to mandate a new vehicular system (usually combined with 1.)
62
Mobile and Vehicular Network Lab
Case Study 1: Car Following
Mobile and Vehicular Network Lab
Case Study 1: Car Following
• Car Following (The SARTRE Project )
– Leader (human) drives for revenue.
– Followers (auto-‐driving) pay for enjoy the trip without human driving.
64
Leader | Dest.: Tainan
Follower | Dest.: Taichung
Leader | Dest.: Tainan
Taichung
EXIT Leave Group & Pay to Leader
Follower | Dest.: Tainan
Taichung EXIT
Mobile and Vehicular Network Lab
Post-‐Accident
• Insurance claim
– Investigation at the scene (agents and polices
• road-‐side surveillances
• costly and time consuming
– Lawsuit
– Bogus insurance claim
• £410m of motor insurance fraud detected in UK, 2009
Case Study 2: Networked DriverCam
Mobile and Vehicular Network Lab
How about an networking solution?
• V2V + V2I
– CSI in
vehicular networks
• Business model
– Incentive for insurance company
• A total package for customers
• Annual renewal of insurance policy (steady cash flow?)
Case Study 2: Networked DriverCam
Mobile and Vehicular Network Lab
• GM OnStar
– V2I: CDMA cellular network
– Stolen Vehicle Assistance, Roadside Assistance, Remote Door Unlock, Remote Horn and Light Flashing, Red Button
Emergency Services and OnStar Remote Vehicle Diagnostics – Subscribe Service:
• Safe & Sound: $18.95/mth ($24.95 in Canada)
67
Case Study 3: GM OnStar
[1] OnStar, https://www.onstar.com/web/portal/landing [2] OnStar, http://en.wikipedia.org/wiki/OnStar, Wiki
Mobile and Vehicular Network Lab
• GM OnStar
– Car Rental Service
68
Case Study 3: GM OnStar
[1] Telematies News, http://telematicsnews.info/2011/10/06/us-‐gm-‐onstar-‐and-‐relayrides-‐launch-‐car-‐sharing-‐program/
[2] automotive.com, http://blogs.automotive.com/gm-‐to-‐begin-‐onstar-‐carsharing-‐anyone-‐have-‐an-‐available-‐corvette-‐59543.html
Mobile and Vehicular Network Lab
Conclusion
• New technology: vehicular networks à
New opportunity: cooperative systems.
• Change our world on the road -‐ higher performance
– More Efficient: Greener, faster, more capacity – Safer: zero injuries & fatalities
– Affordable: each vehicle only needs to know a few things – New applications: remain to be unleashed
• Good research & business opportunities!
• Shameless Plug:
MVNL welcomes you.
69
Mobile and Vehicular Network Lab
Thank you!
Please feel free to contact me for questions:
Prof. Michael Tsai (蔡欣穆)
Mobile and Vehicular Network Lab hsinmu@csie.ntu.edu.tw
Dept. of Computer Science and Information Engineering National Taiwan University
70